1 @c Copyright (C) 1988-2017 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 * Stack and Calling:: Defining which way the stack grows and by how much.
36 * Varargs:: Defining the varargs macros.
37 * Trampolines:: Code set up at run time to enter a nested function.
38 * Library Calls:: Controlling how library routines are implicitly called.
39 * Addressing Modes:: Defining addressing modes valid for memory operands.
40 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
41 * Condition Code:: Defining how insns update the condition code.
42 * Costs:: Defining relative costs of different operations.
43 * Scheduling:: Adjusting the behavior of the instruction scheduler.
44 * Sections:: Dividing storage into text, data, and other sections.
45 * PIC:: Macros for position independent code.
46 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
47 * Debugging Info:: Defining the format of debugging output.
48 * Floating Point:: Handling floating point for cross-compilers.
49 * Mode Switching:: Insertion of mode-switching instructions.
50 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
51 * Emulated TLS:: Emulated TLS support.
52 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
53 * PCH Target:: Validity checking for precompiled headers.
54 * C++ ABI:: Controlling C++ ABI changes.
55 * Named Address Spaces:: Adding support for named address spaces
56 * Misc:: Everything else.
59 @node Target Structure
60 @section The Global @code{targetm} Variable
62 @cindex target functions
64 @deftypevar {struct gcc_target} targetm
65 The target @file{.c} file must define the global @code{targetm} variable
66 which contains pointers to functions and data relating to the target
67 machine. The variable is declared in @file{target.h};
68 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
69 used to initialize the variable, and macros for the default initializers
70 for elements of the structure. The @file{.c} file should override those
71 macros for which the default definition is inappropriate. For example:
74 #include "target-def.h"
76 /* @r{Initialize the GCC target structure.} */
78 #undef TARGET_COMP_TYPE_ATTRIBUTES
79 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
81 struct gcc_target targetm = TARGET_INITIALIZER;
85 Where a macro should be defined in the @file{.c} file in this manner to
86 form part of the @code{targetm} structure, it is documented below as a
87 ``Target Hook'' with a prototype. Many macros will change in future
88 from being defined in the @file{.h} file to being part of the
89 @code{targetm} structure.
91 Similarly, there is a @code{targetcm} variable for hooks that are
92 specific to front ends for C-family languages, documented as ``C
93 Target Hook''. This is declared in @file{c-family/c-target.h}, the
94 initializer @code{TARGETCM_INITIALIZER} in
95 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
96 themselves, they should set @code{target_has_targetcm=yes} in
97 @file{config.gcc}; otherwise a default definition is used.
99 Similarly, there is a @code{targetm_common} variable for hooks that
100 are shared between the compiler driver and the compilers proper,
101 documented as ``Common Target Hook''. This is declared in
102 @file{common/common-target.h}, the initializer
103 @code{TARGETM_COMMON_INITIALIZER} in
104 @file{common/common-target-def.h}. If targets initialize
105 @code{targetm_common} themselves, they should set
106 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
107 default definition is used.
110 @section Controlling the Compilation Driver, @file{gcc}
112 @cindex controlling the compilation driver
114 @c prevent bad page break with this line
115 You can control the compilation driver.
117 @defmac DRIVER_SELF_SPECS
118 A list of specs for the driver itself. It should be a suitable
119 initializer for an array of strings, with no surrounding braces.
121 The driver applies these specs to its own command line between loading
122 default @file{specs} files (but not command-line specified ones) and
123 choosing the multilib directory or running any subcommands. It
124 applies them in the order given, so each spec can depend on the
125 options added by earlier ones. It is also possible to remove options
126 using @samp{%<@var{option}} in the usual way.
128 This macro can be useful when a port has several interdependent target
129 options. It provides a way of standardizing the command line so
130 that the other specs are easier to write.
132 Do not define this macro if it does not need to do anything.
135 @defmac OPTION_DEFAULT_SPECS
136 A list of specs used to support configure-time default options (i.e.@:
137 @option{--with} options) in the driver. It should be a suitable initializer
138 for an array of structures, each containing two strings, without the
139 outermost pair of surrounding braces.
141 The first item in the pair is the name of the default. This must match
142 the code in @file{config.gcc} for the target. The second item is a spec
143 to apply if a default with this name was specified. The string
144 @samp{%(VALUE)} in the spec will be replaced by the value of the default
145 everywhere it occurs.
147 The driver will apply these specs to its own command line between loading
148 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
149 the same mechanism as @code{DRIVER_SELF_SPECS}.
151 Do not define this macro if it does not need to do anything.
155 A C string constant that tells the GCC driver program options to
156 pass to CPP@. It can also specify how to translate options you
157 give to GCC into options for GCC to pass to the CPP@.
159 Do not define this macro if it does not need to do anything.
162 @defmac CPLUSPLUS_CPP_SPEC
163 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
164 than C@. If you do not define this macro, then the value of
165 @code{CPP_SPEC} (if any) will be used instead.
169 A C string constant that tells the GCC driver program options to
170 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
172 It can also specify how to translate options you give to GCC into options
173 for GCC to pass to front ends.
175 Do not define this macro if it does not need to do anything.
179 A C string constant that tells the GCC driver program options to
180 pass to @code{cc1plus}. It can also specify how to translate options you
181 give to GCC into options for GCC to pass to the @code{cc1plus}.
183 Do not define this macro if it does not need to do anything.
184 Note that everything defined in CC1_SPEC is already passed to
185 @code{cc1plus} so there is no need to duplicate the contents of
186 CC1_SPEC in CC1PLUS_SPEC@.
190 A C string constant that tells the GCC driver program options to
191 pass to the assembler. It can also specify how to translate options
192 you give to GCC into options for GCC to pass to the assembler.
193 See the file @file{sun3.h} for an example of this.
195 Do not define this macro if it does not need to do anything.
198 @defmac ASM_FINAL_SPEC
199 A C string constant that tells the GCC driver program how to
200 run any programs which cleanup after the normal assembler.
201 Normally, this is not needed. See the file @file{mips.h} for
204 Do not define this macro if it does not need to do anything.
207 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
208 Define this macro, with no value, if the driver should give the assembler
209 an argument consisting of a single dash, @option{-}, to instruct it to
210 read from its standard input (which will be a pipe connected to the
211 output of the compiler proper). This argument is given after any
212 @option{-o} option specifying the name of the output file.
214 If you do not define this macro, the assembler is assumed to read its
215 standard input if given no non-option arguments. If your assembler
216 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
217 see @file{mips.h} for instance.
221 A C string constant that tells the GCC driver program options to
222 pass to the linker. It can also specify how to translate options you
223 give to GCC into options for GCC to pass to the linker.
225 Do not define this macro if it does not need to do anything.
229 Another C string constant used much like @code{LINK_SPEC}. The difference
230 between the two is that @code{LIB_SPEC} is used at the end of the
231 command given to the linker.
233 If this macro is not defined, a default is provided that
234 loads the standard C library from the usual place. See @file{gcc.c}.
238 Another C string constant that tells the GCC driver program
239 how and when to place a reference to @file{libgcc.a} into the
240 linker command line. This constant is placed both before and after
241 the value of @code{LIB_SPEC}.
243 If this macro is not defined, the GCC driver provides a default that
244 passes the string @option{-lgcc} to the linker.
247 @defmac REAL_LIBGCC_SPEC
248 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
249 @code{LIBGCC_SPEC} is not directly used by the driver program but is
250 instead modified to refer to different versions of @file{libgcc.a}
251 depending on the values of the command line flags @option{-static},
252 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
253 targets where these modifications are inappropriate, define
254 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
255 driver how to place a reference to @file{libgcc} on the link command
256 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
259 @defmac USE_LD_AS_NEEDED
260 A macro that controls the modifications to @code{LIBGCC_SPEC}
261 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
262 generated that uses @option{--as-needed} or equivalent options and the
263 shared @file{libgcc} in place of the
264 static exception handler library, when linking without any of
265 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
269 If defined, this C string constant is added to @code{LINK_SPEC}.
270 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
271 the modifications to @code{LIBGCC_SPEC} mentioned in
272 @code{REAL_LIBGCC_SPEC}.
275 @defmac STARTFILE_SPEC
276 Another C string constant used much like @code{LINK_SPEC}. The
277 difference between the two is that @code{STARTFILE_SPEC} is used at
278 the very beginning of the command given to the linker.
280 If this macro is not defined, a default is provided that loads the
281 standard C startup file from the usual place. See @file{gcc.c}.
285 Another C string constant used much like @code{LINK_SPEC}. The
286 difference between the two is that @code{ENDFILE_SPEC} is used at
287 the very end of the command given to the linker.
289 Do not define this macro if it does not need to do anything.
292 @defmac THREAD_MODEL_SPEC
293 GCC @code{-v} will print the thread model GCC was configured to use.
294 However, this doesn't work on platforms that are multilibbed on thread
295 models, such as AIX 4.3. On such platforms, define
296 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
297 blanks that names one of the recognized thread models. @code{%*}, the
298 default value of this macro, will expand to the value of
299 @code{thread_file} set in @file{config.gcc}.
302 @defmac SYSROOT_SUFFIX_SPEC
303 Define this macro to add a suffix to the target sysroot when GCC is
304 configured with a sysroot. This will cause GCC to search for usr/lib,
305 et al, within sysroot+suffix.
308 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
309 Define this macro to add a headers_suffix to the target sysroot when
310 GCC is configured with a sysroot. This will cause GCC to pass the
311 updated sysroot+headers_suffix to CPP, causing it to search for
312 usr/include, et al, within sysroot+headers_suffix.
316 Define this macro to provide additional specifications to put in the
317 @file{specs} file that can be used in various specifications like
320 The definition should be an initializer for an array of structures,
321 containing a string constant, that defines the specification name, and a
322 string constant that provides the specification.
324 Do not define this macro if it does not need to do anything.
326 @code{EXTRA_SPECS} is useful when an architecture contains several
327 related targets, which have various @code{@dots{}_SPECS} which are similar
328 to each other, and the maintainer would like one central place to keep
331 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
332 define either @code{_CALL_SYSV} when the System V calling sequence is
333 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
336 The @file{config/rs6000/rs6000.h} target file defines:
339 #define EXTRA_SPECS \
340 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
342 #define CPP_SYS_DEFAULT ""
345 The @file{config/rs6000/sysv.h} target file defines:
349 "%@{posix: -D_POSIX_SOURCE @} \
350 %@{mcall-sysv: -D_CALL_SYSV @} \
351 %@{!mcall-sysv: %(cpp_sysv_default) @} \
352 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
354 #undef CPP_SYSV_DEFAULT
355 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
358 while the @file{config/rs6000/eabiaix.h} target file defines
359 @code{CPP_SYSV_DEFAULT} as:
362 #undef CPP_SYSV_DEFAULT
363 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
367 @defmac LINK_LIBGCC_SPECIAL_1
368 Define this macro if the driver program should find the library
369 @file{libgcc.a}. If you do not define this macro, the driver program will pass
370 the argument @option{-lgcc} to tell the linker to do the search.
373 @defmac LINK_GCC_C_SEQUENCE_SPEC
374 The sequence in which libgcc and libc are specified to the linker.
375 By default this is @code{%G %L %G}.
378 @defmac POST_LINK_SPEC
379 Define this macro to add additional steps to be executed after linker.
380 The default value of this macro is empty string.
383 @defmac LINK_COMMAND_SPEC
384 A C string constant giving the complete command line need to execute the
385 linker. When you do this, you will need to update your port each time a
386 change is made to the link command line within @file{gcc.c}. Therefore,
387 define this macro only if you need to completely redefine the command
388 line for invoking the linker and there is no other way to accomplish
389 the effect you need. Overriding this macro may be avoidable by overriding
390 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
393 @deftypevr {Common Target Hook} bool TARGET_ALWAYS_STRIP_DOTDOT
394 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.
397 @defmac MULTILIB_DEFAULTS
398 Define this macro as a C expression for the initializer of an array of
399 string to tell the driver program which options are defaults for this
400 target and thus do not need to be handled specially when using
401 @code{MULTILIB_OPTIONS}.
403 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
404 the target makefile fragment or if none of the options listed in
405 @code{MULTILIB_OPTIONS} are set by default.
406 @xref{Target Fragment}.
409 @defmac RELATIVE_PREFIX_NOT_LINKDIR
410 Define this macro to tell @command{gcc} that it should only translate
411 a @option{-B} prefix into a @option{-L} linker option if the prefix
412 indicates an absolute file name.
415 @defmac MD_EXEC_PREFIX
416 If defined, this macro is an additional prefix to try after
417 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
418 when the compiler is built as a cross
419 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
420 to the list of directories used to find the assembler in @file{configure.ac}.
423 @defmac STANDARD_STARTFILE_PREFIX
424 Define this macro as a C string constant if you wish to override the
425 standard choice of @code{libdir} as the default prefix to
426 try when searching for startup files such as @file{crt0.o}.
427 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
428 is built as a cross compiler.
431 @defmac STANDARD_STARTFILE_PREFIX_1
432 Define this macro as a C string constant if you wish to override the
433 standard choice of @code{/lib} as a prefix to try after the default prefix
434 when searching for startup files such as @file{crt0.o}.
435 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
436 is built as a cross compiler.
439 @defmac STANDARD_STARTFILE_PREFIX_2
440 Define this macro as a C string constant if you wish to override the
441 standard choice of @code{/lib} as yet another prefix to try after the
442 default prefix when searching for startup files such as @file{crt0.o}.
443 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
444 is built as a cross compiler.
447 @defmac MD_STARTFILE_PREFIX
448 If defined, this macro supplies an additional prefix to try after the
449 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
450 compiler is built as a cross compiler.
453 @defmac MD_STARTFILE_PREFIX_1
454 If defined, this macro supplies yet another prefix to try after the
455 standard prefixes. It is not searched when the compiler is built as a
459 @defmac INIT_ENVIRONMENT
460 Define this macro as a C string constant if you wish to set environment
461 variables for programs called by the driver, such as the assembler and
462 loader. The driver passes the value of this macro to @code{putenv} to
463 initialize the necessary environment variables.
466 @defmac LOCAL_INCLUDE_DIR
467 Define this macro as a C string constant if you wish to override the
468 standard choice of @file{/usr/local/include} as the default prefix to
469 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
470 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
471 @file{config.gcc}, normally @file{/usr/include}) in the search order.
473 Cross compilers do not search either @file{/usr/local/include} or its
477 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
478 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
479 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
480 If you do not define this macro, no component is used.
483 @defmac INCLUDE_DEFAULTS
484 Define this macro if you wish to override the entire default search path
485 for include files. For a native compiler, the default search path
486 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
487 @code{GPLUSPLUS_INCLUDE_DIR}, and
488 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
489 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
490 and specify private search areas for GCC@. The directory
491 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
493 The definition should be an initializer for an array of structures.
494 Each array element should have four elements: the directory name (a
495 string constant), the component name (also a string constant), a flag
496 for C++-only directories,
497 and a flag showing that the includes in the directory don't need to be
498 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
499 the array with a null element.
501 The component name denotes what GNU package the include file is part of,
502 if any, in all uppercase letters. For example, it might be @samp{GCC}
503 or @samp{BINUTILS}. If the package is part of a vendor-supplied
504 operating system, code the component name as @samp{0}.
506 For example, here is the definition used for VAX/VMS:
509 #define INCLUDE_DEFAULTS \
511 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
512 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
513 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
520 Here is the order of prefixes tried for exec files:
524 Any prefixes specified by the user with @option{-B}.
527 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
528 is not set and the compiler has not been installed in the configure-time
529 @var{prefix}, the location in which the compiler has actually been installed.
532 The directories specified by the environment variable @code{COMPILER_PATH}.
535 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
536 in the configured-time @var{prefix}.
539 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
542 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
545 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
549 Here is the order of prefixes tried for startfiles:
553 Any prefixes specified by the user with @option{-B}.
556 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
557 value based on the installed toolchain location.
560 The directories specified by the environment variable @code{LIBRARY_PATH}
561 (or port-specific name; native only, cross compilers do not use this).
564 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
565 in the configured @var{prefix} or this is a native compiler.
568 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
571 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
575 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
576 native compiler, or we have a target system root.
579 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
580 native compiler, or we have a target system root.
583 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
584 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
585 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
588 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
589 compiler, or we have a target system root. The default for this macro is
593 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
594 compiler, or we have a target system root. The default for this macro is
598 @node Run-time Target
599 @section Run-time Target Specification
600 @cindex run-time target specification
601 @cindex predefined macros
602 @cindex target specifications
604 @c prevent bad page break with this line
605 Here are run-time target specifications.
607 @defmac TARGET_CPU_CPP_BUILTINS ()
608 This function-like macro expands to a block of code that defines
609 built-in preprocessor macros and assertions for the target CPU, using
610 the functions @code{builtin_define}, @code{builtin_define_std} and
611 @code{builtin_assert}. When the front end
612 calls this macro it provides a trailing semicolon, and since it has
613 finished command line option processing your code can use those
616 @code{builtin_assert} takes a string in the form you pass to the
617 command-line option @option{-A}, such as @code{cpu=mips}, and creates
618 the assertion. @code{builtin_define} takes a string in the form
619 accepted by option @option{-D} and unconditionally defines the macro.
621 @code{builtin_define_std} takes a string representing the name of an
622 object-like macro. If it doesn't lie in the user's namespace,
623 @code{builtin_define_std} defines it unconditionally. Otherwise, it
624 defines a version with two leading underscores, and another version
625 with two leading and trailing underscores, and defines the original
626 only if an ISO standard was not requested on the command line. For
627 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
628 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
629 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
630 defines only @code{_ABI64}.
632 You can also test for the C dialect being compiled. The variable
633 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
634 or @code{clk_objective_c}. Note that if we are preprocessing
635 assembler, this variable will be @code{clk_c} but the function-like
636 macro @code{preprocessing_asm_p()} will return true, so you might want
637 to check for that first. If you need to check for strict ANSI, the
638 variable @code{flag_iso} can be used. The function-like macro
639 @code{preprocessing_trad_p()} can be used to check for traditional
643 @defmac TARGET_OS_CPP_BUILTINS ()
644 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
645 and is used for the target operating system instead.
648 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
649 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
650 and is used for the target object format. @file{elfos.h} uses this
651 macro to define @code{__ELF__}, so you probably do not need to define
655 @deftypevar {extern int} target_flags
656 This variable is declared in @file{options.h}, which is included before
657 any target-specific headers.
660 @deftypevr {Common Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
661 This variable specifies the initial value of @code{target_flags}.
662 Its default setting is 0.
665 @cindex optional hardware or system features
666 @cindex features, optional, in system conventions
668 @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})
669 This hook is called whenever the user specifies one of the
670 target-specific options described by the @file{.opt} definition files
671 (@pxref{Options}). It has the opportunity to do some option-specific
672 processing and should return true if the option is valid. The default
673 definition does nothing but return true.
675 @var{decoded} specifies the option and its arguments. @var{opts} and
676 @var{opts_set} are the @code{gcc_options} structures to be used for
677 storing option state, and @var{loc} is the location at which the
678 option was passed (@code{UNKNOWN_LOCATION} except for options passed
682 @deftypefn {C Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
683 This target hook is called whenever the user specifies one of the
684 target-specific C language family options described by the @file{.opt}
685 definition files(@pxref{Options}). It has the opportunity to do some
686 option-specific processing and should return true if the option is
687 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
688 default definition does nothing but return false.
690 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
691 options. However, if processing an option requires routines that are
692 only available in the C (and related language) front ends, then you
693 should use @code{TARGET_HANDLE_C_OPTION} instead.
696 @deftypefn {C Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING_OBJECT (tree @var{string})
697 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.
700 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE (const char *@var{classname})
701 Declare that Objective C class @var{classname} is referenced by the current TU.
704 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_CLASS_DEFINITION (const char *@var{classname})
705 Declare that Objective C class @var{classname} is defined by the current TU.
708 @deftypefn {C Target Hook} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref})
709 If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object.
712 @deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list})
713 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.
716 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
717 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
718 but is called when the optimize level is changed via an attribute or
719 pragma or when it is reset at the end of the code affected by the
720 attribute or pragma. It is not called at the beginning of compilation
721 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
722 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
723 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
726 @defmac C_COMMON_OVERRIDE_OPTIONS
727 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
728 but is only used in the C
729 language frontends (C, Objective-C, C++, Objective-C++) and so can be
730 used to alter option flag variables which only exist in those
734 @deftypevr {Common Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
735 Some machines may desire to change what optimizations are performed for
736 various optimization levels. This variable, if defined, describes
737 options to enable at particular sets of optimization levels. These
738 options are processed once
739 just after the optimization level is determined and before the remainder
740 of the command options have been parsed, so may be overridden by other
741 options passed explicitly.
743 This processing is run once at program startup and when the optimization
744 options are changed via @code{#pragma GCC optimize} or by using the
745 @code{optimize} attribute.
748 @deftypefn {Common Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
749 Set target-dependent initial values of fields in @var{opts}.
752 @deftypefn {Common Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
753 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
756 @defmac SWITCHABLE_TARGET
757 Some targets need to switch between substantially different subtargets
758 during compilation. For example, the MIPS target has one subtarget for
759 the traditional MIPS architecture and another for MIPS16. Source code
760 can switch between these two subarchitectures using the @code{mips16}
761 and @code{nomips16} attributes.
763 Such subtargets can differ in things like the set of available
764 registers, the set of available instructions, the costs of various
765 operations, and so on. GCC caches a lot of this type of information
766 in global variables, and recomputing them for each subtarget takes a
767 significant amount of time. The compiler therefore provides a facility
768 for maintaining several versions of the global variables and quickly
769 switching between them; see @file{target-globals.h} for details.
771 Define this macro to 1 if your target needs this facility. The default
775 @deftypefn {Target Hook} bool TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P (void)
776 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.
779 @node Per-Function Data
780 @section Defining data structures for per-function information.
781 @cindex per-function data
782 @cindex data structures
784 If the target needs to store information on a per-function basis, GCC
785 provides a macro and a couple of variables to allow this. Note, just
786 using statics to store the information is a bad idea, since GCC supports
787 nested functions, so you can be halfway through encoding one function
788 when another one comes along.
790 GCC defines a data structure called @code{struct function} which
791 contains all of the data specific to an individual function. This
792 structure contains a field called @code{machine} whose type is
793 @code{struct machine_function *}, which can be used by targets to point
794 to their own specific data.
796 If a target needs per-function specific data it should define the type
797 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
798 This macro should be used to initialize the function pointer
799 @code{init_machine_status}. This pointer is explained below.
801 One typical use of per-function, target specific data is to create an
802 RTX to hold the register containing the function's return address. This
803 RTX can then be used to implement the @code{__builtin_return_address}
804 function, for level 0.
806 Note---earlier implementations of GCC used a single data area to hold
807 all of the per-function information. Thus when processing of a nested
808 function began the old per-function data had to be pushed onto a
809 stack, and when the processing was finished, it had to be popped off the
810 stack. GCC used to provide function pointers called
811 @code{save_machine_status} and @code{restore_machine_status} to handle
812 the saving and restoring of the target specific information. Since the
813 single data area approach is no longer used, these pointers are no
816 @defmac INIT_EXPANDERS
817 Macro called to initialize any target specific information. This macro
818 is called once per function, before generation of any RTL has begun.
819 The intention of this macro is to allow the initialization of the
820 function pointer @code{init_machine_status}.
823 @deftypevar {void (*)(struct function *)} init_machine_status
824 If this function pointer is non-@code{NULL} it will be called once per
825 function, before function compilation starts, in order to allow the
826 target to perform any target specific initialization of the
827 @code{struct function} structure. It is intended that this would be
828 used to initialize the @code{machine} of that structure.
830 @code{struct machine_function} structures are expected to be freed by GC@.
831 Generally, any memory that they reference must be allocated by using
832 GC allocation, including the structure itself.
836 @section Storage Layout
837 @cindex storage layout
839 Note that the definitions of the macros in this table which are sizes or
840 alignments measured in bits do not need to be constant. They can be C
841 expressions that refer to static variables, such as the @code{target_flags}.
842 @xref{Run-time Target}.
844 @defmac BITS_BIG_ENDIAN
845 Define this macro to have the value 1 if the most significant bit in a
846 byte has the lowest number; otherwise define it to have the value zero.
847 This means that bit-field instructions count from the most significant
848 bit. If the machine has no bit-field instructions, then this must still
849 be defined, but it doesn't matter which value it is defined to. This
850 macro need not be a constant.
852 This macro does not affect the way structure fields are packed into
853 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
856 @defmac BYTES_BIG_ENDIAN
857 Define this macro to have the value 1 if the most significant byte in a
858 word has the lowest number. This macro need not be a constant.
861 @defmac WORDS_BIG_ENDIAN
862 Define this macro to have the value 1 if, in a multiword object, the
863 most significant word has the lowest number. This applies to both
864 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
865 order of words in memory is not the same as the order in registers. This
866 macro need not be a constant.
869 @defmac REG_WORDS_BIG_ENDIAN
870 On some machines, the order of words in a multiword object differs between
871 registers in memory. In such a situation, define this macro to describe
872 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
873 the order of words in memory.
876 @defmac FLOAT_WORDS_BIG_ENDIAN
877 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
878 @code{TFmode} floating point numbers are stored in memory with the word
879 containing the sign bit at the lowest address; otherwise define it to
880 have the value 0. This macro need not be a constant.
882 You need not define this macro if the ordering is the same as for
886 @defmac BITS_PER_WORD
887 Number of bits in a word. If you do not define this macro, the default
888 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
891 @defmac MAX_BITS_PER_WORD
892 Maximum number of bits in a word. If this is undefined, the default is
893 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
894 largest value that @code{BITS_PER_WORD} can have at run-time.
897 @defmac UNITS_PER_WORD
898 Number of storage units in a word; normally the size of a general-purpose
899 register, a power of two from 1 or 8.
902 @defmac MIN_UNITS_PER_WORD
903 Minimum number of units in a word. If this is undefined, the default is
904 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
905 smallest value that @code{UNITS_PER_WORD} can have at run-time.
909 Width of a pointer, in bits. You must specify a value no wider than the
910 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
911 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
912 a value the default is @code{BITS_PER_WORD}.
915 @defmac POINTERS_EXTEND_UNSIGNED
916 A C expression that determines how pointers should be extended from
917 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
918 greater than zero if pointers should be zero-extended, zero if they
919 should be sign-extended, and negative if some other sort of conversion
920 is needed. In the last case, the extension is done by the target's
921 @code{ptr_extend} instruction.
923 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
924 and @code{word_mode} are all the same width.
927 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
928 A macro to update @var{m} and @var{unsignedp} when an object whose type
929 is @var{type} and which has the specified mode and signedness is to be
930 stored in a register. This macro is only called when @var{type} is a
933 On most RISC machines, which only have operations that operate on a full
934 register, define this macro to set @var{m} to @code{word_mode} if
935 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
936 cases, only integer modes should be widened because wider-precision
937 floating-point operations are usually more expensive than their narrower
940 For most machines, the macro definition does not change @var{unsignedp}.
941 However, some machines, have instructions that preferentially handle
942 either signed or unsigned quantities of certain modes. For example, on
943 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
944 sign-extend the result to 64 bits. On such machines, set
945 @var{unsignedp} according to which kind of extension is more efficient.
947 Do not define this macro if it would never modify @var{m}.
950 @deftypefn {Target Hook} {enum flt_eval_method} TARGET_C_EXCESS_PRECISION (enum excess_precision_type @var{type})
951 Return a value, with the same meaning as the C99 macro @code{FLT_EVAL_METHOD} that describes which excess precision should be applied. @var{type} is either @code{EXCESS_PRECISION_TYPE_IMPLICIT}, @code{EXCESS_PRECISION_TYPE_FAST}, or @code{EXCESS_PRECISION_TYPE_STANDARD}. For @code{EXCESS_PRECISION_TYPE_IMPLICIT}, the target should return which precision and range operations will be implictly evaluated in regardless of the excess precision explicitly added. For @code{EXCESS_PRECISION_TYPE_STANDARD} and @code{EXCESS_PRECISION_TYPE_FAST}, the target should return the explicit excess precision that should be added depending on the value set for @option{-fexcess-precision=@r{[}standard@r{|}fast@r{]}}. Note that unpredictable explicit excess precision does not make sense, so a target should never return @code{FLT_EVAL_METHOD_UNPREDICTABLE} when @var{type} is @code{EXCESS_PRECISION_TYPE_STANDARD} or @code{EXCESS_PRECISION_TYPE_FAST}.
954 @deftypefn {Target Hook} machine_mode TARGET_PROMOTE_FUNCTION_MODE (const_tree @var{type}, machine_mode @var{mode}, int *@var{punsignedp}, const_tree @var{funtype}, int @var{for_return})
955 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
956 function return values. The target hook should return the new mode
957 and possibly change @code{*@var{punsignedp}} if the promotion should
958 change signedness. This function is called only for scalar @emph{or
961 @var{for_return} allows to distinguish the promotion of arguments and
962 return values. If it is @code{1}, a return value is being promoted and
963 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
964 If it is @code{2}, the returned mode should be that of the register in
965 which an incoming parameter is copied, or the outgoing result is computed;
966 then the hook should return the same mode as @code{promote_mode}, though
967 the signedness may be different.
969 @var{type} can be NULL when promoting function arguments of libcalls.
971 The default is to not promote arguments and return values. You can
972 also define the hook to @code{default_promote_function_mode_always_promote}
973 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
976 @defmac PARM_BOUNDARY
977 Normal alignment required for function parameters on the stack, in
978 bits. All stack parameters receive at least this much alignment
979 regardless of data type. On most machines, this is the same as the
983 @defmac STACK_BOUNDARY
984 Define this macro to the minimum alignment enforced by hardware for the
985 stack pointer on this machine. The definition is a C expression for the
986 desired alignment (measured in bits). This value is used as a default
987 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
988 this should be the same as @code{PARM_BOUNDARY}.
991 @defmac PREFERRED_STACK_BOUNDARY
992 Define this macro if you wish to preserve a certain alignment for the
993 stack pointer, greater than what the hardware enforces. The definition
994 is a C expression for the desired alignment (measured in bits). This
995 macro must evaluate to a value equal to or larger than
996 @code{STACK_BOUNDARY}.
999 @defmac INCOMING_STACK_BOUNDARY
1000 Define this macro if the incoming stack boundary may be different
1001 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
1002 to a value equal to or larger than @code{STACK_BOUNDARY}.
1005 @defmac FUNCTION_BOUNDARY
1006 Alignment required for a function entry point, in bits.
1009 @defmac BIGGEST_ALIGNMENT
1010 Biggest alignment that any data type can require on this machine, in
1011 bits. Note that this is not the biggest alignment that is supported,
1012 just the biggest alignment that, when violated, may cause a fault.
1015 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_ABSOLUTE_BIGGEST_ALIGNMENT
1016 If defined, this target hook specifies the absolute biggest alignment
1017 that a type or variable can have on this machine, otherwise,
1018 @code{BIGGEST_ALIGNMENT} is used.
1021 @defmac MALLOC_ABI_ALIGNMENT
1022 Alignment, in bits, a C conformant malloc implementation has to
1023 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1026 @defmac ATTRIBUTE_ALIGNED_VALUE
1027 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1028 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1031 @defmac MINIMUM_ATOMIC_ALIGNMENT
1032 If defined, the smallest alignment, in bits, that can be given to an
1033 object that can be referenced in one operation, without disturbing any
1034 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1035 on machines that don't have byte or half-word store operations.
1038 @defmac BIGGEST_FIELD_ALIGNMENT
1039 Biggest alignment that any structure or union field can require on this
1040 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1041 structure and union fields only, unless the field alignment has been set
1042 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1045 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{type}, @var{computed})
1046 An expression for the alignment of a structure field @var{field} of
1047 type @var{type} if the alignment computed in the usual way (including
1048 applying of @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1049 alignment) is @var{computed}. It overrides alignment only if the
1050 field alignment has not been set by the
1051 @code{__attribute__ ((aligned (@var{n})))} construct. Note that @var{field}
1052 may be @code{NULL_TREE} in case we just query for the minimum alignment
1053 of a field of type @var{type} in structure context.
1056 @defmac MAX_STACK_ALIGNMENT
1057 Biggest stack alignment guaranteed by the backend. Use this macro
1058 to specify the maximum alignment of a variable on stack.
1060 If not defined, the default value is @code{STACK_BOUNDARY}.
1062 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1063 @c But the fix for PR 32893 indicates that we can only guarantee
1064 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1065 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1068 @defmac MAX_OFILE_ALIGNMENT
1069 Biggest alignment supported by the object file format of this machine.
1070 Use this macro to limit the alignment which can be specified using the
1071 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1072 the default value is @code{BIGGEST_ALIGNMENT}.
1074 On systems that use ELF, the default (in @file{config/elfos.h}) is
1075 the largest supported 32-bit ELF section alignment representable on
1076 a 32-bit host e.g. @samp{(((uint64_t) 1 << 28) * 8)}.
1077 On 32-bit ELF the largest supported section alignment in bits is
1078 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1081 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1082 If defined, a C expression to compute the alignment for a variable in
1083 the static store. @var{type} is the data type, and @var{basic-align} is
1084 the alignment that the object would ordinarily have. The value of this
1085 macro is used instead of that alignment to align the object.
1087 If this macro is not defined, then @var{basic-align} is used.
1090 One use of this macro is to increase alignment of medium-size data to
1091 make it all fit in fewer cache lines. Another is to cause character
1092 arrays to be word-aligned so that @code{strcpy} calls that copy
1093 constants to character arrays can be done inline.
1096 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1097 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1098 some alignment increase, instead of optimization only purposes. E.g.@
1099 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1100 must be aligned to 16 byte boundaries.
1102 If this macro is not defined, then @var{basic-align} is used.
1105 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1106 If defined, a C expression to compute the alignment given to a constant
1107 that is being placed in memory. @var{constant} is the constant and
1108 @var{basic-align} is the alignment that the object would ordinarily
1109 have. The value of this macro is used instead of that alignment to
1112 The default definition just returns @var{basic-align}.
1114 The typical use of this macro is to increase alignment for string
1115 constants to be word aligned so that @code{strcpy} calls that copy
1116 constants can be done inline.
1119 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1120 If defined, a C expression to compute the alignment for a variable in
1121 the local store. @var{type} is the data type, and @var{basic-align} is
1122 the alignment that the object would ordinarily have. The value of this
1123 macro is used instead of that alignment to align the object.
1125 If this macro is not defined, then @var{basic-align} is used.
1127 One use of this macro is to increase alignment of medium-size data to
1128 make it all fit in fewer cache lines.
1130 If the value of this macro has a type, it should be an unsigned type.
1133 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_VECTOR_ALIGNMENT (const_tree @var{type})
1134 This hook can be used to define the alignment for a vector of type
1135 @var{type}, in order to comply with a platform ABI. The default is to
1136 require natural alignment for vector types. The alignment returned by
1137 this hook must be a power-of-two multiple of the default alignment of
1138 the vector element type.
1141 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1142 If defined, a C expression to compute the alignment for stack slot.
1143 @var{type} is the data type, @var{mode} is the widest mode available,
1144 and @var{basic-align} is the alignment that the slot would ordinarily
1145 have. The value of this macro is used instead of that alignment to
1148 If this macro is not defined, then @var{basic-align} is used when
1149 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1152 This macro is to set alignment of stack slot to the maximum alignment
1153 of all possible modes which the slot may have.
1155 If the value of this macro has a type, it should be an unsigned type.
1158 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1159 If defined, a C expression to compute the alignment for a local
1160 variable @var{decl}.
1162 If this macro is not defined, then
1163 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1166 One use of this macro is to increase alignment of medium-size data to
1167 make it all fit in fewer cache lines.
1169 If the value of this macro has a type, it should be an unsigned type.
1172 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1173 If defined, a C expression to compute the minimum required alignment
1174 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1175 @var{mode}, assuming normal alignment @var{align}.
1177 If this macro is not defined, then @var{align} will be used.
1180 @defmac EMPTY_FIELD_BOUNDARY
1181 Alignment in bits to be given to a structure bit-field that follows an
1182 empty field such as @code{int : 0;}.
1184 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1187 @defmac STRUCTURE_SIZE_BOUNDARY
1188 Number of bits which any structure or union's size must be a multiple of.
1189 Each structure or union's size is rounded up to a multiple of this.
1191 If you do not define this macro, the default is the same as
1192 @code{BITS_PER_UNIT}.
1195 @defmac STRICT_ALIGNMENT
1196 Define this macro to be the value 1 if instructions will fail to work
1197 if given data not on the nominal alignment. If instructions will merely
1198 go slower in that case, define this macro as 0.
1201 @defmac PCC_BITFIELD_TYPE_MATTERS
1202 Define this if you wish to imitate the way many other C compilers handle
1203 alignment of bit-fields and the structures that contain them.
1205 The behavior is that the type written for a named bit-field (@code{int},
1206 @code{short}, or other integer type) imposes an alignment for the entire
1207 structure, as if the structure really did contain an ordinary field of
1208 that type. In addition, the bit-field is placed within the structure so
1209 that it would fit within such a field, not crossing a boundary for it.
1211 Thus, on most machines, a named bit-field whose type is written as
1212 @code{int} would not cross a four-byte boundary, and would force
1213 four-byte alignment for the whole structure. (The alignment used may
1214 not be four bytes; it is controlled by the other alignment parameters.)
1216 An unnamed bit-field will not affect the alignment of the containing
1219 If the macro is defined, its definition should be a C expression;
1220 a nonzero value for the expression enables this behavior.
1222 Note that if this macro is not defined, or its value is zero, some
1223 bit-fields may cross more than one alignment boundary. The compiler can
1224 support such references if there are @samp{insv}, @samp{extv}, and
1225 @samp{extzv} insns that can directly reference memory.
1227 The other known way of making bit-fields work is to define
1228 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1229 Then every structure can be accessed with fullwords.
1231 Unless the machine has bit-field instructions or you define
1232 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1233 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1235 If your aim is to make GCC use the same conventions for laying out
1236 bit-fields as are used by another compiler, here is how to investigate
1237 what the other compiler does. Compile and run this program:
1256 printf ("Size of foo1 is %d\n",
1257 sizeof (struct foo1));
1258 printf ("Size of foo2 is %d\n",
1259 sizeof (struct foo2));
1264 If this prints 2 and 5, then the compiler's behavior is what you would
1265 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1268 @defmac BITFIELD_NBYTES_LIMITED
1269 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1270 to aligning a bit-field within the structure.
1273 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1274 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1275 whether unnamed bitfields affect the alignment of the containing
1276 structure. The hook should return true if the structure should inherit
1277 the alignment requirements of an unnamed bitfield's type.
1280 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1281 This target hook should return @code{true} if accesses to volatile bitfields
1282 should use the narrowest mode possible. It should return @code{false} if
1283 these accesses should use the bitfield container type.
1285 The default is @code{false}.
1288 @deftypefn {Target Hook} bool TARGET_MEMBER_TYPE_FORCES_BLK (const_tree @var{field}, machine_mode @var{mode})
1289 Return true if a structure, union or array containing @var{field} should
1290 be accessed using @code{BLKMODE}.
1292 If @var{field} is the only field in the structure, @var{mode} is its
1293 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1294 case where structures of one field would require the structure's mode to
1295 retain the field's mode.
1297 Normally, this is not needed.
1300 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1301 Define this macro as an expression for the alignment of a type (given
1302 by @var{type} as a tree node) if the alignment computed in the usual
1303 way is @var{computed} and the alignment explicitly specified was
1306 The default is to use @var{specified} if it is larger; otherwise, use
1307 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1310 @defmac MAX_FIXED_MODE_SIZE
1311 An integer expression for the size in bits of the largest integer
1312 machine mode that should actually be used. All integer machine modes of
1313 this size or smaller can be used for structures and unions with the
1314 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1315 (DImode)} is assumed.
1318 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1319 If defined, an expression of type @code{machine_mode} that
1320 specifies the mode of the save area operand of a
1321 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1322 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1323 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1324 having its mode specified.
1326 You need not define this macro if it always returns @code{Pmode}. You
1327 would most commonly define this macro if the
1328 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1332 @defmac STACK_SIZE_MODE
1333 If defined, an expression of type @code{machine_mode} that
1334 specifies the mode of the size increment operand of an
1335 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1337 You need not define this macro if it always returns @code{word_mode}.
1338 You would most commonly define this macro if the @code{allocate_stack}
1339 pattern needs to support both a 32- and a 64-bit mode.
1342 @deftypefn {Target Hook} scalar_int_mode TARGET_LIBGCC_CMP_RETURN_MODE (void)
1343 This target hook should return the mode to be used for the return value
1344 of compare instructions expanded to libgcc calls. If not defined
1345 @code{word_mode} is returned which is the right choice for a majority of
1349 @deftypefn {Target Hook} scalar_int_mode TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1350 This target hook should return the mode to be used for the shift count operand
1351 of shift instructions expanded to libgcc calls. If not defined
1352 @code{word_mode} is returned which is the right choice for a majority of
1356 @deftypefn {Target Hook} scalar_int_mode TARGET_UNWIND_WORD_MODE (void)
1357 Return machine mode to be used for @code{_Unwind_Word} type.
1358 The default is to use @code{word_mode}.
1361 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1362 This target hook returns @code{true} if bit-fields in the given
1363 @var{record_type} are to be laid out following the rules of Microsoft
1364 Visual C/C++, namely: (i) a bit-field won't share the same storage
1365 unit with the previous bit-field if their underlying types have
1366 different sizes, and the bit-field will be aligned to the highest
1367 alignment of the underlying types of itself and of the previous
1368 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1369 the whole enclosing structure, even if it is unnamed; except that
1370 (iii) a zero-sized bit-field will be disregarded unless it follows
1371 another bit-field of nonzero size. If this hook returns @code{true},
1372 other macros that control bit-field layout are ignored.
1374 When a bit-field is inserted into a packed record, the whole size
1375 of the underlying type is used by one or more same-size adjacent
1376 bit-fields (that is, if its long:3, 32 bits is used in the record,
1377 and any additional adjacent long bit-fields are packed into the same
1378 chunk of 32 bits. However, if the size changes, a new field of that
1379 size is allocated). In an unpacked record, this is the same as using
1380 alignment, but not equivalent when packing.
1382 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1383 the latter will take precedence. If @samp{__attribute__((packed))} is
1384 used on a single field when MS bit-fields are in use, it will take
1385 precedence for that field, but the alignment of the rest of the structure
1386 may affect its placement.
1389 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1390 Returns true if the target supports decimal floating point.
1393 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1394 Returns true if the target supports fixed-point arithmetic.
1397 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1398 This hook is called just before expansion into rtl, allowing the target
1399 to perform additional initializations or analysis before the expansion.
1400 For example, the rs6000 port uses it to allocate a scratch stack slot
1401 for use in copying SDmode values between memory and floating point
1402 registers whenever the function being expanded has any SDmode
1406 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1407 This hook allows the backend to perform additional instantiations on rtl
1408 that are not actually in any insns yet, but will be later.
1411 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1412 If your target defines any fundamental types, or any types your target
1413 uses should be mangled differently from the default, define this hook
1414 to return the appropriate encoding for these types as part of a C++
1415 mangled name. The @var{type} argument is the tree structure representing
1416 the type to be mangled. The hook may be applied to trees which are
1417 not target-specific fundamental types; it should return @code{NULL}
1418 for all such types, as well as arguments it does not recognize. If the
1419 return value is not @code{NULL}, it must point to a statically-allocated
1422 Target-specific fundamental types might be new fundamental types or
1423 qualified versions of ordinary fundamental types. Encode new
1424 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1425 is the name used for the type in source code, and @var{n} is the
1426 length of @var{name} in decimal. Encode qualified versions of
1427 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1428 @var{name} is the name used for the type qualifier in source code,
1429 @var{n} is the length of @var{name} as above, and @var{code} is the
1430 code used to represent the unqualified version of this type. (See
1431 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1432 codes.) In both cases the spaces are for clarity; do not include any
1433 spaces in your string.
1435 This hook is applied to types prior to typedef resolution. If the mangled
1436 name for a particular type depends only on that type's main variant, you
1437 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1440 The default version of this hook always returns @code{NULL}, which is
1441 appropriate for a target that does not define any new fundamental
1446 @section Layout of Source Language Data Types
1448 These macros define the sizes and other characteristics of the standard
1449 basic data types used in programs being compiled. Unlike the macros in
1450 the previous section, these apply to specific features of C and related
1451 languages, rather than to fundamental aspects of storage layout.
1453 @defmac INT_TYPE_SIZE
1454 A C expression for the size in bits of the type @code{int} on the
1455 target machine. If you don't define this, the default is one word.
1458 @defmac SHORT_TYPE_SIZE
1459 A C expression for the size in bits of the type @code{short} on the
1460 target machine. If you don't define this, the default is half a word.
1461 (If this would be less than one storage unit, it is rounded up to one
1465 @defmac LONG_TYPE_SIZE
1466 A C expression for the size in bits of the type @code{long} on the
1467 target machine. If you don't define this, the default is one word.
1470 @defmac ADA_LONG_TYPE_SIZE
1471 On some machines, the size used for the Ada equivalent of the type
1472 @code{long} by a native Ada compiler differs from that used by C@. In
1473 that situation, define this macro to be a C expression to be used for
1474 the size of that type. If you don't define this, the default is the
1475 value of @code{LONG_TYPE_SIZE}.
1478 @defmac LONG_LONG_TYPE_SIZE
1479 A C expression for the size in bits of the type @code{long long} on the
1480 target machine. If you don't define this, the default is two
1481 words. If you want to support GNU Ada on your machine, the value of this
1482 macro must be at least 64.
1485 @defmac CHAR_TYPE_SIZE
1486 A C expression for the size in bits of the type @code{char} on the
1487 target machine. If you don't define this, the default is
1488 @code{BITS_PER_UNIT}.
1491 @defmac BOOL_TYPE_SIZE
1492 A C expression for the size in bits of the C++ type @code{bool} and
1493 C99 type @code{_Bool} on the target machine. If you don't define
1494 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1497 @defmac FLOAT_TYPE_SIZE
1498 A C expression for the size in bits of the type @code{float} on the
1499 target machine. If you don't define this, the default is one word.
1502 @defmac DOUBLE_TYPE_SIZE
1503 A C expression for the size in bits of the type @code{double} on the
1504 target machine. If you don't define this, the default is two
1508 @defmac LONG_DOUBLE_TYPE_SIZE
1509 A C expression for the size in bits of the type @code{long double} on
1510 the target machine. If you don't define this, the default is two
1514 @defmac SHORT_FRACT_TYPE_SIZE
1515 A C expression for the size in bits of the type @code{short _Fract} on
1516 the target machine. If you don't define this, the default is
1517 @code{BITS_PER_UNIT}.
1520 @defmac FRACT_TYPE_SIZE
1521 A C expression for the size in bits of the type @code{_Fract} on
1522 the target machine. If you don't define this, the default is
1523 @code{BITS_PER_UNIT * 2}.
1526 @defmac LONG_FRACT_TYPE_SIZE
1527 A C expression for the size in bits of the type @code{long _Fract} on
1528 the target machine. If you don't define this, the default is
1529 @code{BITS_PER_UNIT * 4}.
1532 @defmac LONG_LONG_FRACT_TYPE_SIZE
1533 A C expression for the size in bits of the type @code{long long _Fract} on
1534 the target machine. If you don't define this, the default is
1535 @code{BITS_PER_UNIT * 8}.
1538 @defmac SHORT_ACCUM_TYPE_SIZE
1539 A C expression for the size in bits of the type @code{short _Accum} on
1540 the target machine. If you don't define this, the default is
1541 @code{BITS_PER_UNIT * 2}.
1544 @defmac ACCUM_TYPE_SIZE
1545 A C expression for the size in bits of the type @code{_Accum} on
1546 the target machine. If you don't define this, the default is
1547 @code{BITS_PER_UNIT * 4}.
1550 @defmac LONG_ACCUM_TYPE_SIZE
1551 A C expression for the size in bits of the type @code{long _Accum} on
1552 the target machine. If you don't define this, the default is
1553 @code{BITS_PER_UNIT * 8}.
1556 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1557 A C expression for the size in bits of the type @code{long long _Accum} on
1558 the target machine. If you don't define this, the default is
1559 @code{BITS_PER_UNIT * 16}.
1562 @defmac LIBGCC2_GNU_PREFIX
1563 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1564 hook and should be defined if that hook is overriden to be true. It
1565 causes function names in libgcc to be changed to use a @code{__gnu_}
1566 prefix for their name rather than the default @code{__}. A port which
1567 uses this macro should also arrange to use @file{t-gnu-prefix} in
1568 the libgcc @file{config.host}.
1571 @defmac WIDEST_HARDWARE_FP_SIZE
1572 A C expression for the size in bits of the widest floating-point format
1573 supported by the hardware. If you define this macro, you must specify a
1574 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1575 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1579 @defmac DEFAULT_SIGNED_CHAR
1580 An expression whose value is 1 or 0, according to whether the type
1581 @code{char} should be signed or unsigned by default. The user can
1582 always override this default with the options @option{-fsigned-char}
1583 and @option{-funsigned-char}.
1586 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1587 This target hook should return true if the compiler should give an
1588 @code{enum} type only as many bytes as it takes to represent the range
1589 of possible values of that type. It should return false if all
1590 @code{enum} types should be allocated like @code{int}.
1592 The default is to return false.
1596 A C expression for a string describing the name of the data type to use
1597 for size values. The typedef name @code{size_t} is defined using the
1598 contents of the string.
1600 The string can contain more than one keyword. If so, separate them with
1601 spaces, and write first any length keyword, then @code{unsigned} if
1602 appropriate, and finally @code{int}. The string must exactly match one
1603 of the data type names defined in the function
1604 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1605 You may not omit @code{int} or change the order---that would cause the
1606 compiler to crash on startup.
1608 If you don't define this macro, the default is @code{"long unsigned
1613 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1614 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1615 dealing with size. This macro is a C expression for a string describing
1616 the name of the data type from which the precision of @code{sizetype}
1619 The string has the same restrictions as @code{SIZE_TYPE} string.
1621 If you don't define this macro, the default is @code{SIZE_TYPE}.
1624 @defmac PTRDIFF_TYPE
1625 A C expression for a string describing the name of the data type to use
1626 for the result of subtracting two pointers. The typedef name
1627 @code{ptrdiff_t} is defined using the contents of the string. See
1628 @code{SIZE_TYPE} above for more information.
1630 If you don't define this macro, the default is @code{"long int"}.
1634 A C expression for a string describing the name of the data type to use
1635 for wide characters. The typedef name @code{wchar_t} is defined using
1636 the contents of the string. See @code{SIZE_TYPE} above for more
1639 If you don't define this macro, the default is @code{"int"}.
1642 @defmac WCHAR_TYPE_SIZE
1643 A C expression for the size in bits of the data type for wide
1644 characters. This is used in @code{cpp}, which cannot make use of
1649 A C expression for a string describing the name of the data type to
1650 use for wide characters passed to @code{printf} and returned from
1651 @code{getwc}. The typedef name @code{wint_t} is defined using the
1652 contents of the string. See @code{SIZE_TYPE} above for more
1655 If you don't define this macro, the default is @code{"unsigned int"}.
1659 A C expression for a string describing the name of the data type that
1660 can represent any value of any standard or extended signed integer type.
1661 The typedef name @code{intmax_t} is defined using the contents of the
1662 string. See @code{SIZE_TYPE} above for more information.
1664 If you don't define this macro, the default is the first of
1665 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1666 much precision as @code{long long int}.
1669 @defmac UINTMAX_TYPE
1670 A C expression for a string describing the name of the data type that
1671 can represent any value of any standard or extended unsigned integer
1672 type. The typedef name @code{uintmax_t} is defined using the contents
1673 of the string. See @code{SIZE_TYPE} above for more information.
1675 If you don't define this macro, the default is the first of
1676 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1677 unsigned int"} that has as much precision as @code{long long unsigned
1681 @defmac SIG_ATOMIC_TYPE
1687 @defmacx UINT16_TYPE
1688 @defmacx UINT32_TYPE
1689 @defmacx UINT64_TYPE
1690 @defmacx INT_LEAST8_TYPE
1691 @defmacx INT_LEAST16_TYPE
1692 @defmacx INT_LEAST32_TYPE
1693 @defmacx INT_LEAST64_TYPE
1694 @defmacx UINT_LEAST8_TYPE
1695 @defmacx UINT_LEAST16_TYPE
1696 @defmacx UINT_LEAST32_TYPE
1697 @defmacx UINT_LEAST64_TYPE
1698 @defmacx INT_FAST8_TYPE
1699 @defmacx INT_FAST16_TYPE
1700 @defmacx INT_FAST32_TYPE
1701 @defmacx INT_FAST64_TYPE
1702 @defmacx UINT_FAST8_TYPE
1703 @defmacx UINT_FAST16_TYPE
1704 @defmacx UINT_FAST32_TYPE
1705 @defmacx UINT_FAST64_TYPE
1706 @defmacx INTPTR_TYPE
1707 @defmacx UINTPTR_TYPE
1708 C expressions for the standard types @code{sig_atomic_t},
1709 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1710 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1711 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1712 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1713 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1714 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1715 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1716 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1717 @code{SIZE_TYPE} above for more information.
1719 If any of these macros evaluates to a null pointer, the corresponding
1720 type is not supported; if GCC is configured to provide
1721 @code{<stdint.h>} in such a case, the header provided may not conform
1722 to C99, depending on the type in question. The defaults for all of
1723 these macros are null pointers.
1726 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1727 The C++ compiler represents a pointer-to-member-function with a struct
1734 ptrdiff_t vtable_index;
1741 The C++ compiler must use one bit to indicate whether the function that
1742 will be called through a pointer-to-member-function is virtual.
1743 Normally, we assume that the low-order bit of a function pointer must
1744 always be zero. Then, by ensuring that the vtable_index is odd, we can
1745 distinguish which variant of the union is in use. But, on some
1746 platforms function pointers can be odd, and so this doesn't work. In
1747 that case, we use the low-order bit of the @code{delta} field, and shift
1748 the remainder of the @code{delta} field to the left.
1750 GCC will automatically make the right selection about where to store
1751 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1752 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1753 set such that functions always start at even addresses, but the lowest
1754 bit of pointers to functions indicate whether the function at that
1755 address is in ARM or Thumb mode. If this is the case of your
1756 architecture, you should define this macro to
1757 @code{ptrmemfunc_vbit_in_delta}.
1759 In general, you should not have to define this macro. On architectures
1760 in which function addresses are always even, according to
1761 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1762 @code{ptrmemfunc_vbit_in_pfn}.
1765 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1766 Normally, the C++ compiler uses function pointers in vtables. This
1767 macro allows the target to change to use ``function descriptors''
1768 instead. Function descriptors are found on targets for whom a
1769 function pointer is actually a small data structure. Normally the
1770 data structure consists of the actual code address plus a data
1771 pointer to which the function's data is relative.
1773 If vtables are used, the value of this macro should be the number
1774 of words that the function descriptor occupies.
1777 @defmac TARGET_VTABLE_ENTRY_ALIGN
1778 By default, the vtable entries are void pointers, the so the alignment
1779 is the same as pointer alignment. The value of this macro specifies
1780 the alignment of the vtable entry in bits. It should be defined only
1781 when special alignment is necessary. */
1784 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1785 There are a few non-descriptor entries in the vtable at offsets below
1786 zero. If these entries must be padded (say, to preserve the alignment
1787 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1788 of words in each data entry.
1792 @section Register Usage
1793 @cindex register usage
1795 This section explains how to describe what registers the target machine
1796 has, and how (in general) they can be used.
1798 The description of which registers a specific instruction can use is
1799 done with register classes; see @ref{Register Classes}. For information
1800 on using registers to access a stack frame, see @ref{Frame Registers}.
1801 For passing values in registers, see @ref{Register Arguments}.
1802 For returning values in registers, see @ref{Scalar Return}.
1805 * Register Basics:: Number and kinds of registers.
1806 * Allocation Order:: Order in which registers are allocated.
1807 * Values in Registers:: What kinds of values each reg can hold.
1808 * Leaf Functions:: Renumbering registers for leaf functions.
1809 * Stack Registers:: Handling a register stack such as 80387.
1812 @node Register Basics
1813 @subsection Basic Characteristics of Registers
1815 @c prevent bad page break with this line
1816 Registers have various characteristics.
1818 @defmac FIRST_PSEUDO_REGISTER
1819 Number of hardware registers known to the compiler. They receive
1820 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1821 pseudo register's number really is assigned the number
1822 @code{FIRST_PSEUDO_REGISTER}.
1825 @defmac FIXED_REGISTERS
1826 @cindex fixed register
1827 An initializer that says which registers are used for fixed purposes
1828 all throughout the compiled code and are therefore not available for
1829 general allocation. These would include the stack pointer, the frame
1830 pointer (except on machines where that can be used as a general
1831 register when no frame pointer is needed), the program counter on
1832 machines where that is considered one of the addressable registers,
1833 and any other numbered register with a standard use.
1835 This information is expressed as a sequence of numbers, separated by
1836 commas and surrounded by braces. The @var{n}th number is 1 if
1837 register @var{n} is fixed, 0 otherwise.
1839 The table initialized from this macro, and the table initialized by
1840 the following one, may be overridden at run time either automatically,
1841 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1842 the user with the command options @option{-ffixed-@var{reg}},
1843 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1846 @defmac CALL_USED_REGISTERS
1847 @cindex call-used register
1848 @cindex call-clobbered register
1849 @cindex call-saved register
1850 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1851 clobbered (in general) by function calls as well as for fixed
1852 registers. This macro therefore identifies the registers that are not
1853 available for general allocation of values that must live across
1856 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1857 automatically saves it on function entry and restores it on function
1858 exit, if the register is used within the function.
1861 @defmac CALL_REALLY_USED_REGISTERS
1862 @cindex call-used register
1863 @cindex call-clobbered register
1864 @cindex call-saved register
1865 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1866 that the entire set of @code{FIXED_REGISTERS} be included.
1867 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1868 This macro is optional. If not specified, it defaults to the value
1869 of @code{CALL_USED_REGISTERS}.
1872 @cindex call-used register
1873 @cindex call-clobbered register
1874 @cindex call-saved register
1875 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_CALL_PART_CLOBBERED (unsigned int @var{regno}, machine_mode @var{mode})
1876 This hook should return true if @var{regno} is partly call-saved and
1877 partly call-clobbered, and if a value of mode @var{mode} would be partly
1878 clobbered by a call. For example, if the low 32 bits of @var{regno} are
1879 preserved across a call but higher bits are clobbered, this hook should
1880 return true for a 64-bit mode but false for a 32-bit mode.
1882 The default implementation returns false, which is correct
1883 for targets that don't have partly call-clobbered registers.
1887 @findex call_used_regs
1890 @findex reg_class_contents
1891 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1892 This hook may conditionally modify five variables
1893 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1894 @code{reg_names}, and @code{reg_class_contents}, to take into account
1895 any dependence of these register sets on target flags. The first three
1896 of these are of type @code{char []} (interpreted as boolean vectors).
1897 @code{global_regs} is a @code{const char *[]}, and
1898 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1899 called, @code{fixed_regs}, @code{call_used_regs},
1900 @code{reg_class_contents}, and @code{reg_names} have been initialized
1901 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1902 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1903 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1904 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1905 command options have been applied.
1907 @cindex disabling certain registers
1908 @cindex controlling register usage
1909 If the usage of an entire class of registers depends on the target
1910 flags, you may indicate this to GCC by using this macro to modify
1911 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1912 registers in the classes which should not be used by GCC@. Also make
1913 @code{define_register_constraint}s return @code{NO_REGS} for constraints
1914 that shouldn't be used.
1916 (However, if this class is not included in @code{GENERAL_REGS} and all
1917 of the insn patterns whose constraints permit this class are
1918 controlled by target switches, then GCC will automatically avoid using
1919 these registers when the target switches are opposed to them.)
1922 @defmac INCOMING_REGNO (@var{out})
1923 Define this macro if the target machine has register windows. This C
1924 expression returns the register number as seen by the called function
1925 corresponding to the register number @var{out} as seen by the calling
1926 function. Return @var{out} if register number @var{out} is not an
1930 @defmac OUTGOING_REGNO (@var{in})
1931 Define this macro if the target machine has register windows. This C
1932 expression returns the register number as seen by the calling function
1933 corresponding to the register number @var{in} as seen by the called
1934 function. Return @var{in} if register number @var{in} is not an inbound
1938 @defmac LOCAL_REGNO (@var{regno})
1939 Define this macro if the target machine has register windows. This C
1940 expression returns true if the register is call-saved but is in the
1941 register window. Unlike most call-saved registers, such registers
1942 need not be explicitly restored on function exit or during non-local
1947 If the program counter has a register number, define this as that
1948 register number. Otherwise, do not define it.
1951 @node Allocation Order
1952 @subsection Order of Allocation of Registers
1953 @cindex order of register allocation
1954 @cindex register allocation order
1956 @c prevent bad page break with this line
1957 Registers are allocated in order.
1959 @defmac REG_ALLOC_ORDER
1960 If defined, an initializer for a vector of integers, containing the
1961 numbers of hard registers in the order in which GCC should prefer
1962 to use them (from most preferred to least).
1964 If this macro is not defined, registers are used lowest numbered first
1965 (all else being equal).
1967 One use of this macro is on machines where the highest numbered
1968 registers must always be saved and the save-multiple-registers
1969 instruction supports only sequences of consecutive registers. On such
1970 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1971 the highest numbered allocable register first.
1974 @defmac ADJUST_REG_ALLOC_ORDER
1975 A C statement (sans semicolon) to choose the order in which to allocate
1976 hard registers for pseudo-registers local to a basic block.
1978 Store the desired register order in the array @code{reg_alloc_order}.
1979 Element 0 should be the register to allocate first; element 1, the next
1980 register; and so on.
1982 The macro body should not assume anything about the contents of
1983 @code{reg_alloc_order} before execution of the macro.
1985 On most machines, it is not necessary to define this macro.
1988 @defmac HONOR_REG_ALLOC_ORDER
1989 Normally, IRA tries to estimate the costs for saving a register in the
1990 prologue and restoring it in the epilogue. This discourages it from
1991 using call-saved registers. If a machine wants to ensure that IRA
1992 allocates registers in the order given by REG_ALLOC_ORDER even if some
1993 call-saved registers appear earlier than call-used ones, then define this
1994 macro as a C expression to nonzero. Default is 0.
1997 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
1998 In some case register allocation order is not enough for the
1999 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2000 If this macro is defined, it should return a floating point value
2001 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2002 be increased by approximately the pseudo's usage frequency times the
2003 value returned by this macro. Not defining this macro is equivalent
2004 to having it always return @code{0.0}.
2006 On most machines, it is not necessary to define this macro.
2009 @node Values in Registers
2010 @subsection How Values Fit in Registers
2012 This section discusses the macros that describe which kinds of values
2013 (specifically, which machine modes) each register can hold, and how many
2014 consecutive registers are needed for a given mode.
2016 @deftypefn {Target Hook} {unsigned int} TARGET_HARD_REGNO_NREGS (unsigned int @var{regno}, machine_mode @var{mode})
2017 This hook returns the number of consecutive hard registers, starting
2018 at register number @var{regno}, required to hold a value of mode
2019 @var{mode}. This hook must never return zero, even if a register
2020 cannot hold the requested mode - indicate that with
2021 @code{TARGET_HARD_REGNO_MODE_OK} and/or
2022 @code{TARGET_CAN_CHANGE_MODE_CLASS} instead.
2024 The default definition returns the number of words in @var{mode}.
2027 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2028 A C expression that is nonzero if a value of mode @var{mode}, stored
2029 in memory, ends with padding that causes it to take up more space than
2030 in registers starting at register number @var{regno} (as determined by
2031 multiplying GCC's notion of the size of the register when containing
2032 this mode by the number of registers returned by
2033 @code{TARGET_HARD_REGNO_NREGS}). By default this is zero.
2035 For example, if a floating-point value is stored in three 32-bit
2036 registers but takes up 128 bits in memory, then this would be
2039 This macros only needs to be defined if there are cases where
2040 @code{subreg_get_info}
2041 would otherwise wrongly determine that a @code{subreg} can be
2042 represented by an offset to the register number, when in fact such a
2043 @code{subreg} would contain some of the padding not stored in
2044 registers and so not be representable.
2047 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2048 For values of @var{regno} and @var{mode} for which
2049 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2050 returning the greater number of registers required to hold the value
2051 including any padding. In the example above, the value would be four.
2054 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2055 Define this macro if the natural size of registers that hold values
2056 of mode @var{mode} is not the word size. It is a C expression that
2057 should give the natural size in bytes for the specified mode. It is
2058 used by the register allocator to try to optimize its results. This
2059 happens for example on SPARC 64-bit where the natural size of
2060 floating-point registers is still 32-bit.
2063 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_MODE_OK (unsigned int @var{regno}, machine_mode @var{mode})
2064 This hook returns true if it is permissible to store a value
2065 of mode @var{mode} in hard register number @var{regno} (or in several
2066 registers starting with that one). The default definition returns true
2069 You need not include code to check for the numbers of fixed registers,
2070 because the allocation mechanism considers them to be always occupied.
2072 @cindex register pairs
2073 On some machines, double-precision values must be kept in even/odd
2074 register pairs. You can implement that by defining this hook to reject
2075 odd register numbers for such modes.
2077 The minimum requirement for a mode to be OK in a register is that the
2078 @samp{mov@var{mode}} instruction pattern support moves between the
2079 register and other hard register in the same class and that moving a
2080 value into the register and back out not alter it.
2082 Since the same instruction used to move @code{word_mode} will work for
2083 all narrower integer modes, it is not necessary on any machine for
2084 this hook to distinguish between these modes, provided you define
2085 patterns @samp{movhi}, etc., to take advantage of this. This is
2086 useful because of the interaction between @code{TARGET_HARD_REGNO_MODE_OK}
2087 and @code{TARGET_MODES_TIEABLE_P}; it is very desirable for all integer
2088 modes to be tieable.
2090 Many machines have special registers for floating point arithmetic.
2091 Often people assume that floating point machine modes are allowed only
2092 in floating point registers. This is not true. Any registers that
2093 can hold integers can safely @emph{hold} a floating point machine
2094 mode, whether or not floating arithmetic can be done on it in those
2095 registers. Integer move instructions can be used to move the values.
2097 On some machines, though, the converse is true: fixed-point machine
2098 modes may not go in floating registers. This is true if the floating
2099 registers normalize any value stored in them, because storing a
2100 non-floating value there would garble it. In this case,
2101 @code{TARGET_HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2102 floating registers. But if the floating registers do not automatically
2103 normalize, if you can store any bit pattern in one and retrieve it
2104 unchanged without a trap, then any machine mode may go in a floating
2105 register, so you can define this hook to say so.
2107 The primary significance of special floating registers is rather that
2108 they are the registers acceptable in floating point arithmetic
2109 instructions. However, this is of no concern to
2110 @code{TARGET_HARD_REGNO_MODE_OK}. You handle it by writing the proper
2111 constraints for those instructions.
2113 On some machines, the floating registers are especially slow to access,
2114 so that it is better to store a value in a stack frame than in such a
2115 register if floating point arithmetic is not being done. As long as the
2116 floating registers are not in class @code{GENERAL_REGS}, they will not
2117 be used unless some pattern's constraint asks for one.
2120 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2121 A C expression that is nonzero if it is OK to rename a hard register
2122 @var{from} to another hard register @var{to}.
2124 One common use of this macro is to prevent renaming of a register to
2125 another register that is not saved by a prologue in an interrupt
2128 The default is always nonzero.
2131 @deftypefn {Target Hook} bool TARGET_MODES_TIEABLE_P (machine_mode @var{mode1}, machine_mode @var{mode2})
2132 This hook returns true if a value of mode @var{mode1} is accessible
2133 in mode @var{mode2} without copying.
2135 If @code{TARGET_HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2136 @code{TARGET_HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always
2137 the same for any @var{r}, then
2138 @code{TARGET_MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2139 should be true. If they differ for any @var{r}, you should define
2140 this hook to return false unless some other mechanism ensures the
2141 accessibility of the value in a narrower mode.
2143 You should define this hook to return true in as many cases as
2144 possible since doing so will allow GCC to perform better register
2145 allocation. The default definition returns true unconditionally.
2148 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2149 This target hook should return @code{true} if it is OK to use a hard register
2150 @var{regno} as scratch reg in peephole2.
2152 One common use of this macro is to prevent using of a register that
2153 is not saved by a prologue in an interrupt handler.
2155 The default version of this hook always returns @code{true}.
2158 @defmac AVOID_CCMODE_COPIES
2159 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2160 registers. You should only define this macro if support for copying to/from
2161 @code{CCmode} is incomplete.
2164 @node Leaf Functions
2165 @subsection Handling Leaf Functions
2167 @cindex leaf functions
2168 @cindex functions, leaf
2169 On some machines, a leaf function (i.e., one which makes no calls) can run
2170 more efficiently if it does not make its own register window. Often this
2171 means it is required to receive its arguments in the registers where they
2172 are passed by the caller, instead of the registers where they would
2175 The special treatment for leaf functions generally applies only when
2176 other conditions are met; for example, often they may use only those
2177 registers for its own variables and temporaries. We use the term ``leaf
2178 function'' to mean a function that is suitable for this special
2179 handling, so that functions with no calls are not necessarily ``leaf
2182 GCC assigns register numbers before it knows whether the function is
2183 suitable for leaf function treatment. So it needs to renumber the
2184 registers in order to output a leaf function. The following macros
2187 @defmac LEAF_REGISTERS
2188 Name of a char vector, indexed by hard register number, which
2189 contains 1 for a register that is allowable in a candidate for leaf
2192 If leaf function treatment involves renumbering the registers, then the
2193 registers marked here should be the ones before renumbering---those that
2194 GCC would ordinarily allocate. The registers which will actually be
2195 used in the assembler code, after renumbering, should not be marked with 1
2198 Define this macro only if the target machine offers a way to optimize
2199 the treatment of leaf functions.
2202 @defmac LEAF_REG_REMAP (@var{regno})
2203 A C expression whose value is the register number to which @var{regno}
2204 should be renumbered, when a function is treated as a leaf function.
2206 If @var{regno} is a register number which should not appear in a leaf
2207 function before renumbering, then the expression should yield @minus{}1, which
2208 will cause the compiler to abort.
2210 Define this macro only if the target machine offers a way to optimize the
2211 treatment of leaf functions, and registers need to be renumbered to do
2215 @findex current_function_is_leaf
2216 @findex current_function_uses_only_leaf_regs
2217 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2218 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2219 specially. They can test the C variable @code{current_function_is_leaf}
2220 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2221 set prior to local register allocation and is valid for the remaining
2222 compiler passes. They can also test the C variable
2223 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2224 functions which only use leaf registers.
2225 @code{current_function_uses_only_leaf_regs} is valid after all passes
2226 that modify the instructions have been run and is only useful if
2227 @code{LEAF_REGISTERS} is defined.
2228 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2229 @c of the next paragraph?! --mew 2feb93
2231 @node Stack Registers
2232 @subsection Registers That Form a Stack
2234 There are special features to handle computers where some of the
2235 ``registers'' form a stack. Stack registers are normally written by
2236 pushing onto the stack, and are numbered relative to the top of the
2239 Currently, GCC can only handle one group of stack-like registers, and
2240 they must be consecutively numbered. Furthermore, the existing
2241 support for stack-like registers is specific to the 80387 floating
2242 point coprocessor. If you have a new architecture that uses
2243 stack-like registers, you will need to do substantial work on
2244 @file{reg-stack.c} and write your machine description to cooperate
2245 with it, as well as defining these macros.
2248 Define this if the machine has any stack-like registers.
2251 @defmac STACK_REG_COVER_CLASS
2252 This is a cover class containing the stack registers. Define this if
2253 the machine has any stack-like registers.
2256 @defmac FIRST_STACK_REG
2257 The number of the first stack-like register. This one is the top
2261 @defmac LAST_STACK_REG
2262 The number of the last stack-like register. This one is the bottom of
2266 @node Register Classes
2267 @section Register Classes
2268 @cindex register class definitions
2269 @cindex class definitions, register
2271 On many machines, the numbered registers are not all equivalent.
2272 For example, certain registers may not be allowed for indexed addressing;
2273 certain registers may not be allowed in some instructions. These machine
2274 restrictions are described to the compiler using @dfn{register classes}.
2276 You define a number of register classes, giving each one a name and saying
2277 which of the registers belong to it. Then you can specify register classes
2278 that are allowed as operands to particular instruction patterns.
2282 In general, each register will belong to several classes. In fact, one
2283 class must be named @code{ALL_REGS} and contain all the registers. Another
2284 class must be named @code{NO_REGS} and contain no registers. Often the
2285 union of two classes will be another class; however, this is not required.
2287 @findex GENERAL_REGS
2288 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2289 terribly special about the name, but the operand constraint letters
2290 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2291 the same as @code{ALL_REGS}, just define it as a macro which expands
2294 Order the classes so that if class @var{x} is contained in class @var{y}
2295 then @var{x} has a lower class number than @var{y}.
2297 The way classes other than @code{GENERAL_REGS} are specified in operand
2298 constraints is through machine-dependent operand constraint letters.
2299 You can define such letters to correspond to various classes, then use
2300 them in operand constraints.
2302 You must define the narrowest register classes for allocatable
2303 registers, so that each class either has no subclasses, or that for
2304 some mode, the move cost between registers within the class is
2305 cheaper than moving a register in the class to or from memory
2308 You should define a class for the union of two classes whenever some
2309 instruction allows both classes. For example, if an instruction allows
2310 either a floating point (coprocessor) register or a general register for a
2311 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2312 which includes both of them. Otherwise you will get suboptimal code,
2313 or even internal compiler errors when reload cannot find a register in the
2314 class computed via @code{reg_class_subunion}.
2316 You must also specify certain redundant information about the register
2317 classes: for each class, which classes contain it and which ones are
2318 contained in it; for each pair of classes, the largest class contained
2321 When a value occupying several consecutive registers is expected in a
2322 certain class, all the registers used must belong to that class.
2323 Therefore, register classes cannot be used to enforce a requirement for
2324 a register pair to start with an even-numbered register. The way to
2325 specify this requirement is with @code{TARGET_HARD_REGNO_MODE_OK}.
2327 Register classes used for input-operands of bitwise-and or shift
2328 instructions have a special requirement: each such class must have, for
2329 each fixed-point machine mode, a subclass whose registers can transfer that
2330 mode to or from memory. For example, on some machines, the operations for
2331 single-byte values (@code{QImode}) are limited to certain registers. When
2332 this is so, each register class that is used in a bitwise-and or shift
2333 instruction must have a subclass consisting of registers from which
2334 single-byte values can be loaded or stored. This is so that
2335 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2337 @deftp {Data type} {enum reg_class}
2338 An enumerated type that must be defined with all the register class names
2339 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2340 must be the last register class, followed by one more enumerated value,
2341 @code{LIM_REG_CLASSES}, which is not a register class but rather
2342 tells how many classes there are.
2344 Each register class has a number, which is the value of casting
2345 the class name to type @code{int}. The number serves as an index
2346 in many of the tables described below.
2349 @defmac N_REG_CLASSES
2350 The number of distinct register classes, defined as follows:
2353 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2357 @defmac REG_CLASS_NAMES
2358 An initializer containing the names of the register classes as C string
2359 constants. These names are used in writing some of the debugging dumps.
2362 @defmac REG_CLASS_CONTENTS
2363 An initializer containing the contents of the register classes, as integers
2364 which are bit masks. The @var{n}th integer specifies the contents of class
2365 @var{n}. The way the integer @var{mask} is interpreted is that
2366 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2368 When the machine has more than 32 registers, an integer does not suffice.
2369 Then the integers are replaced by sub-initializers, braced groupings containing
2370 several integers. Each sub-initializer must be suitable as an initializer
2371 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2372 In this situation, the first integer in each sub-initializer corresponds to
2373 registers 0 through 31, the second integer to registers 32 through 63, and
2377 @defmac REGNO_REG_CLASS (@var{regno})
2378 A C expression whose value is a register class containing hard register
2379 @var{regno}. In general there is more than one such class; choose a class
2380 which is @dfn{minimal}, meaning that no smaller class also contains the
2384 @defmac BASE_REG_CLASS
2385 A macro whose definition is the name of the class to which a valid
2386 base register must belong. A base register is one used in an address
2387 which is the register value plus a displacement.
2390 @defmac MODE_BASE_REG_CLASS (@var{mode})
2391 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2392 the selection of a base register in a mode dependent manner. If
2393 @var{mode} is VOIDmode then it should return the same value as
2394 @code{BASE_REG_CLASS}.
2397 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2398 A C expression whose value is the register class to which a valid
2399 base register must belong in order to be used in a base plus index
2400 register address. You should define this macro if base plus index
2401 addresses have different requirements than other base register uses.
2404 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2405 A C expression whose value is the register class to which a valid
2406 base register for a memory reference in mode @var{mode} to address
2407 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2408 define the context in which the base register occurs. @var{outer_code} is
2409 the code of the immediately enclosing expression (@code{MEM} for the top level
2410 of an address, @code{ADDRESS} for something that occurs in an
2411 @code{address_operand}). @var{index_code} is the code of the corresponding
2412 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2415 @defmac INDEX_REG_CLASS
2416 A macro whose definition is the name of the class to which a valid
2417 index register must belong. An index register is one used in an
2418 address where its value is either multiplied by a scale factor or
2419 added to another register (as well as added to a displacement).
2422 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2423 A C expression which is nonzero if register number @var{num} is
2424 suitable for use as a base register in operand addresses.
2427 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2428 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2429 that expression may examine the mode of the memory reference in
2430 @var{mode}. You should define this macro if the mode of the memory
2431 reference affects whether a register may be used as a base register. If
2432 you define this macro, the compiler will use it instead of
2433 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2434 addresses that appear outside a @code{MEM}, i.e., as an
2435 @code{address_operand}.
2438 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2439 A C expression which is nonzero if register number @var{num} is suitable for
2440 use as a base register in base plus index operand addresses, accessing
2441 memory in mode @var{mode}. It may be either a suitable hard register or a
2442 pseudo register that has been allocated such a hard register. You should
2443 define this macro if base plus index addresses have different requirements
2444 than other base register uses.
2446 Use of this macro is deprecated; please use the more general
2447 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2450 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2451 A C expression which is nonzero if register number @var{num} is
2452 suitable for use as a base register in operand addresses, accessing
2453 memory in mode @var{mode} in address space @var{address_space}.
2454 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2455 that that expression may examine the context in which the register
2456 appears in the memory reference. @var{outer_code} is the code of the
2457 immediately enclosing expression (@code{MEM} if at the top level of the
2458 address, @code{ADDRESS} for something that occurs in an
2459 @code{address_operand}). @var{index_code} is the code of the
2460 corresponding index expression if @var{outer_code} is @code{PLUS};
2461 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2462 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2465 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2466 A C expression which is nonzero if register number @var{num} is
2467 suitable for use as an index register in operand addresses. It may be
2468 either a suitable hard register or a pseudo register that has been
2469 allocated such a hard register.
2471 The difference between an index register and a base register is that
2472 the index register may be scaled. If an address involves the sum of
2473 two registers, neither one of them scaled, then either one may be
2474 labeled the ``base'' and the other the ``index''; but whichever
2475 labeling is used must fit the machine's constraints of which registers
2476 may serve in each capacity. The compiler will try both labelings,
2477 looking for one that is valid, and will reload one or both registers
2478 only if neither labeling works.
2481 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2482 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.
2485 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2486 A target hook that places additional restrictions on the register class
2487 to use when it is necessary to copy value @var{x} into a register in class
2488 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2489 another, smaller class.
2491 The default version of this hook always returns value of @code{rclass} argument.
2493 Sometimes returning a more restrictive class makes better code. For
2494 example, on the 68000, when @var{x} is an integer constant that is in range
2495 for a @samp{moveq} instruction, the value of this macro is always
2496 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2497 Requiring a data register guarantees that a @samp{moveq} will be used.
2499 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2500 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2501 loaded into some register class. By returning @code{NO_REGS} you can
2502 force @var{x} into a memory location. For example, rs6000 can load
2503 immediate values into general-purpose registers, but does not have an
2504 instruction for loading an immediate value into a floating-point
2505 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2506 @var{x} is a floating-point constant. If the constant can't be loaded
2507 into any kind of register, code generation will be better if
2508 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2509 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2511 If an insn has pseudos in it after register allocation, reload will go
2512 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2513 to find the best one. Returning @code{NO_REGS}, in this case, makes
2514 reload add a @code{!} in front of the constraint: the x86 back-end uses
2515 this feature to discourage usage of 387 registers when math is done in
2516 the SSE registers (and vice versa).
2519 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2520 A C expression that places additional restrictions on the register class
2521 to use when it is necessary to copy value @var{x} into a register in class
2522 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2523 another, smaller class. On many machines, the following definition is
2527 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2530 Sometimes returning a more restrictive class makes better code. For
2531 example, on the 68000, when @var{x} is an integer constant that is in range
2532 for a @samp{moveq} instruction, the value of this macro is always
2533 @code{DATA_REGS} as long as @var{class} includes the data registers.
2534 Requiring a data register guarantees that a @samp{moveq} will be used.
2536 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2537 @var{class} is if @var{x} is a legitimate constant which cannot be
2538 loaded into some register class. By returning @code{NO_REGS} you can
2539 force @var{x} into a memory location. For example, rs6000 can load
2540 immediate values into general-purpose registers, but does not have an
2541 instruction for loading an immediate value into a floating-point
2542 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2543 @var{x} is a floating-point constant. If the constant cannot be loaded
2544 into any kind of register, code generation will be better if
2545 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2546 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2548 If an insn has pseudos in it after register allocation, reload will go
2549 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2550 to find the best one. Returning @code{NO_REGS}, in this case, makes
2551 reload add a @code{!} in front of the constraint: the x86 back-end uses
2552 this feature to discourage usage of 387 registers when math is done in
2553 the SSE registers (and vice versa).
2556 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2557 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2560 The default version of this hook always returns value of @code{rclass}
2563 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2564 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2567 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2568 A C expression that places additional restrictions on the register class
2569 to use when it is necessary to be able to hold a value of mode
2570 @var{mode} in a reload register for which class @var{class} would
2573 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2574 there are certain modes that simply cannot go in certain reload classes.
2576 The value is a register class; perhaps @var{class}, or perhaps another,
2579 Don't define this macro unless the target machine has limitations which
2580 require the macro to do something nontrivial.
2583 @deftypefn {Target Hook} reg_class_t TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, reg_class_t @var{reload_class}, machine_mode @var{reload_mode}, secondary_reload_info *@var{sri})
2584 Many machines have some registers that cannot be copied directly to or
2585 from memory or even from other types of registers. An example is the
2586 @samp{MQ} register, which on most machines, can only be copied to or
2587 from general registers, but not memory. Below, we shall be using the
2588 term 'intermediate register' when a move operation cannot be performed
2589 directly, but has to be done by copying the source into the intermediate
2590 register first, and then copying the intermediate register to the
2591 destination. An intermediate register always has the same mode as
2592 source and destination. Since it holds the actual value being copied,
2593 reload might apply optimizations to re-use an intermediate register
2594 and eliding the copy from the source when it can determine that the
2595 intermediate register still holds the required value.
2597 Another kind of secondary reload is required on some machines which
2598 allow copying all registers to and from memory, but require a scratch
2599 register for stores to some memory locations (e.g., those with symbolic
2600 address on the RT, and those with certain symbolic address on the SPARC
2601 when compiling PIC)@. Scratch registers need not have the same mode
2602 as the value being copied, and usually hold a different value than
2603 that being copied. Special patterns in the md file are needed to
2604 describe how the copy is performed with the help of the scratch register;
2605 these patterns also describe the number, register class(es) and mode(s)
2606 of the scratch register(s).
2608 In some cases, both an intermediate and a scratch register are required.
2610 For input reloads, this target hook is called with nonzero @var{in_p},
2611 and @var{x} is an rtx that needs to be copied to a register of class
2612 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2613 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2614 needs to be copied to rtx @var{x} in @var{reload_mode}.
2616 If copying a register of @var{reload_class} from/to @var{x} requires
2617 an intermediate register, the hook @code{secondary_reload} should
2618 return the register class required for this intermediate register.
2619 If no intermediate register is required, it should return NO_REGS.
2620 If more than one intermediate register is required, describe the one
2621 that is closest in the copy chain to the reload register.
2623 If scratch registers are needed, you also have to describe how to
2624 perform the copy from/to the reload register to/from this
2625 closest intermediate register. Or if no intermediate register is
2626 required, but still a scratch register is needed, describe the
2627 copy from/to the reload register to/from the reload operand @var{x}.
2629 You do this by setting @code{sri->icode} to the instruction code of a pattern
2630 in the md file which performs the move. Operands 0 and 1 are the output
2631 and input of this copy, respectively. Operands from operand 2 onward are
2632 for scratch operands. These scratch operands must have a mode, and a
2633 single-register-class
2634 @c [later: or memory]
2637 When an intermediate register is used, the @code{secondary_reload}
2638 hook will be called again to determine how to copy the intermediate
2639 register to/from the reload operand @var{x}, so your hook must also
2640 have code to handle the register class of the intermediate operand.
2642 @c [For later: maybe we'll allow multi-alternative reload patterns -
2643 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2644 @c and match the constraints of input and output to determine the required
2645 @c alternative. A restriction would be that constraints used to match
2646 @c against reloads registers would have to be written as register class
2647 @c constraints, or we need a new target macro / hook that tells us if an
2648 @c arbitrary constraint can match an unknown register of a given class.
2649 @c Such a macro / hook would also be useful in other places.]
2652 @var{x} might be a pseudo-register or a @code{subreg} of a
2653 pseudo-register, which could either be in a hard register or in memory.
2654 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2655 in memory and the hard register number if it is in a register.
2657 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2658 currently not supported. For the time being, you will have to continue
2659 to use @code{TARGET_SECONDARY_MEMORY_NEEDED} for that purpose.
2661 @code{copy_cost} also uses this target hook to find out how values are
2662 copied. If you want it to include some extra cost for the need to allocate
2663 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2664 Or if two dependent moves are supposed to have a lower cost than the sum
2665 of the individual moves due to expected fortuitous scheduling and/or special
2666 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2669 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2670 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2671 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2672 These macros are obsolete, new ports should use the target hook
2673 @code{TARGET_SECONDARY_RELOAD} instead.
2675 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2676 target hook. Older ports still define these macros to indicate to the
2677 reload phase that it may
2678 need to allocate at least one register for a reload in addition to the
2679 register to contain the data. Specifically, if copying @var{x} to a
2680 register @var{class} in @var{mode} requires an intermediate register,
2681 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2682 largest register class all of whose registers can be used as
2683 intermediate registers or scratch registers.
2685 If copying a register @var{class} in @var{mode} to @var{x} requires an
2686 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2687 was supposed to be defined be defined to return the largest register
2688 class required. If the
2689 requirements for input and output reloads were the same, the macro
2690 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2693 The values returned by these macros are often @code{GENERAL_REGS}.
2694 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2695 can be directly copied to or from a register of @var{class} in
2696 @var{mode} without requiring a scratch register. Do not define this
2697 macro if it would always return @code{NO_REGS}.
2699 If a scratch register is required (either with or without an
2700 intermediate register), you were supposed to define patterns for
2701 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2702 (@pxref{Standard Names}. These patterns, which were normally
2703 implemented with a @code{define_expand}, should be similar to the
2704 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2707 These patterns need constraints for the reload register and scratch
2709 contain a single register class. If the original reload register (whose
2710 class is @var{class}) can meet the constraint given in the pattern, the
2711 value returned by these macros is used for the class of the scratch
2712 register. Otherwise, two additional reload registers are required.
2713 Their classes are obtained from the constraints in the insn pattern.
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 These macros should not be used in the case where a particular class of
2721 registers can only be copied to memory and not to another class of
2722 registers. In that case, secondary reload registers are not needed and
2723 would not be helpful. Instead, a stack location must be used to perform
2724 the copy and the @code{mov@var{m}} pattern should use memory as an
2725 intermediate storage. This case often occurs between floating-point and
2729 @deftypefn {Target Hook} bool TARGET_SECONDARY_MEMORY_NEEDED (machine_mode @var{mode}, reg_class_t @var{class1}, reg_class_t @var{class2})
2730 Certain machines have the property that some registers cannot be copied
2731 to some other registers without using memory. Define this hook on
2732 those machines to return true if objects of mode @var{m} in registers
2733 of @var{class1} can only be copied to registers of class @var{class2} by
2734 storing a register of @var{class1} into memory and loading that memory
2735 location into a register of @var{class2}. The default definition returns
2736 false for all inputs.
2739 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2740 Normally when @code{TARGET_SECONDARY_MEMORY_NEEDED} is defined, the compiler
2741 allocates a stack slot for a memory location needed for register copies.
2742 If this macro is defined, the compiler instead uses the memory location
2743 defined by this macro.
2745 Do not define this macro if you do not define
2746 @code{TARGET_SECONDARY_MEMORY_NEEDED}.
2749 @deftypefn {Target Hook} machine_mode TARGET_SECONDARY_MEMORY_NEEDED_MODE (machine_mode @var{mode})
2750 If @code{TARGET_SECONDARY_MEMORY_NEEDED} tells the compiler to use memory
2751 when moving between two particular registers of mode @var{mode},
2752 this hook specifies the mode that the memory should have.
2754 The default depends on @code{TARGET_LRA_P}. Without LRA, the default
2755 is to use a word-sized mode for integral modes that are smaller than a
2756 a word. This is right thing to do on most machines because it ensures
2757 that all bits of the register are copied and prevents accesses to the
2758 registers in a narrower mode, which some machines prohibit for
2759 floating-point registers.
2761 However, this default behavior is not correct on some machines, such as
2762 the DEC Alpha, that store short integers in floating-point registers
2763 differently than in integer registers. On those machines, the default
2764 widening will not work correctly and you must define this hook to
2765 suppress that widening in some cases. See the file @file{alpha.c} for
2768 With LRA, the default is to use @var{mode} unmodified.
2771 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2772 A target hook which returns @code{true} if pseudos that have been assigned
2773 to registers of class @var{rclass} would likely be spilled because
2774 registers of @var{rclass} are needed for spill registers.
2776 The default version of this target hook returns @code{true} if @var{rclass}
2777 has exactly one register and @code{false} otherwise. On most machines, this
2778 default should be used. For generally register-starved machines, such as
2779 i386, or machines with right register constraints, such as SH, this hook
2780 can be used to avoid excessive spilling.
2782 This hook is also used by some of the global intra-procedural code
2783 transformations to throtle code motion, to avoid increasing register
2787 @deftypefn {Target Hook} {unsigned char} TARGET_CLASS_MAX_NREGS (reg_class_t @var{rclass}, machine_mode @var{mode})
2788 A target hook returns the maximum number of consecutive registers
2789 of class @var{rclass} needed to hold a value of mode @var{mode}.
2791 This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}.
2792 In fact, the value returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass},
2793 @var{mode})} target hook should be the maximum value of
2794 @code{TARGET_HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno}
2795 values in the class @var{rclass}.
2797 This target hook helps control the handling of multiple-word values
2800 The default version of this target hook returns the size of @var{mode}
2804 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2805 A C expression for the maximum number of consecutive registers
2806 of class @var{class} needed to hold a value of mode @var{mode}.
2808 This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}. In fact,
2809 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2810 should be the maximum value of @code{TARGET_HARD_REGNO_NREGS (@var{regno},
2811 @var{mode})} for all @var{regno} values in the class @var{class}.
2813 This macro helps control the handling of multiple-word values
2817 @deftypefn {Target Hook} bool TARGET_CAN_CHANGE_MODE_CLASS (machine_mode @var{from}, machine_mode @var{to}, reg_class_t @var{rclass})
2818 This hook returns true if it is possible to bitcast values held in
2819 registers of class @var{rclass} from mode @var{from} to mode @var{to}
2820 and if doing so preserves the low-order bits that are common to both modes.
2821 The result is only meaningful if @var{rclass} has registers that can hold
2822 both @code{from} and @code{to}. The default implementation returns true.
2824 As an example of when such bitcasting is invalid, loading 32-bit integer or
2825 floating-point objects into floating-point registers on Alpha extends them
2826 to 64 bits. Therefore loading a 64-bit object and then storing it as a
2827 32-bit object does not store the low-order 32 bits, as would be the case
2828 for a normal register. Therefore, @file{alpha.h} defines
2829 @code{TARGET_CAN_CHANGE_MODE_CLASS} to return:
2832 (GET_MODE_SIZE (from) == GET_MODE_SIZE (to)
2833 || !reg_classes_intersect_p (FLOAT_REGS, rclass))
2836 Even if storing from a register in mode @var{to} would be valid,
2837 if both @var{from} and @code{raw_reg_mode} for @var{rclass} are wider
2838 than @code{word_mode}, then we must prevent @var{to} narrowing the
2839 mode. This happens when the middle-end assumes that it can load
2840 or store pieces of an @var{N}-word pseudo, and that the pseudo will
2841 eventually be allocated to @var{N} @code{word_mode} hard registers.
2842 Failure to prevent this kind of mode change will result in the
2843 entire @code{raw_reg_mode} being modified instead of the partial
2844 value that the middle-end intended.
2847 @deftypefn {Target Hook} reg_class_t TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS (int, @var{reg_class_t}, @var{reg_class_t})
2848 A target hook which can change allocno class for given pseudo from
2849 allocno and best class calculated by IRA.
2851 The default version of this target hook always returns given class.
2854 @deftypefn {Target Hook} bool TARGET_LRA_P (void)
2855 A target hook which returns true if we use LRA instead of reload pass. The default version of this target hook returns true. New ports should use LRA, and existing ports are encouraged to convert.
2858 @deftypefn {Target Hook} int TARGET_REGISTER_PRIORITY (int)
2859 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.
2862 @deftypefn {Target Hook} bool TARGET_REGISTER_USAGE_LEVELING_P (void)
2863 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.
2866 @deftypefn {Target Hook} bool TARGET_DIFFERENT_ADDR_DISPLACEMENT_P (void)
2867 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.
2870 @deftypefn {Target Hook} bool TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P (rtx @var{subst})
2871 A target hook which returns @code{true} if @var{subst} can't
2872 substitute safely pseudos with equivalent memory values during
2873 register allocation.
2874 The default version of this target hook returns @code{false}.
2875 On most machines, this default should be used. For generally
2876 machines with non orthogonal register usage for addressing, such
2877 as SH, this hook can be used to avoid excessive spilling.
2880 @deftypefn {Target Hook} bool TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT (rtx *@var{disp}, rtx *@var{offset}, machine_mode @var{mode})
2881 A target hook which returns @code{true} if *@var{disp} is
2882 legitimezed to valid address displacement with subtracting *@var{offset}
2883 at memory mode @var{mode}.
2884 The default version of this target hook returns @code{false}.
2885 This hook will benefit machines with limited base plus displacement
2889 @deftypefn {Target Hook} reg_class_t TARGET_SPILL_CLASS (reg_class_t, @var{machine_mode})
2890 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.
2893 @deftypefn {Target Hook} bool TARGET_ADDITIONAL_ALLOCNO_CLASS_P (reg_class_t)
2894 This hook should return @code{true} if given class of registers should be an allocno class in any way. Usually RA uses only one register class from all classes containing the same register set. In some complicated cases, you need to have two or more such classes as allocno ones for RA correct work. Not defining this hook is equivalent to returning @code{false} for all inputs.
2897 @deftypefn {Target Hook} scalar_int_mode TARGET_CSTORE_MODE (enum insn_code @var{icode})
2898 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.
2901 @deftypefn {Target Hook} int TARGET_COMPUTE_PRESSURE_CLASSES (enum reg_class *@var{pressure_classes})
2902 A target hook which lets a backend compute the set of pressure classes to be used by those optimization passes which take register pressure into account, as opposed to letting IRA compute them. It returns the number of register classes stored in the array @var{pressure_classes}.
2905 @node Stack and Calling
2906 @section Stack Layout and Calling Conventions
2907 @cindex calling conventions
2909 @c prevent bad page break with this line
2910 This describes the stack layout and calling conventions.
2914 * Exception Handling::
2919 * Register Arguments::
2921 * Aggregate Return::
2926 * Shrink-wrapping separate components::
2927 * Stack Smashing Protection::
2928 * Miscellaneous Register Hooks::
2932 @subsection Basic Stack Layout
2933 @cindex stack frame layout
2934 @cindex frame layout
2936 @c prevent bad page break with this line
2937 Here is the basic stack layout.
2939 @defmac STACK_GROWS_DOWNWARD
2940 Define this macro to be true if pushing a word onto the stack moves the stack
2941 pointer to a smaller address, and false otherwise.
2944 @defmac STACK_PUSH_CODE
2945 This macro defines the operation used when something is pushed
2946 on the stack. In RTL, a push operation will be
2947 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2949 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2950 and @code{POST_INC}. Which of these is correct depends on
2951 the stack direction and on whether the stack pointer points
2952 to the last item on the stack or whether it points to the
2953 space for the next item on the stack.
2955 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2956 true, which is almost always right, and @code{PRE_INC} otherwise,
2957 which is often wrong.
2960 @defmac FRAME_GROWS_DOWNWARD
2961 Define this macro to nonzero value if the addresses of local variable slots
2962 are at negative offsets from the frame pointer.
2965 @defmac ARGS_GROW_DOWNWARD
2966 Define this macro if successive arguments to a function occupy decreasing
2967 addresses on the stack.
2970 @defmac STARTING_FRAME_OFFSET
2971 Offset from the frame pointer to the first local variable slot to be allocated.
2973 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2974 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2975 Otherwise, it is found by adding the length of the first slot to the
2976 value @code{STARTING_FRAME_OFFSET}.
2977 @c i'm not sure if the above is still correct.. had to change it to get
2978 @c rid of an overfull. --mew 2feb93
2981 @defmac STACK_ALIGNMENT_NEEDED
2982 Define to zero to disable final alignment of the stack during reload.
2983 The nonzero default for this macro is suitable for most ports.
2985 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2986 is a register save block following the local block that doesn't require
2987 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2988 stack alignment and do it in the backend.
2991 @defmac STACK_POINTER_OFFSET
2992 Offset from the stack pointer register to the first location at which
2993 outgoing arguments are placed. If not specified, the default value of
2994 zero is used. This is the proper value for most machines.
2996 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2997 the first location at which outgoing arguments are placed.
3000 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3001 Offset from the argument pointer register to the first argument's
3002 address. On some machines it may depend on the data type of the
3005 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3006 the first argument's address.
3009 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3010 Offset from the stack pointer register to an item dynamically allocated
3011 on the stack, e.g., by @code{alloca}.
3013 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3014 length of the outgoing arguments. The default is correct for most
3015 machines. See @file{function.c} for details.
3018 @defmac INITIAL_FRAME_ADDRESS_RTX
3019 A C expression whose value is RTL representing the address of the initial
3020 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3021 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3022 default value will be used. Define this macro in order to make frame pointer
3023 elimination work in the presence of @code{__builtin_frame_address (count)} and
3024 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3027 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3028 A C expression whose value is RTL representing the address in a stack
3029 frame where the pointer to the caller's frame is stored. Assume that
3030 @var{frameaddr} is an RTL expression for the address of the stack frame
3033 If you don't define this macro, the default is to return the value
3034 of @var{frameaddr}---that is, the stack frame address is also the
3035 address of the stack word that points to the previous frame.
3038 @defmac SETUP_FRAME_ADDRESSES
3039 A C expression that produces the machine-specific code to
3040 setup the stack so that arbitrary frames can be accessed. For example,
3041 on the SPARC, we must flush all of the register windows to the stack
3042 before we can access arbitrary stack frames. You will seldom need to
3043 define this macro. The default is to do nothing.
3046 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3047 This target hook should return an rtx that is used to store
3048 the address of the current frame into the built in @code{setjmp} buffer.
3049 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3050 machines. One reason you may need to define this target hook is if
3051 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3054 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3055 A C expression whose value is RTL representing the value of the frame
3056 address for the current frame. @var{frameaddr} is the frame pointer
3057 of the current frame. This is used for __builtin_frame_address.
3058 You need only define this macro if the frame address is not the same
3059 as the frame pointer. Most machines do not need to define it.
3062 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3063 A C expression whose value is RTL representing the value of the return
3064 address for the frame @var{count} steps up from the current frame, after
3065 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3066 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3067 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero.
3069 The value of the expression must always be the correct address when
3070 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3071 determine the return address of other frames.
3074 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3075 Define this macro to nonzero value if the return address of a particular
3076 stack frame is accessed from the frame pointer of the previous stack
3077 frame. The zero default for this macro is suitable for most ports.
3080 @defmac INCOMING_RETURN_ADDR_RTX
3081 A C expression whose value is RTL representing the location of the
3082 incoming return address at the beginning of any function, before the
3083 prologue. This RTL is either a @code{REG}, indicating that the return
3084 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3087 You only need to define this macro if you want to support call frame
3088 debugging information like that provided by DWARF 2.
3090 If this RTL is a @code{REG}, you should also define
3091 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3094 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3095 A C expression whose value is an integer giving a DWARF 2 column
3096 number that may be used as an alternative return column. The column
3097 must not correspond to any gcc hard register (that is, it must not
3098 be in the range of @code{DWARF_FRAME_REGNUM}).
3100 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3101 general register, but an alternative column needs to be used for signal
3102 frames. Some targets have also used different frame return columns
3106 @defmac DWARF_ZERO_REG
3107 A C expression whose value is an integer giving a DWARF 2 register
3108 number that is considered to always have the value zero. This should
3109 only be defined if the target has an architected zero register, and
3110 someone decided it was a good idea to use that register number to
3111 terminate the stack backtrace. New ports should avoid this.
3114 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3115 This target hook allows the backend to emit frame-related insns that
3116 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3117 info engine will invoke it on insns of the form
3119 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3123 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3125 to let the backend emit the call frame instructions. @var{label} is
3126 the CFI label attached to the insn, @var{pattern} is the pattern of
3127 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3130 @defmac INCOMING_FRAME_SP_OFFSET
3131 A C expression whose value is an integer giving the offset, in bytes,
3132 from the value of the stack pointer register to the top of the stack
3133 frame at the beginning of any function, before the prologue. The top of
3134 the frame is defined to be the value of the stack pointer in the
3135 previous frame, just before the call instruction.
3137 You only need to define this macro if you want to support call frame
3138 debugging information like that provided by DWARF 2.
3141 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3142 A C expression whose value is an integer giving the offset, in bytes,
3143 from the argument pointer to the canonical frame address (cfa). The
3144 final value should coincide with that calculated by
3145 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3146 during virtual register instantiation.
3148 The default value for this macro is
3149 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3150 which is correct for most machines; in general, the arguments are found
3151 immediately before the stack frame. Note that this is not the case on
3152 some targets that save registers into the caller's frame, such as SPARC
3153 and rs6000, and so such targets need to define this macro.
3155 You only need to define this macro if the default is incorrect, and you
3156 want to support call frame debugging information like that provided by
3160 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3161 If defined, a C expression whose value is an integer giving the offset
3162 in bytes from the frame pointer to the canonical frame address (cfa).
3163 The final value should coincide with that calculated by
3164 @code{INCOMING_FRAME_SP_OFFSET}.
3166 Normally the CFA is calculated as an offset from the argument pointer,
3167 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3168 variable due to the ABI, this may not be possible. If this macro is
3169 defined, it implies that the virtual register instantiation should be
3170 based on the frame pointer instead of the argument pointer. Only one
3171 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3175 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3176 If defined, a C expression whose value is an integer giving the offset
3177 in bytes from the canonical frame address (cfa) to the frame base used
3178 in DWARF 2 debug information. The default is zero. A different value
3179 may reduce the size of debug information on some ports.
3182 @node Exception Handling
3183 @subsection Exception Handling Support
3184 @cindex exception handling
3186 @defmac EH_RETURN_DATA_REGNO (@var{N})
3187 A C expression whose value is the @var{N}th register number used for
3188 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3189 @var{N} registers are usable.
3191 The exception handling library routines communicate with the exception
3192 handlers via a set of agreed upon registers. Ideally these registers
3193 should be call-clobbered; it is possible to use call-saved registers,
3194 but may negatively impact code size. The target must support at least
3195 2 data registers, but should define 4 if there are enough free registers.
3197 You must define this macro if you want to support call frame exception
3198 handling like that provided by DWARF 2.
3201 @defmac EH_RETURN_STACKADJ_RTX
3202 A C expression whose value is RTL representing a location in which
3203 to store a stack adjustment to be applied before function return.
3204 This is used to unwind the stack to an exception handler's call frame.
3205 It will be assigned zero on code paths that return normally.
3207 Typically this is a call-clobbered hard register that is otherwise
3208 untouched by the epilogue, but could also be a stack slot.
3210 Do not define this macro if the stack pointer is saved and restored
3211 by the regular prolog and epilog code in the call frame itself; in
3212 this case, the exception handling library routines will update the
3213 stack location to be restored in place. Otherwise, you must define
3214 this macro if you want to support call frame exception handling like
3215 that provided by DWARF 2.
3218 @defmac EH_RETURN_HANDLER_RTX
3219 A C expression whose value is RTL representing a location in which
3220 to store the address of an exception handler to which we should
3221 return. It will not be assigned on code paths that return normally.
3223 Typically this is the location in the call frame at which the normal
3224 return address is stored. For targets that return by popping an
3225 address off the stack, this might be a memory address just below
3226 the @emph{target} call frame rather than inside the current call
3227 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3228 been assigned, so it may be used to calculate the location of the
3231 Some targets have more complex requirements than storing to an
3232 address calculable during initial code generation. In that case
3233 the @code{eh_return} instruction pattern should be used instead.
3235 If you want to support call frame exception handling, you must
3236 define either this macro or the @code{eh_return} instruction pattern.
3239 @defmac RETURN_ADDR_OFFSET
3240 If defined, an integer-valued C expression for which rtl will be generated
3241 to add it to the exception handler address before it is searched in the
3242 exception handling tables, and to subtract it again from the address before
3243 using it to return to the exception handler.
3246 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3247 This macro chooses the encoding of pointers embedded in the exception
3248 handling sections. If at all possible, this should be defined such
3249 that the exception handling section will not require dynamic relocations,
3250 and so may be read-only.
3252 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3253 @var{global} is true if the symbol may be affected by dynamic relocations.
3254 The macro should return a combination of the @code{DW_EH_PE_*} defines
3255 as found in @file{dwarf2.h}.
3257 If this macro is not defined, pointers will not be encoded but
3258 represented directly.
3261 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3262 This macro allows the target to emit whatever special magic is required
3263 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3264 Generic code takes care of pc-relative and indirect encodings; this must
3265 be defined if the target uses text-relative or data-relative encodings.
3267 This is a C statement that branches to @var{done} if the format was
3268 handled. @var{encoding} is the format chosen, @var{size} is the number
3269 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3273 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3274 This macro allows the target to add CPU and operating system specific
3275 code to the call-frame unwinder for use when there is no unwind data
3276 available. The most common reason to implement this macro is to unwind
3277 through signal frames.
3279 This macro is called from @code{uw_frame_state_for} in
3280 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3281 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3282 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3283 for the address of the code being executed and @code{context->cfa} for
3284 the stack pointer value. If the frame can be decoded, the register
3285 save addresses should be updated in @var{fs} and the macro should
3286 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3287 the macro should evaluate to @code{_URC_END_OF_STACK}.
3289 For proper signal handling in Java this macro is accompanied by
3290 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3293 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3294 This macro allows the target to add operating system specific code to the
3295 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3296 usually used for signal or interrupt frames.
3298 This macro is called from @code{uw_update_context} in libgcc's
3299 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3300 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3301 for the abi and context in the @code{.unwabi} directive. If the
3302 @code{.unwabi} directive can be handled, the register save addresses should
3303 be updated in @var{fs}.
3306 @defmac TARGET_USES_WEAK_UNWIND_INFO
3307 A C expression that evaluates to true if the target requires unwind
3308 info to be given comdat linkage. Define it to be @code{1} if comdat
3309 linkage is necessary. The default is @code{0}.
3312 @node Stack Checking
3313 @subsection Specifying How Stack Checking is Done
3315 GCC will check that stack references are within the boundaries of the
3316 stack, if the option @option{-fstack-check} is specified, in one of
3321 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3322 will assume that you have arranged for full stack checking to be done
3323 at appropriate places in the configuration files. GCC will not do
3324 other special processing.
3327 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3328 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3329 that you have arranged for static stack checking (checking of the
3330 static stack frame of functions) to be done at appropriate places
3331 in the configuration files. GCC will only emit code to do dynamic
3332 stack checking (checking on dynamic stack allocations) using the third
3336 If neither of the above are true, GCC will generate code to periodically
3337 ``probe'' the stack pointer using the values of the macros defined below.
3340 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3341 GCC will change its allocation strategy for large objects if the option
3342 @option{-fstack-check} is specified: they will always be allocated
3343 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3345 @defmac STACK_CHECK_BUILTIN
3346 A nonzero value if stack checking is done by the configuration files in a
3347 machine-dependent manner. You should define this macro if stack checking
3348 is required by the ABI of your machine or if you would like to do stack
3349 checking in some more efficient way than the generic approach. The default
3350 value of this macro is zero.
3353 @defmac STACK_CHECK_STATIC_BUILTIN
3354 A nonzero value if static stack checking is done by the configuration files
3355 in a machine-dependent manner. You should define this macro if you would
3356 like to do static stack checking in some more efficient way than the generic
3357 approach. The default value of this macro is zero.
3360 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3361 An integer specifying the interval at which GCC must generate stack probe
3362 instructions, defined as 2 raised to this integer. You will normally
3363 define this macro so that the interval be no larger than the size of
3364 the ``guard pages'' at the end of a stack area. The default value
3365 of 12 (4096-byte interval) is suitable for most systems.
3368 @defmac STACK_CHECK_MOVING_SP
3369 An integer which is nonzero if GCC should move the stack pointer page by page
3370 when doing probes. This can be necessary on systems where the stack pointer
3371 contains the bottom address of the memory area accessible to the executing
3372 thread at any point in time. In this situation an alternate signal stack
3373 is required in order to be able to recover from a stack overflow. The
3374 default value of this macro is zero.
3377 @defmac STACK_CHECK_PROTECT
3378 The number of bytes of stack needed to recover from a stack overflow, for
3379 languages where such a recovery is supported. The default value of 4KB/8KB
3380 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3381 8KB/12KB with other exception handling mechanisms should be adequate for most
3382 architectures and operating systems.
3385 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3386 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3387 in the opposite case.
3389 @defmac STACK_CHECK_MAX_FRAME_SIZE
3390 The maximum size of a stack frame, in bytes. GCC will generate probe
3391 instructions in non-leaf functions to ensure at least this many bytes of
3392 stack are available. If a stack frame is larger than this size, stack
3393 checking will not be reliable and GCC will issue a warning. The
3394 default is chosen so that GCC only generates one instruction on most
3395 systems. You should normally not change the default value of this macro.
3398 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3399 GCC uses this value to generate the above warning message. It
3400 represents the amount of fixed frame used by a function, not including
3401 space for any callee-saved registers, temporaries and user variables.
3402 You need only specify an upper bound for this amount and will normally
3403 use the default of four words.
3406 @defmac STACK_CHECK_MAX_VAR_SIZE
3407 The maximum size, in bytes, of an object that GCC will place in the
3408 fixed area of the stack frame when the user specifies
3409 @option{-fstack-check}.
3410 GCC computed the default from the values of the above macros and you will
3411 normally not need to override that default.
3415 @node Frame Registers
3416 @subsection Registers That Address the Stack Frame
3418 @c prevent bad page break with this line
3419 This discusses registers that address the stack frame.
3421 @defmac STACK_POINTER_REGNUM
3422 The register number of the stack pointer register, which must also be a
3423 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3424 the hardware determines which register this is.
3427 @defmac FRAME_POINTER_REGNUM
3428 The register number of the frame pointer register, which is used to
3429 access automatic variables in the stack frame. On some machines, the
3430 hardware determines which register this is. On other machines, you can
3431 choose any register you wish for this purpose.
3434 @defmac HARD_FRAME_POINTER_REGNUM
3435 On some machines the offset between the frame pointer and starting
3436 offset of the automatic variables is not known until after register
3437 allocation has been done (for example, because the saved registers are
3438 between these two locations). On those machines, define
3439 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3440 be used internally until the offset is known, and define
3441 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3442 used for the frame pointer.
3444 You should define this macro only in the very rare circumstances when it
3445 is not possible to calculate the offset between the frame pointer and
3446 the automatic variables until after register allocation has been
3447 completed. When this macro is defined, you must also indicate in your
3448 definition of @code{ELIMINABLE_REGS} how to eliminate
3449 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3450 or @code{STACK_POINTER_REGNUM}.
3452 Do not define this macro if it would be the same as
3453 @code{FRAME_POINTER_REGNUM}.
3456 @defmac ARG_POINTER_REGNUM
3457 The register number of the arg pointer register, which is used to access
3458 the function's argument list. On some machines, this is the same as the
3459 frame pointer register. On some machines, the hardware determines which
3460 register this is. On other machines, you can choose any register you
3461 wish for this purpose. If this is not the same register as the frame
3462 pointer register, then you must mark it as a fixed register according to
3463 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3464 (@pxref{Elimination}).
3467 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3468 Define this to a preprocessor constant that is nonzero if
3469 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3470 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3471 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3472 definition is not suitable for use in preprocessor conditionals.
3475 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3476 Define this to a preprocessor constant that is nonzero if
3477 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3478 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3479 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3480 definition is not suitable for use in preprocessor conditionals.
3483 @defmac RETURN_ADDRESS_POINTER_REGNUM
3484 The register number of the return address pointer register, which is used to
3485 access the current function's return address from the stack. On some
3486 machines, the return address is not at a fixed offset from the frame
3487 pointer or stack pointer or argument pointer. This register can be defined
3488 to point to the return address on the stack, and then be converted by
3489 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3491 Do not define this macro unless there is no other way to get the return
3492 address from the stack.
3495 @defmac STATIC_CHAIN_REGNUM
3496 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3497 Register numbers used for passing a function's static chain pointer. If
3498 register windows are used, the register number as seen by the called
3499 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3500 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3501 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3504 The static chain register need not be a fixed register.
3506 If the static chain is passed in memory, these macros should not be
3507 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3510 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl_or_type}, bool @var{incoming_p})
3511 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3512 targets that may use different static chain locations for different
3513 nested functions. This may be required if the target has function
3514 attributes that affect the calling conventions of the function and
3515 those calling conventions use different static chain locations.
3517 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3519 If the static chain is passed in memory, this hook should be used to
3520 provide rtx giving @code{mem} expressions that denote where they are stored.
3521 Often the @code{mem} expression as seen by the caller will be at an offset
3522 from the stack pointer and the @code{mem} expression as seen by the callee
3523 will be at an offset from the frame pointer.
3524 @findex stack_pointer_rtx
3525 @findex frame_pointer_rtx
3526 @findex arg_pointer_rtx
3527 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3528 @code{arg_pointer_rtx} will have been initialized and should be used
3529 to refer to those items.
3532 @defmac DWARF_FRAME_REGISTERS
3533 This macro specifies the maximum number of hard registers that can be
3534 saved in a call frame. This is used to size data structures used in
3535 DWARF2 exception handling.
3537 Prior to GCC 3.0, this macro was needed in order to establish a stable
3538 exception handling ABI in the face of adding new hard registers for ISA
3539 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3540 in the number of hard registers. Nevertheless, this macro can still be
3541 used to reduce the runtime memory requirements of the exception handling
3542 routines, which can be substantial if the ISA contains a lot of
3543 registers that are not call-saved.
3545 If this macro is not defined, it defaults to
3546 @code{FIRST_PSEUDO_REGISTER}.
3549 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3551 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3552 for backward compatibility in pre GCC 3.0 compiled code.
3554 If this macro is not defined, it defaults to
3555 @code{DWARF_FRAME_REGISTERS}.
3558 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3560 Define this macro if the target's representation for dwarf registers
3561 is different than the internal representation for unwind column.
3562 Given a dwarf register, this macro should return the internal unwind
3563 column number to use instead.
3566 @defmac DWARF_FRAME_REGNUM (@var{regno})
3568 Define this macro if the target's representation for dwarf registers
3569 used in .eh_frame or .debug_frame is different from that used in other
3570 debug info sections. Given a GCC hard register number, this macro
3571 should return the .eh_frame register number. The default is
3572 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3576 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3578 Define this macro to map register numbers held in the call frame info
3579 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3580 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3581 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3582 return @code{@var{regno}}.
3586 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3588 Define this macro if the target stores register values as
3589 @code{_Unwind_Word} type in unwind context. It should be defined if
3590 target register size is larger than the size of @code{void *}. The
3591 default is to store register values as @code{void *} type.
3595 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3597 Define this macro to be 1 if the target always uses extended unwind
3598 context with version, args_size and by_value fields. If it is undefined,
3599 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3600 defined and 0 otherwise.
3605 @subsection Eliminating Frame Pointer and Arg Pointer
3607 @c prevent bad page break with this line
3608 This is about eliminating the frame pointer and arg pointer.
3610 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3611 This target hook should return @code{true} if a function must have and use
3612 a frame pointer. This target hook is called in the reload pass. If its return
3613 value is @code{true} the function will have a frame pointer.
3615 This target hook can in principle examine the current function and decide
3616 according to the facts, but on most machines the constant @code{false} or the
3617 constant @code{true} suffices. Use @code{false} when the machine allows code
3618 to be generated with no frame pointer, and doing so saves some time or space.
3619 Use @code{true} when there is no possible advantage to avoiding a frame
3622 In certain cases, the compiler does not know how to produce valid code
3623 without a frame pointer. The compiler recognizes those cases and
3624 automatically gives the function a frame pointer regardless of what
3625 @code{targetm.frame_pointer_required} returns. You don't need to worry about
3628 In a function that does not require a frame pointer, the frame pointer
3629 register can be allocated for ordinary usage, unless you mark it as a
3630 fixed register. See @code{FIXED_REGISTERS} for more information.
3632 Default return value is @code{false}.
3635 @defmac ELIMINABLE_REGS
3636 This macro specifies a table of register pairs used to eliminate
3637 unneeded registers that point into the stack frame.
3639 The definition of this macro is a list of structure initializations, each
3640 of which specifies an original and replacement register.
3642 On some machines, the position of the argument pointer is not known until
3643 the compilation is completed. In such a case, a separate hard register
3644 must be used for the argument pointer. This register can be eliminated by
3645 replacing it with either the frame pointer or the argument pointer,
3646 depending on whether or not the frame pointer has been eliminated.
3648 In this case, you might specify:
3650 #define ELIMINABLE_REGS \
3651 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3652 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3653 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3656 Note that the elimination of the argument pointer with the stack pointer is
3657 specified first since that is the preferred elimination.
3660 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3661 This target hook should return @code{true} if the compiler is allowed to
3662 try to replace register number @var{from_reg} with register number
3663 @var{to_reg}. This target hook will usually be @code{true}, since most of the
3664 cases preventing register elimination are things that the compiler already
3667 Default return value is @code{true}.
3670 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3671 This macro returns the initial difference between the specified pair
3672 of registers. The value would be computed from information
3673 such as the result of @code{get_frame_size ()} and the tables of
3674 registers @code{df_regs_ever_live_p} and @code{call_used_regs}.
3677 @deftypefn {Target Hook} void TARGET_COMPUTE_FRAME_LAYOUT (void)
3678 This target hook is called once each time the frame layout needs to be
3679 recalculated. The calculations can be cached by the target and can then
3680 be used by @code{INITIAL_ELIMINATION_OFFSET} instead of re-computing the
3681 layout on every invocation of that hook. This is particularly useful
3682 for targets that have an expensive frame layout function. Implementing
3683 this callback is optional.
3686 @node Stack Arguments
3687 @subsection Passing Function Arguments on the Stack
3688 @cindex arguments on stack
3689 @cindex stack arguments
3691 The macros in this section control how arguments are passed
3692 on the stack. See the following section for other macros that
3693 control passing certain arguments in registers.
3695 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3696 This target hook returns @code{true} if an argument declared in a
3697 prototype as an integral type smaller than @code{int} should actually be
3698 passed as an @code{int}. In addition to avoiding errors in certain
3699 cases of mismatch, it also makes for better code on certain machines.
3700 The default is to not promote prototypes.
3704 A C expression. If nonzero, push insns will be used to pass
3706 If the target machine does not have a push instruction, set it to zero.
3707 That directs GCC to use an alternate strategy: to
3708 allocate the entire argument block and then store the arguments into
3709 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3712 @defmac PUSH_ARGS_REVERSED
3713 A C expression. If nonzero, function arguments will be evaluated from
3714 last to first, rather than from first to last. If this macro is not
3715 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3716 and args grow in opposite directions, and 0 otherwise.
3719 @defmac PUSH_ROUNDING (@var{npushed})
3720 A C expression that is the number of bytes actually pushed onto the
3721 stack when an instruction attempts to push @var{npushed} bytes.
3723 On some machines, the definition
3726 #define PUSH_ROUNDING(BYTES) (BYTES)
3730 will suffice. But on other machines, instructions that appear
3731 to push one byte actually push two bytes in an attempt to maintain
3732 alignment. Then the definition should be
3735 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3738 If the value of this macro has a type, it should be an unsigned type.
3741 @findex outgoing_args_size
3742 @findex crtl->outgoing_args_size
3743 @defmac ACCUMULATE_OUTGOING_ARGS
3744 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3745 will be computed and placed into
3746 @code{crtl->outgoing_args_size}. No space will be pushed
3747 onto the stack for each call; instead, the function prologue should
3748 increase the stack frame size by this amount.
3750 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3754 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3755 Define this macro if functions should assume that stack space has been
3756 allocated for arguments even when their values are passed in
3759 The value of this macro is the size, in bytes, of the area reserved for
3760 arguments passed in registers for the function represented by @var{fndecl},
3761 which can be zero if GCC is calling a library function.
3762 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3765 This space can be allocated by the caller, or be a part of the
3766 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3769 @c above is overfull. not sure what to do. --mew 5feb93 did
3770 @c something, not sure if it looks good. --mew 10feb93
3772 @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3773 Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3774 Define this macro if space guaranteed when compiling a function body
3775 is different to space required when making a call, a situation that
3776 can arise with K&R style function definitions.
3779 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3780 Define this to a nonzero value if it is the responsibility of the
3781 caller to allocate the area reserved for arguments passed in registers
3782 when calling a function of @var{fntype}. @var{fntype} may be NULL
3783 if the function called is a library function.
3785 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3786 whether the space for these arguments counts in the value of
3787 @code{crtl->outgoing_args_size}.
3790 @defmac STACK_PARMS_IN_REG_PARM_AREA
3791 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3792 stack parameters don't skip the area specified by it.
3793 @c i changed this, makes more sens and it should have taken care of the
3794 @c overfull.. not as specific, tho. --mew 5feb93
3796 Normally, when a parameter is not passed in registers, it is placed on the
3797 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3798 suppresses this behavior and causes the parameter to be passed on the
3799 stack in its natural location.
3802 @deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size})
3803 This target hook returns the number of bytes of its own arguments that
3804 a function pops on returning, or 0 if the function pops no arguments
3805 and the caller must therefore pop them all after the function returns.
3807 @var{fundecl} is a C variable whose value is a tree node that describes
3808 the function in question. Normally it is a node of type
3809 @code{FUNCTION_DECL} that describes the declaration of the function.
3810 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3812 @var{funtype} is a C variable whose value is a tree node that
3813 describes the function in question. Normally it is a node of type
3814 @code{FUNCTION_TYPE} that describes the data type of the function.
3815 From this it is possible to obtain the data types of the value and
3816 arguments (if known).
3818 When a call to a library function is being considered, @var{fundecl}
3819 will contain an identifier node for the library function. Thus, if
3820 you need to distinguish among various library functions, you can do so
3821 by their names. Note that ``library function'' in this context means
3822 a function used to perform arithmetic, whose name is known specially
3823 in the compiler and was not mentioned in the C code being compiled.
3825 @var{size} is the number of bytes of arguments passed on the
3826 stack. If a variable number of bytes is passed, it is zero, and
3827 argument popping will always be the responsibility of the calling function.
3829 On the VAX, all functions always pop their arguments, so the definition
3830 of this macro is @var{size}. On the 68000, using the standard
3831 calling convention, no functions pop their arguments, so the value of
3832 the macro is always 0 in this case. But an alternative calling
3833 convention is available in which functions that take a fixed number of
3834 arguments pop them but other functions (such as @code{printf}) pop
3835 nothing (the caller pops all). When this convention is in use,
3836 @var{funtype} is examined to determine whether a function takes a fixed
3837 number of arguments.
3840 @defmac CALL_POPS_ARGS (@var{cum})
3841 A C expression that should indicate the number of bytes a call sequence
3842 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3843 when compiling a function call.
3845 @var{cum} is the variable in which all arguments to the called function
3846 have been accumulated.
3848 On certain architectures, such as the SH5, a call trampoline is used
3849 that pops certain registers off the stack, depending on the arguments
3850 that have been passed to the function. Since this is a property of the
3851 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3855 @node Register Arguments
3856 @subsection Passing Arguments in Registers
3857 @cindex arguments in registers
3858 @cindex registers arguments
3860 This section describes the macros which let you control how various
3861 types of arguments are passed in registers or how they are arranged in
3864 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3865 Return an RTX indicating whether a function argument is passed in a
3866 register and if so, which register.
3868 The arguments are @var{ca}, which summarizes all the previous
3869 arguments; @var{mode}, the machine mode of the argument; @var{type},
3870 the data type of the argument as a tree node or 0 if that is not known
3871 (which happens for C support library functions); and @var{named},
3872 which is @code{true} for an ordinary argument and @code{false} for
3873 nameless arguments that correspond to @samp{@dots{}} in the called
3874 function's prototype. @var{type} can be an incomplete type if a
3875 syntax error has previously occurred.
3877 The return value is usually either a @code{reg} RTX for the hard
3878 register in which to pass the argument, or zero to pass the argument
3881 The return value can be a @code{const_int} which means argument is
3882 passed in a target specific slot with specified number. Target hooks
3883 should be used to store or load argument in such case. See
3884 @code{TARGET_STORE_BOUNDS_FOR_ARG} and @code{TARGET_LOAD_BOUNDS_FOR_ARG}
3885 for more information.
3887 The value of the expression can also be a @code{parallel} RTX@. This is
3888 used when an argument is passed in multiple locations. The mode of the
3889 @code{parallel} should be the mode of the entire argument. The
3890 @code{parallel} holds any number of @code{expr_list} pairs; each one
3891 describes where part of the argument is passed. In each
3892 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3893 register in which to pass this part of the argument, and the mode of the
3894 register RTX indicates how large this part of the argument is. The
3895 second operand of the @code{expr_list} is a @code{const_int} which gives
3896 the offset in bytes into the entire argument of where this part starts.
3897 As a special exception the first @code{expr_list} in the @code{parallel}
3898 RTX may have a first operand of zero. This indicates that the entire
3899 argument is also stored on the stack.
3901 The last time this hook is called, it is called with @code{MODE ==
3902 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3903 pattern as operands 2 and 3 respectively.
3905 @cindex @file{stdarg.h} and register arguments
3906 The usual way to make the ISO library @file{stdarg.h} work on a
3907 machine where some arguments are usually passed in registers, is to
3908 cause nameless arguments to be passed on the stack instead. This is
3909 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
3910 @var{named} is @code{false}.
3912 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
3913 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
3914 You may use the hook @code{targetm.calls.must_pass_in_stack}
3915 in the definition of this macro to determine if this argument is of a
3916 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3917 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
3918 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3919 defined, the argument will be computed in the stack and then loaded into
3923 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (machine_mode @var{mode}, const_tree @var{type})
3924 This target hook should return @code{true} if we should not pass @var{type}
3925 solely in registers. The file @file{expr.h} defines a
3926 definition that is usually appropriate, refer to @file{expr.h} for additional
3930 @deftypefn {Target Hook} rtx TARGET_FUNCTION_INCOMING_ARG (cumulative_args_t @var{ca}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3931 Define this hook if the caller and callee on the target have different
3932 views of where arguments are passed. Also define this hook if there are
3933 functions that are never directly called, but are invoked by the hardware
3934 and which have nonstandard calling conventions.
3936 In this case @code{TARGET_FUNCTION_ARG} computes the register in
3937 which the caller passes the value, and
3938 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
3939 fashion to tell the function being called where the arguments will
3942 @code{TARGET_FUNCTION_INCOMING_ARG} can also return arbitrary address
3943 computation using hard register, which can be forced into a register,
3944 so that it can be used to pass special arguments.
3946 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
3947 @code{TARGET_FUNCTION_ARG} serves both purposes.
3950 @deftypefn {Target Hook} bool TARGET_USE_PSEUDO_PIC_REG (void)
3951 This hook should return 1 in case pseudo register should be created
3952 for pic_offset_table_rtx during function expand.
3955 @deftypefn {Target Hook} void TARGET_INIT_PIC_REG (void)
3956 Perform a target dependent initialization of pic_offset_table_rtx.
3957 This hook is called at the start of register allocation.
3960 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, machine_mode @var{mode}, tree @var{type}, bool @var{named})
3961 This target hook returns the number of bytes at the beginning of an
3962 argument that must be put in registers. The value must be zero for
3963 arguments that are passed entirely in registers or that are entirely
3964 pushed on the stack.
3966 On some machines, certain arguments must be passed partially in
3967 registers and partially in memory. On these machines, typically the
3968 first few words of arguments are passed in registers, and the rest
3969 on the stack. If a multi-word argument (a @code{double} or a
3970 structure) crosses that boundary, its first few words must be passed
3971 in registers and the rest must be pushed. This macro tells the
3972 compiler when this occurs, and how many bytes should go in registers.
3974 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
3975 register to be used by the caller for this argument; likewise
3976 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
3979 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (cumulative_args_t @var{cum}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3980 This target hook should return @code{true} if an argument at the
3981 position indicated by @var{cum} should be passed by reference. This
3982 predicate is queried after target independent reasons for being
3983 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3985 If the hook returns true, a copy of that argument is made in memory and a
3986 pointer to the argument is passed instead of the argument itself.
3987 The pointer is passed in whatever way is appropriate for passing a pointer
3991 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3992 The function argument described by the parameters to this hook is
3993 known to be passed by reference. The hook should return true if the
3994 function argument should be copied by the callee instead of copied
3997 For any argument for which the hook returns true, if it can be
3998 determined that the argument is not modified, then a copy need
4001 The default version of this hook always returns false.
4004 @defmac CUMULATIVE_ARGS
4005 A C type for declaring a variable that is used as the first argument
4006 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4007 target machines, the type @code{int} suffices and can hold the number
4008 of bytes of argument so far.
4010 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4011 arguments that have been passed on the stack. The compiler has other
4012 variables to keep track of that. For target machines on which all
4013 arguments are passed on the stack, there is no need to store anything in
4014 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4015 should not be empty, so use @code{int}.
4018 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4019 If defined, this macro is called before generating any code for a
4020 function, but after the @var{cfun} descriptor for the function has been
4021 created. The back end may use this macro to update @var{cfun} to
4022 reflect an ABI other than that which would normally be used by default.
4023 If the compiler is generating code for a compiler-generated function,
4024 @var{fndecl} may be @code{NULL}.
4027 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4028 A C statement (sans semicolon) for initializing the variable
4029 @var{cum} for the state at the beginning of the argument list. The
4030 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4031 is the tree node for the data type of the function which will receive
4032 the args, or 0 if the args are to a compiler support library function.
4033 For direct calls that are not libcalls, @var{fndecl} contain the
4034 declaration node of the function. @var{fndecl} is also set when
4035 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4036 being compiled. @var{n_named_args} is set to the number of named
4037 arguments, including a structure return address if it is passed as a
4038 parameter, when making a call. When processing incoming arguments,
4039 @var{n_named_args} is set to @minus{}1.
4041 When processing a call to a compiler support library function,
4042 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4043 contains the name of the function, as a string. @var{libname} is 0 when
4044 an ordinary C function call is being processed. Thus, each time this
4045 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4046 never both of them at once.
4049 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4050 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4051 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4052 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4053 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4054 0)} is used instead.
4057 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4058 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4059 finding the arguments for the function being compiled. If this macro is
4060 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4062 The value passed for @var{libname} is always 0, since library routines
4063 with special calling conventions are never compiled with GCC@. The
4064 argument @var{libname} exists for symmetry with
4065 @code{INIT_CUMULATIVE_ARGS}.
4066 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4067 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4070 @deftypefn {Target Hook} void TARGET_FUNCTION_ARG_ADVANCE (cumulative_args_t @var{ca}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4071 This hook updates the summarizer variable pointed to by @var{ca} to
4072 advance past an argument in the argument list. The values @var{mode},
4073 @var{type} and @var{named} describe that argument. Once this is done,
4074 the variable @var{cum} is suitable for analyzing the @emph{following}
4075 argument with @code{TARGET_FUNCTION_ARG}, etc.
4077 This hook need not do anything if the argument in question was passed
4078 on the stack. The compiler knows how to track the amount of stack space
4079 used for arguments without any special help.
4082 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4083 If defined, a C expression that is the number of bytes to add to the
4084 offset of the argument passed in memory. This is needed for the SPU,
4085 which passes @code{char} and @code{short} arguments in the preferred
4086 slot that is in the middle of the quad word instead of starting at the
4090 @deftypefn {Target Hook} pad_direction TARGET_FUNCTION_ARG_PADDING (machine_mode @var{mode}, const_tree @var{type})
4091 This hook determines whether, and in which direction, to pad out
4092 an argument of mode @var{mode} and type @var{type}. It returns
4093 @code{PAD_UPWARD} to insert padding above the argument, @code{PAD_DOWNWARD}
4094 to insert padding below the argument, or @code{PAD_NONE} to inhibit padding.
4096 The @emph{amount} of padding is not controlled by this hook, but by
4097 @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is always just enough
4098 to reach the next multiple of that boundary.
4100 This hook has a default definition that is right for most systems.
4101 For little-endian machines, the default is to pad upward. For
4102 big-endian machines, the default is to pad downward for an argument of
4103 constant size shorter than an @code{int}, and upward otherwise.
4106 @defmac PAD_VARARGS_DOWN
4107 If defined, a C expression which determines whether the default
4108 implementation of va_arg will attempt to pad down before reading the
4109 next argument, if that argument is smaller than its aligned space as
4110 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4111 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4114 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4115 Specify padding for the last element of a block move between registers and
4116 memory. @var{first} is nonzero if this is the only element. Defining this
4117 macro allows better control of register function parameters on big-endian
4118 machines, without using @code{PARALLEL} rtl. In particular,
4119 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4120 registers, as there is no longer a "wrong" part of a register; For example,
4121 a three byte aggregate may be passed in the high part of a register if so
4125 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (machine_mode @var{mode}, const_tree @var{type})
4126 This hook returns the alignment boundary, in bits, of an argument
4127 with the specified mode and type. The default hook returns
4128 @code{PARM_BOUNDARY} for all arguments.
4131 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_ROUND_BOUNDARY (machine_mode @var{mode}, const_tree @var{type})
4132 Normally, the size of an argument is rounded up to @code{PARM_BOUNDARY},
4133 which is the default value for this hook. You can define this hook to
4134 return a different value if an argument size must be rounded to a larger
4138 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4139 A C expression that is nonzero if @var{regno} is the number of a hard
4140 register in which function arguments are sometimes passed. This does
4141 @emph{not} include implicit arguments such as the static chain and
4142 the structure-value address. On many machines, no registers can be
4143 used for this purpose since all function arguments are pushed on the
4147 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4148 This hook should return true if parameter of type @var{type} are passed
4149 as two scalar parameters. By default, GCC will attempt to pack complex
4150 arguments into the target's word size. Some ABIs require complex arguments
4151 to be split and treated as their individual components. For example, on
4152 AIX64, complex floats should be passed in a pair of floating point
4153 registers, even though a complex float would fit in one 64-bit floating
4156 The default value of this hook is @code{NULL}, which is treated as always
4160 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4161 This hook returns a type node for @code{va_list} for the target.
4162 The default version of the hook returns @code{void*}.
4165 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4166 This target hook is used in function @code{c_common_nodes_and_builtins}
4167 to iterate through the target specific builtin types for va_list. The
4168 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4169 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4171 The arguments @var{pname} and @var{ptree} are used to store the result of
4172 this macro and are set to the name of the va_list builtin type and its
4174 If the return value of this macro is zero, then there is no more element.
4175 Otherwise the @var{IDX} should be increased for the next call of this
4176 macro to iterate through all types.
4179 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4180 This hook returns the va_list type of the calling convention specified by
4182 The default version of this hook returns @code{va_list_type_node}.
4185 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4186 This hook returns the va_list type of the calling convention specified by the
4187 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4191 @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})
4192 This hook performs target-specific gimplification of
4193 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4194 arguments to @code{va_arg}; the latter two are as in
4195 @code{gimplify.c:gimplify_expr}.
4198 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (scalar_int_mode @var{mode})
4199 Define this to return nonzero if the port can handle pointers
4200 with machine mode @var{mode}. The default version of this
4201 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4204 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (struct ao_ref *@var{ref})
4205 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.
4208 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (scalar_mode @var{mode})
4209 Define this to return nonzero if the port is prepared to handle
4210 insns involving scalar mode @var{mode}. For a scalar mode to be
4211 considered supported, all the basic arithmetic and comparisons
4214 The default version of this hook returns true for any mode
4215 required to handle the basic C types (as defined by the port).
4216 Included here are the double-word arithmetic supported by the
4217 code in @file{optabs.c}.
4220 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (machine_mode @var{mode})
4221 Define this to return nonzero if the port is prepared to handle
4222 insns involving vector mode @var{mode}. At the very least, it
4223 must have move patterns for this mode.
4226 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4227 Return true if GCC should try to use a scalar mode to store an array
4228 of @var{nelems} elements, given that each element has mode @var{mode}.
4229 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4230 and allows GCC to use any defined integer mode.
4232 One use of this hook is to support vector load and store operations
4233 that operate on several homogeneous vectors. For example, ARM NEON
4234 has operations like:
4237 int8x8x3_t vld3_s8 (const int8_t *)
4240 where the return type is defined as:
4243 typedef struct int8x8x3_t
4249 If this hook allows @code{val} to have a scalar mode, then
4250 @code{int8x8x3_t} can have the same mode. GCC can then store
4251 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4254 @deftypefn {Target Hook} bool TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P (scalar_float_mode @var{mode})
4255 Define this to return nonzero if libgcc provides support for the
4256 floating-point mode @var{mode}, which is known to pass
4257 @code{TARGET_SCALAR_MODE_SUPPORTED_P}. The default version of this
4258 hook returns true for all of @code{SFmode}, @code{DFmode},
4259 @code{XFmode} and @code{TFmode}, if such modes exist.
4262 @deftypefn {Target Hook} opt_scalar_float_mode TARGET_FLOATN_MODE (int @var{n}, bool @var{extended})
4263 Define this to return the machine mode to use for the type
4264 @code{_Float@var{n}}, if @var{extended} is false, or the type
4265 @code{_Float@var{n}x}, if @var{extended} is true. If such a type is not
4266 supported, return @code{opt_scalar_float_mode ()}. The default version of
4267 this hook returns @code{SFmode} for @code{_Float32}, @code{DFmode} for
4268 @code{_Float64} and @code{_Float32x} and @code{TFmode} for
4269 @code{_Float128}, if those modes exist and satisfy the requirements for
4270 those types and pass @code{TARGET_SCALAR_MODE_SUPPORTED_P} and
4271 @code{TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P}; for @code{_Float64x}, it
4272 returns the first of @code{XFmode} and @code{TFmode} that exists and
4273 satisfies the same requirements; for other types, it returns
4274 @code{opt_scalar_float_mode ()}. The hook is only called for values
4275 of @var{n} and @var{extended} that are valid according to
4276 ISO/IEC TS 18661-3:2015; that is, @var{n} is one of 32, 64, 128, or,
4277 if @var{extended} is false, 16 or greater than 128 and a multiple of 32.
4280 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (machine_mode @var{mode})
4281 Define this to return nonzero for machine modes for which the port has
4282 small register classes. If this target hook returns nonzero for a given
4283 @var{mode}, the compiler will try to minimize the lifetime of registers
4284 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4285 In this case, the hook is expected to return nonzero if it returns nonzero
4288 On some machines, it is risky to let hard registers live across arbitrary
4289 insns. Typically, these machines have instructions that require values
4290 to be in specific registers (like an accumulator), and reload will fail
4291 if the required hard register is used for another purpose across such an
4294 Passes before reload do not know which hard registers will be used
4295 in an instruction, but the machine modes of the registers set or used in
4296 the instruction are already known. And for some machines, register
4297 classes are small for, say, integer registers but not for floating point
4298 registers. For example, the AMD x86-64 architecture requires specific
4299 registers for the legacy x86 integer instructions, but there are many
4300 SSE registers for floating point operations. On such targets, a good
4301 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4302 machine modes but zero for the SSE register classes.
4304 The default version of this hook returns false for any mode. It is always
4305 safe to redefine this hook to return with a nonzero value. But if you
4306 unnecessarily define it, you will reduce the amount of optimizations
4307 that can be performed in some cases. If you do not define this hook
4308 to return a nonzero value when it is required, the compiler will run out
4309 of spill registers and print a fatal error message.
4313 @subsection How Scalar Function Values Are Returned
4314 @cindex return values in registers
4315 @cindex values, returned by functions
4316 @cindex scalars, returned as values
4318 This section discusses the macros that control returning scalars as
4319 values---values that can fit in registers.
4321 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4323 Define this to return an RTX representing the place where a function
4324 returns or receives a value of data type @var{ret_type}, a tree node
4325 representing a data type. @var{fn_decl_or_type} is a tree node
4326 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4327 function being called. If @var{outgoing} is false, the hook should
4328 compute the register in which the caller will see the return value.
4329 Otherwise, the hook should return an RTX representing the place where
4330 a function returns a value.
4332 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4333 (Actually, on most machines, scalar values are returned in the same
4334 place regardless of mode.) The value of the expression is usually a
4335 @code{reg} RTX for the hard register where the return value is stored.
4336 The value can also be a @code{parallel} RTX, if the return value is in
4337 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4338 @code{parallel} form. Note that the callee will populate every
4339 location specified in the @code{parallel}, but if the first element of
4340 the @code{parallel} contains the whole return value, callers will use
4341 that element as the canonical location and ignore the others. The m68k
4342 port uses this type of @code{parallel} to return pointers in both
4343 @samp{%a0} (the canonical location) and @samp{%d0}.
4345 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4346 the same promotion rules specified in @code{PROMOTE_MODE} if
4347 @var{valtype} is a scalar type.
4349 If the precise function being called is known, @var{func} is a tree
4350 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4351 pointer. This makes it possible to use a different value-returning
4352 convention for specific functions when all their calls are
4355 Some target machines have ``register windows'' so that the register in
4356 which a function returns its value is not the same as the one in which
4357 the caller sees the value. For such machines, you should return
4358 different RTX depending on @var{outgoing}.
4360 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4361 aggregate data types, because these are returned in another way. See
4362 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4365 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4366 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4367 a new target instead.
4370 @defmac LIBCALL_VALUE (@var{mode})
4371 A C expression to create an RTX representing the place where a library
4372 function returns a value of mode @var{mode}.
4374 Note that ``library function'' in this context means a compiler
4375 support routine, used to perform arithmetic, whose name is known
4376 specially by the compiler and was not mentioned in the C code being
4380 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (machine_mode @var{mode}, const_rtx @var{fun})
4381 Define this hook if the back-end needs to know the name of the libcall
4382 function in order to determine where the result should be returned.
4384 The mode of the result is given by @var{mode} and the name of the called
4385 library function is given by @var{fun}. The hook should return an RTX
4386 representing the place where the library function result will be returned.
4388 If this hook is not defined, then LIBCALL_VALUE will be used.
4391 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4392 A C expression that is nonzero if @var{regno} is the number of a hard
4393 register in which the values of called function may come back.
4395 A register whose use for returning values is limited to serving as the
4396 second of a pair (for a value of type @code{double}, say) need not be
4397 recognized by this macro. So for most machines, this definition
4401 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4404 If the machine has register windows, so that the caller and the called
4405 function use different registers for the return value, this macro
4406 should recognize only the caller's register numbers.
4408 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4409 for a new target instead.
4412 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4413 A target hook that return @code{true} if @var{regno} is the number of a hard
4414 register in which the values of called function may come back.
4416 A register whose use for returning values is limited to serving as the
4417 second of a pair (for a value of type @code{double}, say) need not be
4418 recognized by this target hook.
4420 If the machine has register windows, so that the caller and the called
4421 function use different registers for the return value, this target hook
4422 should recognize only the caller's register numbers.
4424 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4427 @defmac APPLY_RESULT_SIZE
4428 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4429 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4430 saving and restoring an arbitrary return value.
4433 @deftypevr {Target Hook} bool TARGET_OMIT_STRUCT_RETURN_REG
4434 Normally, when a function returns a structure by memory, the address
4435 is passed as an invisible pointer argument, but the compiler also
4436 arranges to return the address from the function like it would a normal
4437 pointer return value. Define this to true if that behavior is
4438 undesirable on your target.
4441 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4442 This hook should return true if values of type @var{type} are returned
4443 at the most significant end of a register (in other words, if they are
4444 padded at the least significant end). You can assume that @var{type}
4445 is returned in a register; the caller is required to check this.
4447 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4448 be able to hold the complete return value. For example, if a 1-, 2-
4449 or 3-byte structure is returned at the most significant end of a
4450 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4454 @node Aggregate Return
4455 @subsection How Large Values Are Returned
4456 @cindex aggregates as return values
4457 @cindex large return values
4458 @cindex returning aggregate values
4459 @cindex structure value address
4461 When a function value's mode is @code{BLKmode} (and in some other
4462 cases), the value is not returned according to
4463 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4464 caller passes the address of a block of memory in which the value
4465 should be stored. This address is called the @dfn{structure value
4468 This section describes how to control returning structure values in
4471 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4472 This target hook should return a nonzero value to say to return the
4473 function value in memory, just as large structures are always returned.
4474 Here @var{type} will be the data type of the value, and @var{fntype}
4475 will be the type of the function doing the returning, or @code{NULL} for
4478 Note that values of mode @code{BLKmode} must be explicitly handled
4479 by this function. Also, the option @option{-fpcc-struct-return}
4480 takes effect regardless of this macro. On most systems, it is
4481 possible to leave the hook undefined; this causes a default
4482 definition to be used, whose value is the constant 1 for @code{BLKmode}
4483 values, and 0 otherwise.
4485 Do not use this hook to indicate that structures and unions should always
4486 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4490 @defmac DEFAULT_PCC_STRUCT_RETURN
4491 Define this macro to be 1 if all structure and union return values must be
4492 in memory. Since this results in slower code, this should be defined
4493 only if needed for compatibility with other compilers or with an ABI@.
4494 If you define this macro to be 0, then the conventions used for structure
4495 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4498 If not defined, this defaults to the value 1.
4501 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4502 This target hook should return the location of the structure value
4503 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4504 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4505 be @code{NULL}, for libcalls. You do not need to define this target
4506 hook if the address is always passed as an ``invisible'' first
4509 On some architectures the place where the structure value address
4510 is found by the called function is not the same place that the
4511 caller put it. This can be due to register windows, or it could
4512 be because the function prologue moves it to a different place.
4513 @var{incoming} is @code{1} or @code{2} when the location is needed in
4514 the context of the called function, and @code{0} in the context of
4517 If @var{incoming} is nonzero and the address is to be found on the
4518 stack, return a @code{mem} which refers to the frame pointer. If
4519 @var{incoming} is @code{2}, the result is being used to fetch the
4520 structure value address at the beginning of a function. If you need
4521 to emit adjusting code, you should do it at this point.
4524 @defmac PCC_STATIC_STRUCT_RETURN
4525 Define this macro if the usual system convention on the target machine
4526 for returning structures and unions is for the called function to return
4527 the address of a static variable containing the value.
4529 Do not define this if the usual system convention is for the caller to
4530 pass an address to the subroutine.
4532 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4533 nothing when you use @option{-freg-struct-return} mode.
4536 @deftypefn {Target Hook} machine_mode TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4537 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.
4540 @deftypefn {Target Hook} machine_mode TARGET_GET_RAW_ARG_MODE (int @var{regno})
4541 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.
4545 @subsection Caller-Saves Register Allocation
4547 If you enable it, GCC can save registers around function calls. This
4548 makes it possible to use call-clobbered registers to hold variables that
4549 must live across calls.
4551 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4552 A C expression specifying which mode is required for saving @var{nregs}
4553 of a pseudo-register in call-clobbered hard register @var{regno}. If
4554 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4555 returned. For most machines this macro need not be defined since GCC
4556 will select the smallest suitable mode.
4559 @node Function Entry
4560 @subsection Function Entry and Exit
4561 @cindex function entry and exit
4565 This section describes the macros that output function entry
4566 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4568 @deftypefn {Target Hook} void TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY (FILE *@var{file}, unsigned HOST_WIDE_INT @var{patch_area_size}, bool @var{record_p})
4569 Generate a patchable area at the function start, consisting of
4570 @var{patch_area_size} NOP instructions. If the target supports named
4571 sections and if @var{record_p} is true, insert a pointer to the current
4572 location in the table of patchable functions. The default implementation
4573 of the hook places the table of pointers in the special section named
4574 @code{__patchable_function_entries}.
4577 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file})
4578 If defined, a function that outputs the assembler code for entry to a
4579 function. The prologue is responsible for setting up the stack frame,
4580 initializing the frame pointer register, saving registers that must be
4581 saved, and allocating @var{size} additional bytes of storage for the
4582 local variables. @var{file} is a stdio stream to which the assembler
4583 code should be output.
4585 The label for the beginning of the function need not be output by this
4586 macro. That has already been done when the macro is run.
4588 @findex regs_ever_live
4589 To determine which registers to save, the macro can refer to the array
4590 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4591 @var{r} is used anywhere within the function. This implies the function
4592 prologue should save register @var{r}, provided it is not one of the
4593 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4594 @code{regs_ever_live}.)
4596 On machines that have ``register windows'', the function entry code does
4597 not save on the stack the registers that are in the windows, even if
4598 they are supposed to be preserved by function calls; instead it takes
4599 appropriate steps to ``push'' the register stack, if any non-call-used
4600 registers are used in the function.
4602 @findex frame_pointer_needed
4603 On machines where functions may or may not have frame-pointers, the
4604 function entry code must vary accordingly; it must set up the frame
4605 pointer if one is wanted, and not otherwise. To determine whether a
4606 frame pointer is in wanted, the macro can refer to the variable
4607 @code{frame_pointer_needed}. The variable's value will be 1 at run
4608 time in a function that needs a frame pointer. @xref{Elimination}.
4610 The function entry code is responsible for allocating any stack space
4611 required for the function. This stack space consists of the regions
4612 listed below. In most cases, these regions are allocated in the
4613 order listed, with the last listed region closest to the top of the
4614 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4615 the highest address if it is not defined). You can use a different order
4616 for a machine if doing so is more convenient or required for
4617 compatibility reasons. Except in cases where required by standard
4618 or by a debugger, there is no reason why the stack layout used by GCC
4619 need agree with that used by other compilers for a machine.
4622 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4623 If defined, a function that outputs assembler code at the end of a
4624 prologue. This should be used when the function prologue is being
4625 emitted as RTL, and you have some extra assembler that needs to be
4626 emitted. @xref{prologue instruction pattern}.
4629 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4630 If defined, a function that outputs assembler code at the start of an
4631 epilogue. This should be used when the function epilogue is being
4632 emitted as RTL, and you have some extra assembler that needs to be
4633 emitted. @xref{epilogue instruction pattern}.
4636 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file})
4637 If defined, a function that outputs the assembler code for exit from a
4638 function. The epilogue is responsible for restoring the saved
4639 registers and stack pointer to their values when the function was
4640 called, and returning control to the caller. This macro takes the
4641 same argument as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4642 registers to restore are determined from @code{regs_ever_live} and
4643 @code{CALL_USED_REGISTERS} in the same way.
4645 On some machines, there is a single instruction that does all the work
4646 of returning from the function. On these machines, give that
4647 instruction the name @samp{return} and do not define the macro
4648 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4650 Do not define a pattern named @samp{return} if you want the
4651 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4652 switches to control whether return instructions or epilogues are used,
4653 define a @samp{return} pattern with a validity condition that tests the
4654 target switches appropriately. If the @samp{return} pattern's validity
4655 condition is false, epilogues will be used.
4657 On machines where functions may or may not have frame-pointers, the
4658 function exit code must vary accordingly. Sometimes the code for these
4659 two cases is completely different. To determine whether a frame pointer
4660 is wanted, the macro can refer to the variable
4661 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4662 a function that needs a frame pointer.
4664 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4665 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4666 The C variable @code{current_function_is_leaf} is nonzero for such a
4667 function. @xref{Leaf Functions}.
4669 On some machines, some functions pop their arguments on exit while
4670 others leave that for the caller to do. For example, the 68020 when
4671 given @option{-mrtd} pops arguments in functions that take a fixed
4672 number of arguments.
4675 @findex crtl->args.pops_args
4676 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4677 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4678 needs to know what was decided. The number of bytes of the current
4679 function's arguments that this function should pop is available in
4680 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4685 @findex pretend_args_size
4686 @findex crtl->args.pretend_args_size
4687 A region of @code{crtl->args.pretend_args_size} bytes of
4688 uninitialized space just underneath the first argument arriving on the
4689 stack. (This may not be at the very start of the allocated stack region
4690 if the calling sequence has pushed anything else since pushing the stack
4691 arguments. But usually, on such machines, nothing else has been pushed
4692 yet, because the function prologue itself does all the pushing.) This
4693 region is used on machines where an argument may be passed partly in
4694 registers and partly in memory, and, in some cases to support the
4695 features in @code{<stdarg.h>}.
4698 An area of memory used to save certain registers used by the function.
4699 The size of this area, which may also include space for such things as
4700 the return address and pointers to previous stack frames, is
4701 machine-specific and usually depends on which registers have been used
4702 in the function. Machines with register windows often do not require
4706 A region of at least @var{size} bytes, possibly rounded up to an allocation
4707 boundary, to contain the local variables of the function. On some machines,
4708 this region and the save area may occur in the opposite order, with the
4709 save area closer to the top of the stack.
4712 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4713 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4714 @code{crtl->outgoing_args_size} bytes to be used for outgoing
4715 argument lists of the function. @xref{Stack Arguments}.
4718 @defmac EXIT_IGNORE_STACK
4719 Define this macro as a C expression that is nonzero if the return
4720 instruction or the function epilogue ignores the value of the stack
4721 pointer; in other words, if it is safe to delete an instruction to
4722 adjust the stack pointer before a return from the function. The
4725 Note that this macro's value is relevant only for functions for which
4726 frame pointers are maintained. It is never safe to delete a final
4727 stack adjustment in a function that has no frame pointer, and the
4728 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4731 @defmac EPILOGUE_USES (@var{regno})
4732 Define this macro as a C expression that is nonzero for registers that are
4733 used by the epilogue or the @samp{return} pattern. The stack and frame
4734 pointer registers are already assumed to be used as needed.
4737 @defmac EH_USES (@var{regno})
4738 Define this macro as a C expression that is nonzero for registers that are
4739 used by the exception handling mechanism, and so should be considered live
4740 on entry to an exception edge.
4743 @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})
4744 A function that outputs the assembler code for a thunk
4745 function, used to implement C++ virtual function calls with multiple
4746 inheritance. The thunk acts as a wrapper around a virtual function,
4747 adjusting the implicit object parameter before handing control off to
4750 First, emit code to add the integer @var{delta} to the location that
4751 contains the incoming first argument. Assume that this argument
4752 contains a pointer, and is the one used to pass the @code{this} pointer
4753 in C++. This is the incoming argument @emph{before} the function prologue,
4754 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4755 all other incoming arguments.
4757 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4758 made after adding @code{delta}. In particular, if @var{p} is the
4759 adjusted pointer, the following adjustment should be made:
4762 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4765 After the additions, emit code to jump to @var{function}, which is a
4766 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4767 not touch the return address. Hence returning from @var{FUNCTION} will
4768 return to whoever called the current @samp{thunk}.
4770 The effect must be as if @var{function} had been called directly with
4771 the adjusted first argument. This macro is responsible for emitting all
4772 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4773 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4775 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4776 have already been extracted from it.) It might possibly be useful on
4777 some targets, but probably not.
4779 If you do not define this macro, the target-independent code in the C++
4780 front end will generate a less efficient heavyweight thunk that calls
4781 @var{function} instead of jumping to it. The generic approach does
4782 not support varargs.
4785 @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})
4786 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4787 to output the assembler code for the thunk function specified by the
4788 arguments it is passed, and false otherwise. In the latter case, the
4789 generic approach will be used by the C++ front end, with the limitations
4794 @subsection Generating Code for Profiling
4795 @cindex profiling, code generation
4797 These macros will help you generate code for profiling.
4799 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4800 A C statement or compound statement to output to @var{file} some
4801 assembler code to call the profiling subroutine @code{mcount}.
4804 The details of how @code{mcount} expects to be called are determined by
4805 your operating system environment, not by GCC@. To figure them out,
4806 compile a small program for profiling using the system's installed C
4807 compiler and look at the assembler code that results.
4809 Older implementations of @code{mcount} expect the address of a counter
4810 variable to be loaded into some register. The name of this variable is
4811 @samp{LP} followed by the number @var{labelno}, so you would generate
4812 the name using @samp{LP%d} in a @code{fprintf}.
4815 @defmac PROFILE_HOOK
4816 A C statement or compound statement to output to @var{file} some assembly
4817 code to call the profiling subroutine @code{mcount} even the target does
4818 not support profiling.
4821 @defmac NO_PROFILE_COUNTERS
4822 Define this macro to be an expression with a nonzero value if the
4823 @code{mcount} subroutine on your system does not need a counter variable
4824 allocated for each function. This is true for almost all modern
4825 implementations. If you define this macro, you must not use the
4826 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4829 @defmac PROFILE_BEFORE_PROLOGUE
4830 Define this macro if the code for function profiling should come before
4831 the function prologue. Normally, the profiling code comes after.
4834 @deftypefn {Target Hook} bool TARGET_KEEP_LEAF_WHEN_PROFILED (void)
4835 This target hook returns true if the target wants the leaf flag for the current function to stay true even if it calls mcount. This might make sense for targets using the leaf flag only to determine whether a stack frame needs to be generated or not and for which the call to mcount is generated before the function prologue.
4839 @subsection Permitting tail calls
4842 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4843 True if it is OK to do sibling call optimization for the specified
4844 call expression @var{exp}. @var{decl} will be the called function,
4845 or @code{NULL} if this is an indirect call.
4847 It is not uncommon for limitations of calling conventions to prevent
4848 tail calls to functions outside the current unit of translation, or
4849 during PIC compilation. The hook is used to enforce these restrictions,
4850 as the @code{sibcall} md pattern can not fail, or fall over to a
4851 ``normal'' call. The criteria for successful sibling call optimization
4852 may vary greatly between different architectures.
4855 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4856 Add any hard registers to @var{regs} that are live on entry to the
4857 function. This hook only needs to be defined to provide registers that
4858 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4859 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4860 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4861 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4864 @deftypefn {Target Hook} void TARGET_SET_UP_BY_PROLOGUE (struct hard_reg_set_container *@var{})
4865 This hook should add additional registers that are computed by the prologue to the hard regset for shrink-wrapping optimization purposes.
4868 @deftypefn {Target Hook} bool TARGET_WARN_FUNC_RETURN (tree)
4869 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.
4872 @node Shrink-wrapping separate components
4873 @subsection Shrink-wrapping separate components
4874 @cindex shrink-wrapping separate components
4876 The prologue may perform a variety of target dependent tasks such as
4877 saving callee-saved registers, saving the return address, aligning the
4878 stack, creating a stack frame, initializing the PIC register, setting
4879 up the static chain, etc.
4881 On some targets some of these tasks may be independent of others and
4882 thus may be shrink-wrapped separately. These independent tasks are
4883 referred to as components and are handled generically by the target
4884 independent parts of GCC.
4886 Using the following hooks those prologue or epilogue components can be
4887 shrink-wrapped separately, so that the initialization (and possibly
4888 teardown) those components do is not done as frequently on execution
4889 paths where this would unnecessary.
4891 What exactly those components are is up to the target code; the generic
4892 code treats them abstractly, as a bit in an @code{sbitmap}. These
4893 @code{sbitmap}s are allocated by the @code{shrink_wrap.get_separate_components}
4894 and @code{shrink_wrap.components_for_bb} hooks, and deallocated by the
4897 @deftypefn {Target Hook} sbitmap TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS (void)
4898 This hook should return an @code{sbitmap} with the bits set for those
4899 components that can be separately shrink-wrapped in the current function.
4900 Return @code{NULL} if the current function should not get any separate
4902 Don't define this hook if it would always return @code{NULL}.
4903 If it is defined, the other hooks in this group have to be defined as well.
4906 @deftypefn {Target Hook} sbitmap TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB (basic_block)
4907 This hook should return an @code{sbitmap} with the bits set for those
4908 components where either the prologue component has to be executed before
4909 the @code{basic_block}, or the epilogue component after it, or both.
4912 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS (sbitmap @var{components}, edge @var{e}, sbitmap @var{edge_components}, bool @var{is_prologue})
4913 This hook should clear the bits in the @var{components} bitmap for those
4914 components in @var{edge_components} that the target cannot handle on edge
4915 @var{e}, where @var{is_prologue} says if this is for a prologue or an
4919 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS (sbitmap)
4920 Emit prologue insns for the components indicated by the parameter.
4923 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS (sbitmap)
4924 Emit epilogue insns for the components indicated by the parameter.
4927 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS (sbitmap)
4928 Mark the components in the parameter as handled, so that the
4929 @code{prologue} and @code{epilogue} named patterns know to ignore those
4930 components. The target code should not hang on to the @code{sbitmap}, it
4931 will be deleted after this call.
4934 @node Stack Smashing Protection
4935 @subsection Stack smashing protection
4936 @cindex stack smashing protection
4938 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4939 This hook returns a @code{DECL} node for the external variable to use
4940 for the stack protection guard. This variable is initialized by the
4941 runtime to some random value and is used to initialize the guard value
4942 that is placed at the top of the local stack frame. The type of this
4943 variable must be @code{ptr_type_node}.
4945 The default version of this hook creates a variable called
4946 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4949 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4950 This hook returns a @code{CALL_EXPR} that alerts the runtime that the
4951 stack protect guard variable has been modified. This expression should
4952 involve a call to a @code{noreturn} function.
4954 The default version of this hook invokes a function called
4955 @samp{__stack_chk_fail}, taking no arguments. This function is
4956 normally defined in @file{libgcc2.c}.
4959 @deftypefn {Target Hook} bool TARGET_STACK_PROTECT_RUNTIME_ENABLED_P (void)
4960 Returns true if the target wants GCC's default stack protect runtime support, otherwise return false. The default implementation always returns true.
4963 @deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
4964 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
4967 @node Miscellaneous Register Hooks
4968 @subsection Miscellaneous register hooks
4969 @cindex miscellaneous register hooks
4971 @deftypevr {Target Hook} bool TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
4972 Set to true if each call that binds to a local definition explicitly
4973 clobbers or sets all non-fixed registers modified by performing the call.
4974 That is, by the call pattern itself, or by code that might be inserted by the
4975 linker (e.g. stubs, veneers, branch islands), but not including those
4976 modifiable by the callee. The affected registers may be mentioned explicitly
4977 in the call pattern, or included as clobbers in CALL_INSN_FUNCTION_USAGE.
4978 The default version of this hook is set to false. The purpose of this hook
4979 is to enable the fipa-ra optimization.
4983 @section Implementing the Varargs Macros
4984 @cindex varargs implementation
4986 GCC comes with an implementation of @code{<varargs.h>} and
4987 @code{<stdarg.h>} that work without change on machines that pass arguments
4988 on the stack. Other machines require their own implementations of
4989 varargs, and the two machine independent header files must have
4990 conditionals to include it.
4992 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4993 the calling convention for @code{va_start}. The traditional
4994 implementation takes just one argument, which is the variable in which
4995 to store the argument pointer. The ISO implementation of
4996 @code{va_start} takes an additional second argument. The user is
4997 supposed to write the last named argument of the function here.
4999 However, @code{va_start} should not use this argument. The way to find
5000 the end of the named arguments is with the built-in functions described
5003 @defmac __builtin_saveregs ()
5004 Use this built-in function to save the argument registers in memory so
5005 that the varargs mechanism can access them. Both ISO and traditional
5006 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
5007 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
5009 On some machines, @code{__builtin_saveregs} is open-coded under the
5010 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
5011 other machines, it calls a routine written in assembler language,
5012 found in @file{libgcc2.c}.
5014 Code generated for the call to @code{__builtin_saveregs} appears at the
5015 beginning of the function, as opposed to where the call to
5016 @code{__builtin_saveregs} is written, regardless of what the code is.
5017 This is because the registers must be saved before the function starts
5018 to use them for its own purposes.
5019 @c i rewrote the first sentence above to fix an overfull hbox. --mew
5023 @defmac __builtin_next_arg (@var{lastarg})
5024 This builtin returns the address of the first anonymous stack
5025 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
5026 returns the address of the location above the first anonymous stack
5027 argument. Use it in @code{va_start} to initialize the pointer for
5028 fetching arguments from the stack. Also use it in @code{va_start} to
5029 verify that the second parameter @var{lastarg} is the last named argument
5030 of the current function.
5033 @defmac __builtin_classify_type (@var{object})
5034 Since each machine has its own conventions for which data types are
5035 passed in which kind of register, your implementation of @code{va_arg}
5036 has to embody these conventions. The easiest way to categorize the
5037 specified data type is to use @code{__builtin_classify_type} together
5038 with @code{sizeof} and @code{__alignof__}.
5040 @code{__builtin_classify_type} ignores the value of @var{object},
5041 considering only its data type. It returns an integer describing what
5042 kind of type that is---integer, floating, pointer, structure, and so on.
5044 The file @file{typeclass.h} defines an enumeration that you can use to
5045 interpret the values of @code{__builtin_classify_type}.
5048 These machine description macros help implement varargs:
5050 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5051 If defined, this hook produces the machine-specific code for a call to
5052 @code{__builtin_saveregs}. This code will be moved to the very
5053 beginning of the function, before any parameter access are made. The
5054 return value of this function should be an RTX that contains the value
5055 to use as the return of @code{__builtin_saveregs}.
5058 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (cumulative_args_t @var{args_so_far}, machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
5059 This target hook offers an alternative to using
5060 @code{__builtin_saveregs} and defining the hook
5061 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5062 register arguments into the stack so that all the arguments appear to
5063 have been passed consecutively on the stack. Once this is done, you can
5064 use the standard implementation of varargs that works for machines that
5065 pass all their arguments on the stack.
5067 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5068 structure, containing the values that are obtained after processing the
5069 named arguments. The arguments @var{mode} and @var{type} describe the
5070 last named argument---its machine mode and its data type as a tree node.
5072 The target hook should do two things: first, push onto the stack all the
5073 argument registers @emph{not} used for the named arguments, and second,
5074 store the size of the data thus pushed into the @code{int}-valued
5075 variable pointed to by @var{pretend_args_size}. The value that you
5076 store here will serve as additional offset for setting up the stack
5079 Because you must generate code to push the anonymous arguments at
5080 compile time without knowing their data types,
5081 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5082 have just a single category of argument register and use it uniformly
5085 If the argument @var{second_time} is nonzero, it means that the
5086 arguments of the function are being analyzed for the second time. This
5087 happens for an inline function, which is not actually compiled until the
5088 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5089 not generate any instructions in this case.
5092 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca})
5093 Define this hook to return @code{true} if the location where a function
5094 argument is passed depends on whether or not it is a named argument.
5096 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5097 is set for varargs and stdarg functions. If this hook returns
5098 @code{true}, the @var{named} argument is always true for named
5099 arguments, and false for unnamed arguments. If it returns @code{false},
5100 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5101 then all arguments are treated as named. Otherwise, all named arguments
5102 except the last are treated as named.
5104 You need not define this hook if it always returns @code{false}.
5107 @deftypefn {Target Hook} void TARGET_CALL_ARGS (rtx, @var{tree})
5108 While generating RTL for a function call, this target hook is invoked once
5109 for each argument passed to the function, either a register returned by
5110 @code{TARGET_FUNCTION_ARG} or a memory location. It is called just
5111 before the point where argument registers are stored. The type of the
5112 function to be called is also passed as the second argument; it is
5113 @code{NULL_TREE} for libcalls. The @code{TARGET_END_CALL_ARGS} hook is
5114 invoked just after the code to copy the return reg has been emitted.
5115 This functionality can be used to perform special setup of call argument
5116 registers if a target needs it.
5117 For functions without arguments, the hook is called once with @code{pc_rtx}
5118 passed instead of an argument register.
5119 Most ports do not need to implement anything for this hook.
5122 @deftypefn {Target Hook} void TARGET_END_CALL_ARGS (void)
5123 This target hook is invoked while generating RTL for a function call,
5124 just after the point where the return reg is copied into a pseudo. It
5125 signals that all the call argument and return registers for the just
5126 emitted call are now no longer in use.
5127 Most ports do not need to implement anything for this hook.
5130 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca})
5131 If you need to conditionally change ABIs so that one works with
5132 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5133 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5134 defined, then define this hook to return @code{true} if
5135 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5136 Otherwise, you should not define this hook.
5139 @deftypefn {Target Hook} rtx TARGET_LOAD_BOUNDS_FOR_ARG (rtx @var{slot}, rtx @var{arg}, rtx @var{slot_no})
5140 This hook is used by expand pass to emit insn to load bounds of
5141 @var{arg} passed in @var{slot}. Expand pass uses this hook in case
5142 bounds of @var{arg} are not passed in register. If @var{slot} is a
5143 memory, then bounds are loaded as for regular pointer loaded from
5144 memory. If @var{slot} is not a memory then @var{slot_no} is an integer
5145 constant holding number of the target dependent special slot which
5146 should be used to obtain bounds. Hook returns RTX holding loaded bounds.
5149 @deftypefn {Target Hook} void TARGET_STORE_BOUNDS_FOR_ARG (rtx @var{arg}, rtx @var{slot}, rtx @var{bounds}, rtx @var{slot_no})
5150 This hook is used by expand pass to emit insns to store @var{bounds} of
5151 @var{arg} passed in @var{slot}. Expand pass uses this hook in case
5152 @var{bounds} of @var{arg} are not passed in register. If @var{slot} is a
5153 memory, then @var{bounds} are stored as for regular pointer stored in
5154 memory. If @var{slot} is not a memory then @var{slot_no} is an integer
5155 constant holding number of the target dependent special slot which
5156 should be used to store @var{bounds}.
5159 @deftypefn {Target Hook} rtx TARGET_LOAD_RETURNED_BOUNDS (rtx @var{slot})
5160 This hook is used by expand pass to emit insn to load bounds
5161 returned by function call in @var{slot}. Hook returns RTX holding
5165 @deftypefn {Target Hook} void TARGET_STORE_RETURNED_BOUNDS (rtx @var{slot}, rtx @var{bounds})
5166 This hook is used by expand pass to emit insn to store @var{bounds}
5167 returned by function call into @var{slot}.
5170 @deftypefn {Target Hook} rtx TARGET_CHKP_FUNCTION_VALUE_BOUNDS (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
5171 Define this to return an RTX representing the place where a function
5172 returns bounds for returned pointers. Arguments meaning is similar to
5173 @code{TARGET_FUNCTION_VALUE}.
5176 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARG_BOUNDS (cumulative_args_t @var{args_so_far}, machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
5177 Use it to store bounds for anonymous register arguments stored
5178 into the stack. Arguments meaning is similar to
5179 @code{TARGET_SETUP_INCOMING_VARARGS}.
5183 @section Trampolines for Nested Functions
5184 @cindex trampolines for nested functions
5185 @cindex nested functions, trampolines for
5187 A @dfn{trampoline} is a small piece of code that is created at run time
5188 when the address of a nested function is taken. It normally resides on
5189 the stack, in the stack frame of the containing function. These macros
5190 tell GCC how to generate code to allocate and initialize a
5193 The instructions in the trampoline must do two things: load a constant
5194 address into the static chain register, and jump to the real address of
5195 the nested function. On CISC machines such as the m68k, this requires
5196 two instructions, a move immediate and a jump. Then the two addresses
5197 exist in the trampoline as word-long immediate operands. On RISC
5198 machines, it is often necessary to load each address into a register in
5199 two parts. Then pieces of each address form separate immediate
5202 The code generated to initialize the trampoline must store the variable
5203 parts---the static chain value and the function address---into the
5204 immediate operands of the instructions. On a CISC machine, this is
5205 simply a matter of copying each address to a memory reference at the
5206 proper offset from the start of the trampoline. On a RISC machine, it
5207 may be necessary to take out pieces of the address and store them
5210 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5211 This hook is called by @code{assemble_trampoline_template} to output,
5212 on the stream @var{f}, assembler code for a block of data that contains
5213 the constant parts of a trampoline. This code should not include a
5214 label---the label is taken care of automatically.
5216 If you do not define this hook, it means no template is needed
5217 for the target. Do not define this hook on systems where the block move
5218 code to copy the trampoline into place would be larger than the code
5219 to generate it on the spot.
5222 @defmac TRAMPOLINE_SECTION
5223 Return the section into which the trampoline template is to be placed
5224 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5227 @defmac TRAMPOLINE_SIZE
5228 A C expression for the size in bytes of the trampoline, as an integer.
5231 @defmac TRAMPOLINE_ALIGNMENT
5232 Alignment required for trampolines, in bits.
5234 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5235 is used for aligning trampolines.
5238 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5239 This hook is called to initialize a trampoline.
5240 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5241 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5242 RTX for the static chain value that should be passed to the function
5245 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5246 first thing this hook should do is emit a block move into @var{m_tramp}
5247 from the memory block returned by @code{assemble_trampoline_template}.
5248 Note that the block move need only cover the constant parts of the
5249 trampoline. If the target isolates the variable parts of the trampoline
5250 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5252 If the target requires any other actions, such as flushing caches or
5253 enabling stack execution, these actions should be performed after
5254 initializing the trampoline proper.
5257 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5258 This hook should perform any machine-specific adjustment in
5259 the address of the trampoline. Its argument contains the address of the
5260 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5261 the address to be used for a function call should be different from the
5262 address at which the template was stored, the different address should
5263 be returned; otherwise @var{addr} should be returned unchanged.
5264 If this hook is not defined, @var{addr} will be used for function calls.
5267 @deftypevr {Target Hook} int TARGET_CUSTOM_FUNCTION_DESCRIPTORS
5268 This hook should be defined to a power of 2 if the target will benefit
5269 from the use of custom descriptors for nested functions instead of the
5270 standard trampolines. Such descriptors are created at run time on the
5271 stack and made up of data only, but they are non-standard so the generated
5272 code must be prepared to deal with them. This hook should be defined to 0
5273 if the target uses function descriptors for its standard calling sequence,
5274 like for example HP-PA or IA-64. Using descriptors for nested functions
5275 eliminates the need for trampolines that reside on the stack and require
5276 it to be made executable.
5278 The value of the macro is used to parameterize the run-time identification
5279 scheme implemented to distinguish descriptors from function addresses: it
5280 gives the number of bytes by which their address is misaligned compared
5281 with function addresses. The value of 1 will generally work, unless it is
5282 already reserved by the target for another purpose, like for example on ARM.
5285 Implementing trampolines is difficult on many machines because they have
5286 separate instruction and data caches. Writing into a stack location
5287 fails to clear the memory in the instruction cache, so when the program
5288 jumps to that location, it executes the old contents.
5290 Here are two possible solutions. One is to clear the relevant parts of
5291 the instruction cache whenever a trampoline is set up. The other is to
5292 make all trampolines identical, by having them jump to a standard
5293 subroutine. The former technique makes trampoline execution faster; the
5294 latter makes initialization faster.
5296 To clear the instruction cache when a trampoline is initialized, define
5297 the following macro.
5299 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5300 If defined, expands to a C expression clearing the @emph{instruction
5301 cache} in the specified interval. The definition of this macro would
5302 typically be a series of @code{asm} statements. Both @var{beg} and
5303 @var{end} are both pointer expressions.
5306 To use a standard subroutine, define the following macro. In addition,
5307 you must make sure that the instructions in a trampoline fill an entire
5308 cache line with identical instructions, or else ensure that the
5309 beginning of the trampoline code is always aligned at the same point in
5310 its cache line. Look in @file{m68k.h} as a guide.
5312 @defmac TRANSFER_FROM_TRAMPOLINE
5313 Define this macro if trampolines need a special subroutine to do their
5314 work. The macro should expand to a series of @code{asm} statements
5315 which will be compiled with GCC@. They go in a library function named
5316 @code{__transfer_from_trampoline}.
5318 If you need to avoid executing the ordinary prologue code of a compiled
5319 C function when you jump to the subroutine, you can do so by placing a
5320 special label of your own in the assembler code. Use one @code{asm}
5321 statement to generate an assembler label, and another to make the label
5322 global. Then trampolines can use that label to jump directly to your
5323 special assembler code.
5327 @section Implicit Calls to Library Routines
5328 @cindex library subroutine names
5329 @cindex @file{libgcc.a}
5331 @c prevent bad page break with this line
5332 Here is an explanation of implicit calls to library routines.
5334 @defmac DECLARE_LIBRARY_RENAMES
5335 This macro, if defined, should expand to a piece of C code that will get
5336 expanded when compiling functions for libgcc.a. It can be used to
5337 provide alternate names for GCC's internal library functions if there
5338 are ABI-mandated names that the compiler should provide.
5341 @findex set_optab_libfunc
5342 @findex init_one_libfunc
5343 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5344 This hook should declare additional library routines or rename
5345 existing ones, using the functions @code{set_optab_libfunc} and
5346 @code{init_one_libfunc} defined in @file{optabs.c}.
5347 @code{init_optabs} calls this macro after initializing all the normal
5350 The default is to do nothing. Most ports don't need to define this hook.
5353 @deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX
5354 If false (the default), internal library routines start with two
5355 underscores. If set to true, these routines start with @code{__gnu_}
5356 instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This
5357 currently only affects functions defined in @file{libgcc2.c}. If this
5358 is set to true, the @file{tm.h} file must also
5359 @code{#define LIBGCC2_GNU_PREFIX}.
5362 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5363 This macro should return @code{true} if the library routine that
5364 implements the floating point comparison operator @var{comparison} in
5365 mode @var{mode} will return a boolean, and @var{false} if it will
5368 GCC's own floating point libraries return tristates from the
5369 comparison operators, so the default returns false always. Most ports
5370 don't need to define this macro.
5373 @defmac TARGET_LIB_INT_CMP_BIASED
5374 This macro should evaluate to @code{true} if the integer comparison
5375 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5376 operand is smaller than the second, 1 to indicate that they are equal,
5377 and 2 to indicate that the first operand is greater than the second.
5378 If this macro evaluates to @code{false} the comparison functions return
5379 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5380 in @file{libgcc.a}, you do not need to define this macro.
5383 @defmac TARGET_HAS_NO_HW_DIVIDE
5384 This macro should be defined if the target has no hardware divide
5385 instructions. If this macro is defined, GCC will use an algorithm which
5386 make use of simple logical and arithmetic operations for 64-bit
5387 division. If the macro is not defined, GCC will use an algorithm which
5388 make use of a 64-bit by 32-bit divide primitive.
5391 @cindex @code{EDOM}, implicit usage
5394 The value of @code{EDOM} on the target machine, as a C integer constant
5395 expression. If you don't define this macro, GCC does not attempt to
5396 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5397 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5400 If you do not define @code{TARGET_EDOM}, then compiled code reports
5401 domain errors by calling the library function and letting it report the
5402 error. If mathematical functions on your system use @code{matherr} when
5403 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5404 that @code{matherr} is used normally.
5407 @cindex @code{errno}, implicit usage
5408 @defmac GEN_ERRNO_RTX
5409 Define this macro as a C expression to create an rtl expression that
5410 refers to the global ``variable'' @code{errno}. (On certain systems,
5411 @code{errno} may not actually be a variable.) If you don't define this
5412 macro, a reasonable default is used.
5415 @deftypefn {Target Hook} bool TARGET_LIBC_HAS_FUNCTION (enum function_class @var{fn_class})
5416 This hook determines whether a function from a class of functions
5417 @var{fn_class} is present at the runtime.
5420 @defmac NEXT_OBJC_RUNTIME
5421 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
5422 by default. This calling convention involves passing the object, the selector
5423 and the method arguments all at once to the method-lookup library function.
5424 This is the usual setting when targeting Darwin/Mac OS X systems, which have
5425 the NeXT runtime installed.
5427 If the macro is set to 0, the "GNU" Objective-C message sending convention
5428 will be used by default. This convention passes just the object and the
5429 selector to the method-lookup function, which returns a pointer to the method.
5431 In either case, it remains possible to select code-generation for the alternate
5432 scheme, by means of compiler command line switches.
5435 @node Addressing Modes
5436 @section Addressing Modes
5437 @cindex addressing modes
5439 @c prevent bad page break with this line
5440 This is about addressing modes.
5442 @defmac HAVE_PRE_INCREMENT
5443 @defmacx HAVE_PRE_DECREMENT
5444 @defmacx HAVE_POST_INCREMENT
5445 @defmacx HAVE_POST_DECREMENT
5446 A C expression that is nonzero if the machine supports pre-increment,
5447 pre-decrement, post-increment, or post-decrement addressing respectively.
5450 @defmac HAVE_PRE_MODIFY_DISP
5451 @defmacx HAVE_POST_MODIFY_DISP
5452 A C expression that is nonzero if the machine supports pre- or
5453 post-address side-effect generation involving constants other than
5454 the size of the memory operand.
5457 @defmac HAVE_PRE_MODIFY_REG
5458 @defmacx HAVE_POST_MODIFY_REG
5459 A C expression that is nonzero if the machine supports pre- or
5460 post-address side-effect generation involving a register displacement.
5463 @defmac CONSTANT_ADDRESS_P (@var{x})
5464 A C expression that is 1 if the RTX @var{x} is a constant which
5465 is a valid address. On most machines the default definition of
5466 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5467 is acceptable, but a few machines are more restrictive as to which
5468 constant addresses are supported.
5471 @defmac CONSTANT_P (@var{x})
5472 @code{CONSTANT_P}, which is defined by target-independent code,
5473 accepts integer-values expressions whose values are not explicitly
5474 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5475 expressions and @code{const} arithmetic expressions, in addition to
5476 @code{const_int} and @code{const_double} expressions.
5479 @defmac MAX_REGS_PER_ADDRESS
5480 A number, the maximum number of registers that can appear in a valid
5481 memory address. Note that it is up to you to specify a value equal to
5482 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5486 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5487 A function that returns whether @var{x} (an RTX) is a legitimate memory
5488 address on the target machine for a memory operand of mode @var{mode}.
5490 Legitimate addresses are defined in two variants: a strict variant and a
5491 non-strict one. The @var{strict} parameter chooses which variant is
5492 desired by the caller.
5494 The strict variant is used in the reload pass. It must be defined so
5495 that any pseudo-register that has not been allocated a hard register is
5496 considered a memory reference. This is because in contexts where some
5497 kind of register is required, a pseudo-register with no hard register
5498 must be rejected. For non-hard registers, the strict variant should look
5499 up the @code{reg_renumber} array; it should then proceed using the hard
5500 register number in the array, or treat the pseudo as a memory reference
5501 if the array holds @code{-1}.
5503 The non-strict variant is used in other passes. It must be defined to
5504 accept all pseudo-registers in every context where some kind of
5505 register is required.
5507 Normally, constant addresses which are the sum of a @code{symbol_ref}
5508 and an integer are stored inside a @code{const} RTX to mark them as
5509 constant. Therefore, there is no need to recognize such sums
5510 specifically as legitimate addresses. Normally you would simply
5511 recognize any @code{const} as legitimate.
5513 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5514 sums that are not marked with @code{const}. It assumes that a naked
5515 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5516 naked constant sums as illegitimate addresses, so that none of them will
5517 be given to @code{PRINT_OPERAND_ADDRESS}.
5519 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5520 On some machines, whether a symbolic address is legitimate depends on
5521 the section that the address refers to. On these machines, define the
5522 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5523 into the @code{symbol_ref}, and then check for it here. When you see a
5524 @code{const}, you will have to look inside it to find the
5525 @code{symbol_ref} in order to determine the section. @xref{Assembler
5528 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5529 Some ports are still using a deprecated legacy substitute for
5530 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5534 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5538 and should @code{goto @var{label}} if the address @var{x} is a valid
5539 address on the target machine for a memory operand of mode @var{mode}.
5541 @findex REG_OK_STRICT
5542 Compiler source files that want to use the strict variant of this
5543 macro define the macro @code{REG_OK_STRICT}. You should use an
5544 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5545 that case and the non-strict variant otherwise.
5547 Using the hook is usually simpler because it limits the number of
5548 files that are recompiled when changes are made.
5551 @defmac TARGET_MEM_CONSTRAINT
5552 A single character to be used instead of the default @code{'m'}
5553 character for general memory addresses. This defines the constraint
5554 letter which matches the memory addresses accepted by
5555 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5556 support new address formats in your back end without changing the
5557 semantics of the @code{'m'} constraint. This is necessary in order to
5558 preserve functionality of inline assembly constructs using the
5559 @code{'m'} constraint.
5562 @defmac FIND_BASE_TERM (@var{x})
5563 A C expression to determine the base term of address @var{x},
5564 or to provide a simplified version of @var{x} from which @file{alias.c}
5565 can easily find the base term. This macro is used in only two places:
5566 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5568 It is always safe for this macro to not be defined. It exists so
5569 that alias analysis can understand machine-dependent addresses.
5571 The typical use of this macro is to handle addresses containing
5572 a label_ref or symbol_ref within an UNSPEC@.
5575 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode})
5576 This hook is given an invalid memory address @var{x} for an
5577 operand of mode @var{mode} and should try to return a valid memory
5580 @findex break_out_memory_refs
5581 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5582 and @var{oldx} will be the operand that was given to that function to produce
5585 The code of the hook should not alter the substructure of
5586 @var{x}. If it transforms @var{x} into a more legitimate form, it
5587 should return the new @var{x}.
5589 It is not necessary for this hook to come up with a legitimate address,
5590 with the exception of native TLS addresses (@pxref{Emulated TLS}).
5591 The compiler has standard ways of doing so in all cases. In fact, if
5592 the target supports only emulated TLS, it
5593 is safe to omit this hook or make it return @var{x} if it cannot find
5594 a valid way to legitimize the address. But often a machine-dependent
5595 strategy can generate better code.
5598 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5599 A C compound statement that attempts to replace @var{x}, which is an address
5600 that needs reloading, with a valid memory address for an operand of mode
5601 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5602 It is not necessary to define this macro, but it might be useful for
5603 performance reasons.
5605 For example, on the i386, it is sometimes possible to use a single
5606 reload register instead of two by reloading a sum of two pseudo
5607 registers into a register. On the other hand, for number of RISC
5608 processors offsets are limited so that often an intermediate address
5609 needs to be generated in order to address a stack slot. By defining
5610 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5611 generated for adjacent some stack slots can be made identical, and thus
5614 @emph{Note}: This macro should be used with caution. It is necessary
5615 to know something of how reload works in order to effectively use this,
5616 and it is quite easy to produce macros that build in too much knowledge
5617 of reload internals.
5619 @emph{Note}: This macro must be able to reload an address created by a
5620 previous invocation of this macro. If it fails to handle such addresses
5621 then the compiler may generate incorrect code or abort.
5624 The macro definition should use @code{push_reload} to indicate parts that
5625 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5626 suitable to be passed unaltered to @code{push_reload}.
5628 The code generated by this macro must not alter the substructure of
5629 @var{x}. If it transforms @var{x} into a more legitimate form, it
5630 should assign @var{x} (which will always be a C variable) a new value.
5631 This also applies to parts that you change indirectly by calling
5634 @findex strict_memory_address_p
5635 The macro definition may use @code{strict_memory_address_p} to test if
5636 the address has become legitimate.
5639 If you want to change only a part of @var{x}, one standard way of doing
5640 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5641 single level of rtl. Thus, if the part to be changed is not at the
5642 top level, you'll need to replace first the top level.
5643 It is not necessary for this macro to come up with a legitimate
5644 address; but often a machine-dependent strategy can generate better code.
5647 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr}, addr_space_t @var{addrspace})
5648 This hook returns @code{true} if memory address @var{addr} in address
5649 space @var{addrspace} can have
5650 different meanings depending on the machine mode of the memory
5651 reference it is used for or if the address is valid for some modes
5654 Autoincrement and autodecrement addresses typically have mode-dependent
5655 effects because the amount of the increment or decrement is the size
5656 of the operand being addressed. Some machines have other mode-dependent
5657 addresses. Many RISC machines have no mode-dependent addresses.
5659 You may assume that @var{addr} is a valid address for the machine.
5661 The default version of this hook returns @code{false}.
5664 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (machine_mode @var{mode}, rtx @var{x})
5665 This hook returns true if @var{x} is a legitimate constant for a
5666 @var{mode}-mode immediate operand on the target machine. You can assume that
5667 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5669 The default definition returns true.
5672 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5673 This hook is used to undo the possibly obfuscating effects of the
5674 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5675 macros. Some backend implementations of these macros wrap symbol
5676 references inside an @code{UNSPEC} rtx to represent PIC or similar
5677 addressing modes. This target hook allows GCC's optimizers to understand
5678 the semantics of these opaque @code{UNSPEC}s by converting them back
5679 into their original form.
5682 @deftypefn {Target Hook} bool TARGET_CONST_NOT_OK_FOR_DEBUG_P (rtx @var{x})
5683 This hook should return true if @var{x} should not be emitted into
5687 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (machine_mode @var{mode}, rtx @var{x})
5688 This hook should return true if @var{x} is of a form that cannot (or
5689 should not) be spilled to the constant pool. @var{mode} is the mode
5692 The default version of this hook returns false.
5694 The primary reason to define this hook is to prevent reload from
5695 deciding that a non-legitimate constant would be better reloaded
5696 from the constant pool instead of spilling and reloading a register
5697 holding the constant. This restriction is often true of addresses
5698 of TLS symbols for various targets.
5701 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (machine_mode @var{mode}, const_rtx @var{x})
5702 This hook should return true if pool entries for constant @var{x} can
5703 be placed in an @code{object_block} structure. @var{mode} is the mode
5706 The default version returns false for all constants.
5709 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_DECL_P (const_tree @var{decl})
5710 This hook should return true if pool entries for @var{decl} should
5711 be placed in an @code{object_block} structure.
5713 The default version returns true for all decls.
5716 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (tree @var{fndecl})
5717 This hook should return the DECL of a function that implements the
5718 reciprocal of the machine-specific builtin function @var{fndecl}, or
5719 @code{NULL_TREE} if such a function is not available.
5722 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5723 This hook should return the DECL of a function @var{f} that given an
5724 address @var{addr} as an argument returns a mask @var{m} that can be
5725 used to extract from two vectors the relevant data that resides in
5726 @var{addr} in case @var{addr} is not properly aligned.
5728 The autovectorizer, when vectorizing a load operation from an address
5729 @var{addr} that may be unaligned, will generate two vector loads from
5730 the two aligned addresses around @var{addr}. It then generates a
5731 @code{REALIGN_LOAD} operation to extract the relevant data from the
5732 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5733 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5734 the third argument, @var{OFF}, defines how the data will be extracted
5735 from these two vectors: if @var{OFF} is 0, then the returned vector is
5736 @var{v2}; otherwise, the returned vector is composed from the last
5737 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5738 @var{OFF} elements of @var{v2}.
5740 If this hook is defined, the autovectorizer will generate a call
5741 to @var{f} (using the DECL tree that this hook returns) and will
5742 use the return value of @var{f} as the argument @var{OFF} to
5743 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5744 should comply with the semantics expected by @code{REALIGN_LOAD}
5746 If this hook is not defined, then @var{addr} will be used as
5747 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5748 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5751 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5752 Returns cost of different scalar or vector statements for vectorization cost model.
5753 For vector memory operations the cost may depend on type (@var{vectype}) and
5754 misalignment value (@var{misalign}).
5757 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5758 Return true if vector alignment is reachable (by peeling N iterations) for the given scalar type @var{type}. @var{is_packed} is false if the scalar access using @var{type} is known to be naturally aligned.
5761 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VEC_PERM_CONST_OK (machine_mode, const unsigned char *@var{sel})
5762 Return true if a vector created for @code{vec_perm_const} is valid.
5765 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5766 This hook should return the DECL of a function that implements conversion of the
5767 input vector of type @var{src_type} to type @var{dest_type}.
5768 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5769 specifies how the conversion is to be applied
5770 (truncation, rounding, etc.).
5772 If this hook is defined, the autovectorizer will use the
5773 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5774 conversion. Otherwise, it will return @code{NULL_TREE}.
5777 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (unsigned @var{code}, tree @var{vec_type_out}, tree @var{vec_type_in})
5778 This hook should return the decl of a function that implements the
5779 vectorized variant of the function with the @code{combined_fn} code
5780 @var{code} or @code{NULL_TREE} if such a function is not available.
5781 The return type of the vectorized function shall be of vector type
5782 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5785 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5786 This hook should return the decl of a function that implements the
5787 vectorized variant of target built-in function @code{fndecl}. The
5788 return type of the vectorized function shall be of vector type
5789 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5792 @deftypefn {Target Hook} bool TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT (machine_mode @var{mode}, const_tree @var{type}, int @var{misalignment}, bool @var{is_packed})
5793 This hook should return true if the target supports misaligned vector
5794 store/load of a specific factor denoted in the @var{misalignment}
5795 parameter. The vector store/load should be of machine mode @var{mode} and
5796 the elements in the vectors should be of type @var{type}. @var{is_packed}
5797 parameter is true if the memory access is defined in a packed struct.
5800 @deftypefn {Target Hook} machine_mode TARGET_VECTORIZE_PREFERRED_SIMD_MODE (scalar_mode @var{mode})
5801 This hook should return the preferred mode for vectorizing scalar
5802 mode @var{mode}. The default is
5803 equal to @code{word_mode}, because the vectorizer can do some
5804 transformations even in absence of specialized @acronym{SIMD} hardware.
5807 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void)
5808 This hook should return a mask of sizes that should be iterated over
5809 after trying to autovectorize using the vector size derived from the
5810 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5811 The default is zero which means to not iterate over other vector sizes.
5814 @deftypefn {Target Hook} opt_machine_mode TARGET_VECTORIZE_GET_MASK_MODE (unsigned @var{nunits}, unsigned @var{length})
5815 A vector mask is a value that holds one boolean result for every element
5816 in a vector. This hook returns the machine mode that should be used to
5817 represent such a mask when the vector in question is @var{length} bytes
5818 long and contains @var{nunits} elements. The hook returns an empty
5819 @code{opt_machine_mode} if no such mode exists.
5821 The default implementation returns the mode of an integer vector that
5822 is @var{length} bytes long and that contains @var{nunits} elements,
5823 if such a mode exists.
5826 @deftypefn {Target Hook} {void *} TARGET_VECTORIZE_INIT_COST (struct loop *@var{loop_info})
5827 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.
5830 @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})
5831 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.
5834 @deftypefn {Target Hook} void TARGET_VECTORIZE_FINISH_COST (void *@var{data}, unsigned *@var{prologue_cost}, unsigned *@var{body_cost}, unsigned *@var{epilogue_cost})
5835 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.
5838 @deftypefn {Target Hook} void TARGET_VECTORIZE_DESTROY_COST_DATA (void *@var{data})
5839 This hook should release @var{data} and any related data structures allocated by TARGET_VECTORIZE_INIT_COST. The default releases the accumulator.
5842 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_GATHER (const_tree @var{mem_vectype}, const_tree @var{index_type}, int @var{scale})
5843 Target builtin that implements vector gather operation. @var{mem_vectype}
5844 is the vector type of the load and @var{index_type} is scalar type of
5845 the index, scaled by @var{scale}.
5846 The default is @code{NULL_TREE} which means to not vectorize gather
5850 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_SCATTER (const_tree @var{vectype}, const_tree @var{index_type}, int @var{scale})
5851 Target builtin that implements vector scatter operation. @var{vectype}
5852 is the vector type of the store and @var{index_type} is scalar type of
5853 the index, scaled by @var{scale}.
5854 The default is @code{NULL_TREE} which means to not vectorize scatter
5858 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN (struct cgraph_node *@var{}, struct cgraph_simd_clone *@var{}, @var{tree}, @var{int})
5859 This hook should set @var{vecsize_mangle}, @var{vecsize_int}, @var{vecsize_float}
5860 fields in @var{simd_clone} structure pointed by @var{clone_info} argument and also
5861 @var{simdlen} field if it was previously 0.
5862 The hook should return 0 if SIMD clones shouldn't be emitted,
5863 or number of @var{vecsize_mangle} variants that should be emitted.
5866 @deftypefn {Target Hook} void TARGET_SIMD_CLONE_ADJUST (struct cgraph_node *@var{})
5867 This hook should add implicit @code{attribute(target("..."))} attribute
5868 to SIMD clone @var{node} if needed.
5871 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_USABLE (struct cgraph_node *@var{})
5872 This hook should return -1 if SIMD clone @var{node} shouldn't be used
5873 in vectorized loops in current function, or non-negative number if it is
5874 usable. In that case, the smaller the number is, the more desirable it is
5878 @deftypefn {Target Hook} int TARGET_SIMT_VF (void)
5879 Return number of threads in SIMT thread group on the target.
5882 @deftypefn {Target Hook} bool TARGET_GOACC_VALIDATE_DIMS (tree @var{decl}, int *@var{dims}, int @var{fn_level})
5883 This hook should check the launch dimensions provided for an OpenACC
5884 compute region, or routine. Defaulted values are represented as -1
5885 and non-constant values as 0. The @var{fn_level} is negative for the
5886 function corresponding to the compute region. For a routine is is the
5887 outermost level at which partitioned execution may be spawned. The hook
5888 should verify non-default values. If DECL is NULL, global defaults
5889 are being validated and unspecified defaults should be filled in.
5890 Diagnostics should be issued as appropriate. Return
5891 true, if changes have been made. You must override this hook to
5892 provide dimensions larger than 1.
5895 @deftypefn {Target Hook} int TARGET_GOACC_DIM_LIMIT (int @var{axis})
5896 This hook should return the maximum size of a particular dimension,
5897 or zero if unbounded.
5900 @deftypefn {Target Hook} bool TARGET_GOACC_FORK_JOIN (gcall *@var{call}, const int *@var{dims}, bool @var{is_fork})
5901 This hook can be used to convert IFN_GOACC_FORK and IFN_GOACC_JOIN
5902 function calls to target-specific gimple, or indicate whether they
5903 should be retained. It is executed during the oacc_device_lower pass.
5904 It should return true, if the call should be retained. It should
5905 return false, if it is to be deleted (either because target-specific
5906 gimple has been inserted before it, or there is no need for it).
5907 The default hook returns false, if there are no RTL expanders for them.
5910 @deftypefn {Target Hook} void TARGET_GOACC_REDUCTION (gcall *@var{call})
5911 This hook is used by the oacc_transform pass to expand calls to the
5912 @var{GOACC_REDUCTION} internal function, into a sequence of gimple
5913 instructions. @var{call} is gimple statement containing the call to
5914 the function. This hook removes statement @var{call} after the
5915 expanded sequence has been inserted. This hook is also responsible
5916 for allocating any storage for reductions when necessary.
5919 @node Anchored Addresses
5920 @section Anchored Addresses
5921 @cindex anchored addresses
5922 @cindex @option{-fsection-anchors}
5924 GCC usually addresses every static object as a separate entity.
5925 For example, if we have:
5929 int foo (void) @{ return a + b + c; @}
5932 the code for @code{foo} will usually calculate three separate symbolic
5933 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5934 it would be better to calculate just one symbolic address and access
5935 the three variables relative to it. The equivalent pseudocode would
5941 register int *xr = &x;
5942 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5946 (which isn't valid C). We refer to shared addresses like @code{x} as
5947 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5949 The hooks below describe the target properties that GCC needs to know
5950 in order to make effective use of section anchors. It won't use
5951 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5952 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5954 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5955 The minimum offset that should be applied to a section anchor.
5956 On most targets, it should be the smallest offset that can be
5957 applied to a base register while still giving a legitimate address
5958 for every mode. The default value is 0.
5961 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5962 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5963 offset that should be applied to section anchors. The default
5967 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5968 Write the assembly code to define section anchor @var{x}, which is a
5969 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5970 The hook is called with the assembly output position set to the beginning
5971 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5973 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5974 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5975 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5976 is @code{NULL}, which disables the use of section anchors altogether.
5979 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
5980 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5981 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5982 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5984 The default version is correct for most targets, but you might need to
5985 intercept this hook to handle things like target-specific attributes
5986 or target-specific sections.
5989 @node Condition Code
5990 @section Condition Code Status
5991 @cindex condition code status
5993 The macros in this section can be split in two families, according to the
5994 two ways of representing condition codes in GCC.
5996 The first representation is the so called @code{(cc0)} representation
5997 (@pxref{Jump Patterns}), where all instructions can have an implicit
5998 clobber of the condition codes. The second is the condition code
5999 register representation, which provides better schedulability for
6000 architectures that do have a condition code register, but on which
6001 most instructions do not affect it. The latter category includes
6004 The implicit clobbering poses a strong restriction on the placement of
6005 the definition and use of the condition code. In the past the definition
6006 and use were always adjacent. However, recent changes to support trapping
6007 arithmatic may result in the definition and user being in different blocks.
6008 Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally,
6009 the definition may be the source of exception handling edges.
6011 These restrictions can prevent important
6012 optimizations on some machines. For example, on the IBM RS/6000, there
6013 is a delay for taken branches unless the condition code register is set
6014 three instructions earlier than the conditional branch. The instruction
6015 scheduler cannot perform this optimization if it is not permitted to
6016 separate the definition and use of the condition code register.
6018 For this reason, it is possible and suggested to use a register to
6019 represent the condition code for new ports. If there is a specific
6020 condition code register in the machine, use a hard register. If the
6021 condition code or comparison result can be placed in any general register,
6022 or if there are multiple condition registers, use a pseudo register.
6023 Registers used to store the condition code value will usually have a mode
6024 that is in class @code{MODE_CC}.
6026 Alternatively, you can use @code{BImode} if the comparison operator is
6027 specified already in the compare instruction. In this case, you are not
6028 interested in most macros in this section.
6031 * CC0 Condition Codes:: Old style representation of condition codes.
6032 * MODE_CC Condition Codes:: Modern representation of condition codes.
6035 @node CC0 Condition Codes
6036 @subsection Representation of condition codes using @code{(cc0)}
6040 The file @file{conditions.h} defines a variable @code{cc_status} to
6041 describe how the condition code was computed (in case the interpretation of
6042 the condition code depends on the instruction that it was set by). This
6043 variable contains the RTL expressions on which the condition code is
6044 currently based, and several standard flags.
6046 Sometimes additional machine-specific flags must be defined in the machine
6047 description header file. It can also add additional machine-specific
6048 information by defining @code{CC_STATUS_MDEP}.
6050 @defmac CC_STATUS_MDEP
6051 C code for a data type which is used for declaring the @code{mdep}
6052 component of @code{cc_status}. It defaults to @code{int}.
6054 This macro is not used on machines that do not use @code{cc0}.
6057 @defmac CC_STATUS_MDEP_INIT
6058 A C expression to initialize the @code{mdep} field to ``empty''.
6059 The default definition does nothing, since most machines don't use
6060 the field anyway. If you want to use the field, you should probably
6061 define this macro to initialize it.
6063 This macro is not used on machines that do not use @code{cc0}.
6066 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
6067 A C compound statement to set the components of @code{cc_status}
6068 appropriately for an insn @var{insn} whose body is @var{exp}. It is
6069 this macro's responsibility to recognize insns that set the condition
6070 code as a byproduct of other activity as well as those that explicitly
6073 This macro is not used on machines that do not use @code{cc0}.
6075 If there are insns that do not set the condition code but do alter
6076 other machine registers, this macro must check to see whether they
6077 invalidate the expressions that the condition code is recorded as
6078 reflecting. For example, on the 68000, insns that store in address
6079 registers do not set the condition code, which means that usually
6080 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
6081 insns. But suppose that the previous insn set the condition code
6082 based on location @samp{a4@@(102)} and the current insn stores a new
6083 value in @samp{a4}. Although the condition code is not changed by
6084 this, it will no longer be true that it reflects the contents of
6085 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
6086 @code{cc_status} in this case to say that nothing is known about the
6087 condition code value.
6089 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
6090 with the results of peephole optimization: insns whose patterns are
6091 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
6092 constants which are just the operands. The RTL structure of these
6093 insns is not sufficient to indicate what the insns actually do. What
6094 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
6095 @code{CC_STATUS_INIT}.
6097 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
6098 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
6099 @samp{cc}. This avoids having detailed information about patterns in
6100 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
6103 @node MODE_CC Condition Codes
6104 @subsection Representation of condition codes using registers
6108 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
6109 On many machines, the condition code may be produced by other instructions
6110 than compares, for example the branch can use directly the condition
6111 code set by a subtract instruction. However, on some machines
6112 when the condition code is set this way some bits (such as the overflow
6113 bit) are not set in the same way as a test instruction, so that a different
6114 branch instruction must be used for some conditional branches. When
6115 this happens, use the machine mode of the condition code register to
6116 record different formats of the condition code register. Modes can
6117 also be used to record which compare instruction (e.g. a signed or an
6118 unsigned comparison) produced the condition codes.
6120 If other modes than @code{CCmode} are required, add them to
6121 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
6122 a mode given an operand of a compare. This is needed because the modes
6123 have to be chosen not only during RTL generation but also, for example,
6124 by instruction combination. The result of @code{SELECT_CC_MODE} should
6125 be consistent with the mode used in the patterns; for example to support
6126 the case of the add on the SPARC discussed above, we have the pattern
6132 (plus:SI (match_operand:SI 0 "register_operand" "%r")
6133 (match_operand:SI 1 "arith_operand" "rI"))
6140 together with a @code{SELECT_CC_MODE} that returns @code{CCNZmode}
6141 for comparisons whose argument is a @code{plus}:
6144 #define SELECT_CC_MODE(OP,X,Y) \
6145 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
6146 ? ((OP == LT || OP == LE || OP == GT || OP == GE) \
6147 ? CCFPEmode : CCFPmode) \
6148 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
6149 || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \
6150 ? CCNZmode : CCmode))
6153 Another reason to use modes is to retain information on which operands
6154 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
6157 You should define this macro if and only if you define extra CC modes
6158 in @file{@var{machine}-modes.def}.
6161 @deftypefn {Target Hook} void TARGET_CANONICALIZE_COMPARISON (int *@var{code}, rtx *@var{op0}, rtx *@var{op1}, bool @var{op0_preserve_value})
6162 On some machines not all possible comparisons are defined, but you can
6163 convert an invalid comparison into a valid one. For example, the Alpha
6164 does not have a @code{GT} comparison, but you can use an @code{LT}
6165 comparison instead and swap the order of the operands.
6167 On such machines, implement this hook to do any required conversions.
6168 @var{code} is the initial comparison code and @var{op0} and @var{op1}
6169 are the left and right operands of the comparison, respectively. If
6170 @var{op0_preserve_value} is @code{true} the implementation is not
6171 allowed to change the value of @var{op0} since the value might be used
6172 in RTXs which aren't comparisons. E.g. the implementation is not
6173 allowed to swap operands in that case.
6175 GCC will not assume that the comparison resulting from this macro is
6176 valid but will see if the resulting insn matches a pattern in the
6179 You need not to implement this hook if it would never change the
6180 comparison code or operands.
6183 @defmac REVERSIBLE_CC_MODE (@var{mode})
6184 A C expression whose value is one if it is always safe to reverse a
6185 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6186 can ever return @var{mode} for a floating-point inequality comparison,
6187 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6189 You need not define this macro if it would always returns zero or if the
6190 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6191 For example, here is the definition used on the SPARC, where floating-point
6192 inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}:
6195 #define REVERSIBLE_CC_MODE(MODE) \
6196 ((MODE) != CCFPEmode && (MODE) != CCFPmode)
6200 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6201 A C expression whose value is reversed condition code of the @var{code} for
6202 comparison done in CC_MODE @var{mode}. The macro is used only in case
6203 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6204 machine has some non-standard way how to reverse certain conditionals. For
6205 instance in case all floating point conditions are non-trapping, compiler may
6206 freely convert unordered compares to ordered ones. Then definition may look
6210 #define REVERSE_CONDITION(CODE, MODE) \
6211 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6212 : reverse_condition_maybe_unordered (CODE))
6216 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6217 On targets which do not use @code{(cc0)}, and which use a hard
6218 register rather than a pseudo-register to hold condition codes, the
6219 regular CSE passes are often not able to identify cases in which the
6220 hard register is set to a common value. Use this hook to enable a
6221 small pass which optimizes such cases. This hook should return true
6222 to enable this pass, and it should set the integers to which its
6223 arguments point to the hard register numbers used for condition codes.
6224 When there is only one such register, as is true on most systems, the
6225 integer pointed to by @var{p2} should be set to
6226 @code{INVALID_REGNUM}.
6228 The default version of this hook returns false.
6231 @deftypefn {Target Hook} machine_mode TARGET_CC_MODES_COMPATIBLE (machine_mode @var{m1}, machine_mode @var{m2})
6232 On targets which use multiple condition code modes in class
6233 @code{MODE_CC}, it is sometimes the case that a comparison can be
6234 validly done in more than one mode. On such a system, define this
6235 target hook to take two mode arguments and to return a mode in which
6236 both comparisons may be validly done. If there is no such mode,
6237 return @code{VOIDmode}.
6239 The default version of this hook checks whether the modes are the
6240 same. If they are, it returns that mode. If they are different, it
6241 returns @code{VOIDmode}.
6244 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
6245 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.
6249 @section Describing Relative Costs of Operations
6250 @cindex costs of instructions
6251 @cindex relative costs
6252 @cindex speed of instructions
6254 These macros let you describe the relative speed of various operations
6255 on the target machine.
6257 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6258 A C expression for the cost of moving data of mode @var{mode} from a
6259 register in class @var{from} to one in class @var{to}. The classes are
6260 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6261 value of 2 is the default; other values are interpreted relative to
6264 It is not required that the cost always equal 2 when @var{from} is the
6265 same as @var{to}; on some machines it is expensive to move between
6266 registers if they are not general registers.
6268 If reload sees an insn consisting of a single @code{set} between two
6269 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6270 classes returns a value of 2, reload does not check to ensure that the
6271 constraints of the insn are met. Setting a cost of other than 2 will
6272 allow reload to verify that the constraints are met. You should do this
6273 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6275 These macros are obsolete, new ports should use the target hook
6276 @code{TARGET_REGISTER_MOVE_COST} instead.
6279 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6280 This target hook should return the cost of moving data of mode @var{mode}
6281 from a register in class @var{from} to one in class @var{to}. The classes
6282 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6283 A value of 2 is the default; other values are interpreted relative to
6286 It is not required that the cost always equal 2 when @var{from} is the
6287 same as @var{to}; on some machines it is expensive to move between
6288 registers if they are not general registers.
6290 If reload sees an insn consisting of a single @code{set} between two
6291 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6292 classes returns a value of 2, reload does not check to ensure that the
6293 constraints of the insn are met. Setting a cost of other than 2 will
6294 allow reload to verify that the constraints are met. You should do this
6295 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6297 The default version of this function returns 2.
6300 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6301 A C expression for the cost of moving data of mode @var{mode} between a
6302 register of class @var{class} and memory; @var{in} is zero if the value
6303 is to be written to memory, nonzero if it is to be read in. This cost
6304 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6305 registers and memory is more expensive than between two registers, you
6306 should define this macro to express the relative cost.
6308 If you do not define this macro, GCC uses a default cost of 4 plus
6309 the cost of copying via a secondary reload register, if one is
6310 needed. If your machine requires a secondary reload register to copy
6311 between memory and a register of @var{class} but the reload mechanism is
6312 more complex than copying via an intermediate, define this macro to
6313 reflect the actual cost of the move.
6315 GCC defines the function @code{memory_move_secondary_cost} if
6316 secondary reloads are needed. It computes the costs due to copying via
6317 a secondary register. If your machine copies from memory using a
6318 secondary register in the conventional way but the default base value of
6319 4 is not correct for your machine, define this macro to add some other
6320 value to the result of that function. The arguments to that function
6321 are the same as to this macro.
6323 These macros are obsolete, new ports should use the target hook
6324 @code{TARGET_MEMORY_MOVE_COST} instead.
6327 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6328 This target hook should return the cost of moving data of mode @var{mode}
6329 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6330 if the value is to be written to memory, @code{true} if it is to be read in.
6331 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6332 If moving between registers and memory is more expensive than between two
6333 registers, you should add this target hook to express the relative cost.
6335 If you do not add this target hook, GCC uses a default cost of 4 plus
6336 the cost of copying via a secondary reload register, if one is
6337 needed. If your machine requires a secondary reload register to copy
6338 between memory and a register of @var{rclass} but the reload mechanism is
6339 more complex than copying via an intermediate, use this target hook to
6340 reflect the actual cost of the move.
6342 GCC defines the function @code{memory_move_secondary_cost} if
6343 secondary reloads are needed. It computes the costs due to copying via
6344 a secondary register. If your machine copies from memory using a
6345 secondary register in the conventional way but the default base value of
6346 4 is not correct for your machine, use this target hook to add some other
6347 value to the result of that function. The arguments to that function
6348 are the same as to this target hook.
6351 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6352 A C expression for the cost of a branch instruction. A value of 1 is
6353 the default; other values are interpreted relative to that. Parameter
6354 @var{speed_p} is true when the branch in question should be optimized
6355 for speed. When it is false, @code{BRANCH_COST} should return a value
6356 optimal for code size rather than performance. @var{predictable_p} is
6357 true for well-predicted branches. On many architectures the
6358 @code{BRANCH_COST} can be reduced then.
6361 Here are additional macros which do not specify precise relative costs,
6362 but only that certain actions are more expensive than GCC would
6365 @defmac SLOW_BYTE_ACCESS
6366 Define this macro as a C expression which is nonzero if accessing less
6367 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6368 faster than accessing a word of memory, i.e., if such access
6369 require more than one instruction or if there is no difference in cost
6370 between byte and (aligned) word loads.
6372 When this macro is not defined, the compiler will access a field by
6373 finding the smallest containing object; when it is defined, a fullword
6374 load will be used if alignment permits. Unless bytes accesses are
6375 faster than word accesses, using word accesses is preferable since it
6376 may eliminate subsequent memory access if subsequent accesses occur to
6377 other fields in the same word of the structure, but to different bytes.
6380 @deftypefn {Target Hook} bool TARGET_SLOW_UNALIGNED_ACCESS (machine_mode @var{mode}, unsigned int @var{align})
6381 This hook returns true if memory accesses described by the
6382 @var{mode} and @var{alignment} parameters have a cost many times greater
6383 than aligned accesses, for example if they are emulated in a trap handler.
6384 This hook is invoked only for unaligned accesses, i.e. when
6385 @code{@var{alignment} < GET_MODE_ALIGNMENT (@var{mode})}.
6387 When this hook returns true, the compiler will act as if
6388 @code{STRICT_ALIGNMENT} were true when generating code for block
6389 moves. This can cause significantly more instructions to be produced.
6390 Therefore, do not make this hook return true if unaligned accesses only
6391 add a cycle or two to the time for a memory access.
6393 The hook must return true whenever @code{STRICT_ALIGNMENT} is true.
6394 The default implementation returns @code{STRICT_ALIGNMENT}.
6397 @defmac MOVE_RATIO (@var{speed})
6398 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6399 which a sequence of insns should be generated instead of a
6400 string move insn or a library call. Increasing the value will always
6401 make code faster, but eventually incurs high cost in increased code size.
6403 Note that on machines where the corresponding move insn is a
6404 @code{define_expand} that emits a sequence of insns, this macro counts
6405 the number of such sequences.
6407 The parameter @var{speed} is true if the code is currently being
6408 optimized for speed rather than size.
6410 If you don't define this, a reasonable default is used.
6413 @deftypefn {Target Hook} bool TARGET_USE_BY_PIECES_INFRASTRUCTURE_P (unsigned HOST_WIDE_INT @var{size}, unsigned int @var{alignment}, enum by_pieces_operation @var{op}, bool @var{speed_p})
6414 GCC will attempt several strategies when asked to copy between
6415 two areas of memory, or to set, clear or store to memory, for example
6416 when copying a @code{struct}. The @code{by_pieces} infrastructure
6417 implements such memory operations as a sequence of load, store or move
6418 insns. Alternate strategies are to expand the
6419 @code{movmem} or @code{setmem} optabs, to emit a library call, or to emit
6420 unit-by-unit, loop-based operations.
6422 This target hook should return true if, for a memory operation with a
6423 given @var{size} and @var{alignment}, using the @code{by_pieces}
6424 infrastructure is expected to result in better code generation.
6425 Both @var{size} and @var{alignment} are measured in terms of storage
6428 The parameter @var{op} is one of: @code{CLEAR_BY_PIECES},
6429 @code{MOVE_BY_PIECES}, @code{SET_BY_PIECES}, @code{STORE_BY_PIECES} or
6430 @code{COMPARE_BY_PIECES}. These describe the type of memory operation
6431 under consideration.
6433 The parameter @var{speed_p} is true if the code is currently being
6434 optimized for speed rather than size.
6436 Returning true for higher values of @var{size} can improve code generation
6437 for speed if the target does not provide an implementation of the
6438 @code{movmem} or @code{setmem} standard names, if the @code{movmem} or
6439 @code{setmem} implementation would be more expensive than a sequence of
6440 insns, or if the overhead of a library call would dominate that of
6441 the body of the memory operation.
6443 Returning true for higher values of @code{size} may also cause an increase
6444 in code size, for example where the number of insns emitted to perform a
6445 move would be greater than that of a library call.
6448 @deftypefn {Target Hook} int TARGET_COMPARE_BY_PIECES_BRANCH_RATIO (machine_mode @var{mode})
6449 When expanding a block comparison in MODE, gcc can try to reduce the
6450 number of branches at the expense of more memory operations. This hook
6451 allows the target to override the default choice. It should return the
6452 factor by which branches should be reduced over the plain expansion with
6453 one comparison per @var{mode}-sized piece. A port can also prevent a
6454 particular mode from being used for block comparisons by returning a
6455 negative number from this hook.
6458 @defmac MOVE_MAX_PIECES
6459 A C expression used by @code{move_by_pieces} to determine the largest unit
6460 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6463 @defmac STORE_MAX_PIECES
6464 A C expression used by @code{store_by_pieces} to determine the largest unit
6465 a store used to memory is. Defaults to @code{MOVE_MAX_PIECES}, or two times
6466 the size of @code{HOST_WIDE_INT}, whichever is smaller.
6469 @defmac COMPARE_MAX_PIECES
6470 A C expression used by @code{compare_by_pieces} to determine the largest unit
6471 a load or store used to compare memory is. Defaults to
6472 @code{MOVE_MAX_PIECES}.
6475 @defmac CLEAR_RATIO (@var{speed})
6476 The threshold of number of scalar move insns, @emph{below} which a sequence
6477 of insns should be generated to clear memory instead of a string clear insn
6478 or a library call. Increasing the value will always make code faster, but
6479 eventually incurs high cost in increased code size.
6481 The parameter @var{speed} is true if the code is currently being
6482 optimized for speed rather than size.
6484 If you don't define this, a reasonable default is used.
6487 @defmac SET_RATIO (@var{speed})
6488 The threshold of number of scalar move insns, @emph{below} which a sequence
6489 of insns should be generated to set memory to a constant value, instead of
6490 a block set insn or a library call.
6491 Increasing the value will always make code faster, but
6492 eventually incurs high cost in increased code size.
6494 The parameter @var{speed} is true if the code is currently being
6495 optimized for speed rather than size.
6497 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6500 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6501 A C expression used to determine whether a load postincrement is a good
6502 thing to use for a given mode. Defaults to the value of
6503 @code{HAVE_POST_INCREMENT}.
6506 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6507 A C expression used to determine whether a load postdecrement is a good
6508 thing to use for a given mode. Defaults to the value of
6509 @code{HAVE_POST_DECREMENT}.
6512 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6513 A C expression used to determine whether a load preincrement is a good
6514 thing to use for a given mode. Defaults to the value of
6515 @code{HAVE_PRE_INCREMENT}.
6518 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6519 A C expression used to determine whether a load predecrement is a good
6520 thing to use for a given mode. Defaults to the value of
6521 @code{HAVE_PRE_DECREMENT}.
6524 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6525 A C expression used to determine whether a store postincrement is a good
6526 thing to use for a given mode. Defaults to the value of
6527 @code{HAVE_POST_INCREMENT}.
6530 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6531 A C expression used to determine whether a store postdecrement is a good
6532 thing to use for a given mode. Defaults to the value of
6533 @code{HAVE_POST_DECREMENT}.
6536 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6537 This macro is used to determine whether a store preincrement is a good
6538 thing to use for a given mode. Defaults to the value of
6539 @code{HAVE_PRE_INCREMENT}.
6542 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6543 This macro is used to determine whether a store predecrement is a good
6544 thing to use for a given mode. Defaults to the value of
6545 @code{HAVE_PRE_DECREMENT}.
6548 @defmac NO_FUNCTION_CSE
6549 Define this macro to be true if it is as good or better to call a constant
6550 function address than to call an address kept in a register.
6553 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
6554 Define this macro if a non-short-circuit operation produced by
6555 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6556 @code{BRANCH_COST} is greater than or equal to the value 2.
6559 @deftypefn {Target Hook} bool TARGET_OPTAB_SUPPORTED_P (int @var{op}, machine_mode @var{mode1}, machine_mode @var{mode2}, optimization_type @var{opt_type})
6560 Return true if the optimizers should use optab @var{op} with
6561 modes @var{mode1} and @var{mode2} for optimization type @var{opt_type}.
6562 The optab is known to have an associated @file{.md} instruction
6563 whose C condition is true. @var{mode2} is only meaningful for conversion
6564 optabs; for direct optabs it is a copy of @var{mode1}.
6566 For example, when called with @var{op} equal to @code{rint_optab} and
6567 @var{mode1} equal to @code{DFmode}, the hook should say whether the
6568 optimizers should use optab @code{rintdf2}.
6570 The default hook returns true for all inputs.
6573 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, machine_mode @var{mode}, int @var{outer_code}, int @var{opno}, int *@var{total}, bool @var{speed})
6574 This target hook describes the relative costs of RTL expressions.
6576 The cost may depend on the precise form of the expression, which is
6577 available for examination in @var{x}, and the fact that @var{x} appears
6578 as operand @var{opno} of an expression with rtx code @var{outer_code}.
6579 That is, the hook can assume that there is some rtx @var{y} such
6580 that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that
6581 either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or
6582 (b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}.
6584 @var{mode} is @var{x}'s machine mode, or for cases like @code{const_int} that
6585 do not have a mode, the mode in which @var{x} is used.
6587 In implementing this hook, you can use the construct
6588 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6591 On entry to the hook, @code{*@var{total}} contains a default estimate
6592 for the cost of the expression. The hook should modify this value as
6593 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6594 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6595 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6597 When optimizing for code size, i.e.@: when @code{speed} is
6598 false, this target hook should be used to estimate the relative
6599 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6601 The hook returns true when all subexpressions of @var{x} have been
6602 processed, and false when @code{rtx_cost} should recurse.
6605 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, machine_mode @var{mode}, addr_space_t @var{as}, bool @var{speed})
6606 This hook computes the cost of an addressing mode that contains
6607 @var{address}. If not defined, the cost is computed from
6608 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6610 For most CISC machines, the default cost is a good approximation of the
6611 true cost of the addressing mode. However, on RISC machines, all
6612 instructions normally have the same length and execution time. Hence
6613 all addresses will have equal costs.
6615 In cases where more than one form of an address is known, the form with
6616 the lowest cost will be used. If multiple forms have the same, lowest,
6617 cost, the one that is the most complex will be used.
6619 For example, suppose an address that is equal to the sum of a register
6620 and a constant is used twice in the same basic block. When this macro
6621 is not defined, the address will be computed in a register and memory
6622 references will be indirect through that register. On machines where
6623 the cost of the addressing mode containing the sum is no higher than
6624 that of a simple indirect reference, this will produce an additional
6625 instruction and possibly require an additional register. Proper
6626 specification of this macro eliminates this overhead for such machines.
6628 This hook is never called with an invalid address.
6630 On machines where an address involving more than one register is as
6631 cheap as an address computation involving only one register, defining
6632 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6633 be live over a region of code where only one would have been if
6634 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6635 should be considered in the definition of this macro. Equivalent costs
6636 should probably only be given to addresses with different numbers of
6637 registers on machines with lots of registers.
6640 @deftypefn {Target Hook} {unsigned int} TARGET_MAX_NOCE_IFCVT_SEQ_COST (edge @var{e})
6641 This hook returns a value in the same units as @code{TARGET_RTX_COSTS},
6642 giving the maximum acceptable cost for a sequence generated by the RTL
6643 if-conversion pass when conditional execution is not available.
6644 The RTL if-conversion pass attempts to convert conditional operations
6645 that would require a branch to a series of unconditional operations and
6646 @code{mov@var{mode}cc} insns. This hook returns the maximum cost of the
6647 unconditional instructions and the @code{mov@var{mode}cc} insns.
6648 RTL if-conversion is cancelled if the cost of the converted sequence
6649 is greater than the value returned by this hook.
6651 @code{e} is the edge between the basic block containing the conditional
6652 branch to the basic block which would be executed if the condition
6655 The default implementation of this hook uses the
6656 @code{max-rtl-if-conversion-[un]predictable} parameters if they are set,
6657 and uses a multiple of @code{BRANCH_COST} otherwise.
6660 @deftypefn {Target Hook} bool TARGET_NOCE_CONVERSION_PROFITABLE_P (rtx_insn *@var{seq}, struct noce_if_info *@var{if_info})
6661 This hook returns true if the instruction sequence @code{seq} is a good
6662 candidate as a replacement for the if-convertible sequence described in
6666 @deftypefn {Target Hook} bool TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P (void)
6667 This predicate controls the use of the eager delay slot filler to disallow
6668 speculatively executed instructions being placed in delay slots. Targets
6669 such as certain MIPS architectures possess both branches with and without
6670 delay slots. As the eager delay slot filler can decrease performance,
6671 disabling it is beneficial when ordinary branches are available. Use of
6672 delay slot branches filled using the basic filler is often still desirable
6673 as the delay slot can hide a pipeline bubble.
6677 @section Adjusting the Instruction Scheduler
6679 The instruction scheduler may need a fair amount of machine-specific
6680 adjustment in order to produce good code. GCC provides several target
6681 hooks for this purpose. It is usually enough to define just a few of
6682 them: try the first ones in this list first.
6684 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6685 This hook returns the maximum number of instructions that can ever
6686 issue at the same time on the target machine. The default is one.
6687 Although the insn scheduler can define itself the possibility of issue
6688 an insn on the same cycle, the value can serve as an additional
6689 constraint to issue insns on the same simulated processor cycle (see
6690 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6691 This value must be constant over the entire compilation. If you need
6692 it to vary depending on what the instructions are, you must use
6693 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6696 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx_insn *@var{insn}, int @var{more})
6697 This hook is executed by the scheduler after it has scheduled an insn
6698 from the ready list. It should return the number of insns which can
6699 still be issued in the current cycle. The default is
6700 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6701 @code{USE}, which normally are not counted against the issue rate.
6702 You should define this hook if some insns take more machine resources
6703 than others, so that fewer insns can follow them in the same cycle.
6704 @var{file} is either a null pointer, or a stdio stream to write any
6705 debug output to. @var{verbose} is the verbose level provided by
6706 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6710 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx_insn *@var{insn}, int @var{dep_type1}, rtx_insn *@var{dep_insn}, int @var{cost}, unsigned int @var{dw})
6711 This function corrects the value of @var{cost} based on the
6712 relationship between @var{insn} and @var{dep_insn} through a
6713 dependence of type dep_type, and strength @var{dw}. It should return the new
6714 value. The default is to make no adjustment to @var{cost}. This can be
6715 used for example to specify to the scheduler using the traditional pipeline
6716 description that an output- or anti-dependence does not incur the same cost
6717 as a data-dependence. If the scheduler using the automaton based pipeline
6718 description, the cost of anti-dependence is zero and the cost of
6719 output-dependence is maximum of one and the difference of latency
6720 times of the first and the second insns. If these values are not
6721 acceptable, you could use the hook to modify them too. See also
6722 @pxref{Processor pipeline description}.
6725 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx_insn *@var{insn}, int @var{priority})
6726 This hook adjusts the integer scheduling priority @var{priority} of
6727 @var{insn}. It should return the new priority. Increase the priority to
6728 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6729 later. Do not define this hook if you do not need to adjust the
6730 scheduling priorities of insns.
6733 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock})
6734 This hook is executed by the scheduler after it has scheduled the ready
6735 list, to allow the machine description to reorder it (for example to
6736 combine two small instructions together on @samp{VLIW} machines).
6737 @var{file} is either a null pointer, or a stdio stream to write any
6738 debug output to. @var{verbose} is the verbose level provided by
6739 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6740 list of instructions that are ready to be scheduled. @var{n_readyp} is
6741 a pointer to the number of elements in the ready list. The scheduler
6742 reads the ready list in reverse order, starting with
6743 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6744 is the timer tick of the scheduler. You may modify the ready list and
6745 the number of ready insns. The return value is the number of insns that
6746 can issue this cycle; normally this is just @code{issue_rate}. See also
6747 @samp{TARGET_SCHED_REORDER2}.
6750 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock})
6751 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6752 function is called whenever the scheduler starts a new cycle. This one
6753 is called once per iteration over a cycle, immediately after
6754 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6755 return the number of insns to be scheduled in the same cycle. Defining
6756 this hook can be useful if there are frequent situations where
6757 scheduling one insn causes other insns to become ready in the same
6758 cycle. These other insns can then be taken into account properly.
6761 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_P (void)
6762 This hook is used to check whether target platform supports macro fusion.
6765 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_PAIR_P (rtx_insn *@var{prev}, rtx_insn *@var{curr})
6766 This hook is used to check whether two insns should be macro fused for
6767 a target microarchitecture. If this hook returns true for the given insn pair
6768 (@var{prev} and @var{curr}), the scheduler will put them into a sched
6769 group, and they will not be scheduled apart. The two insns will be either
6770 two SET insns or a compare and a conditional jump and this hook should
6771 validate any dependencies needed to fuse the two insns together.
6774 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx_insn *@var{head}, rtx_insn *@var{tail})
6775 This hook is called after evaluation forward dependencies of insns in
6776 chain given by two parameter values (@var{head} and @var{tail}
6777 correspondingly) but before insns scheduling of the insn chain. For
6778 example, it can be used for better insn classification if it requires
6779 analysis of dependencies. This hook can use backward and forward
6780 dependencies of the insn scheduler because they are already
6784 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6785 This hook is executed by the scheduler at the beginning of each block of
6786 instructions that are to be scheduled. @var{file} is either a null
6787 pointer, or a stdio stream to write any debug output to. @var{verbose}
6788 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6789 @var{max_ready} is the maximum number of insns in the current scheduling
6790 region that can be live at the same time. This can be used to allocate
6791 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6794 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6795 This hook is executed by the scheduler at the end of each block of
6796 instructions that are to be scheduled. It can be used to perform
6797 cleanup of any actions done by the other scheduling hooks. @var{file}
6798 is either a null pointer, or a stdio stream to write any debug output
6799 to. @var{verbose} is the verbose level provided by
6800 @option{-fsched-verbose-@var{n}}.
6803 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6804 This hook is executed by the scheduler after function level initializations.
6805 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6806 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6807 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6810 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6811 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6812 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6813 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6816 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6817 The hook returns an RTL insn. The automaton state used in the
6818 pipeline hazard recognizer is changed as if the insn were scheduled
6819 when the new simulated processor cycle starts. Usage of the hook may
6820 simplify the automaton pipeline description for some @acronym{VLIW}
6821 processors. If the hook is defined, it is used only for the automaton
6822 based pipeline description. The default is not to change the state
6823 when the new simulated processor cycle starts.
6826 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6827 The hook can be used to initialize data used by the previous hook.
6830 @deftypefn {Target Hook} {rtx_insn *} TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6831 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6832 to changed the state as if the insn were scheduled when the new
6833 simulated processor cycle finishes.
6836 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6837 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6838 used to initialize data used by the previous hook.
6841 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
6842 The hook to notify target that the current simulated cycle is about to finish.
6843 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6844 to change the state in more complicated situations - e.g., when advancing
6845 state on a single insn is not enough.
6848 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
6849 The hook to notify target that new simulated cycle has just started.
6850 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6851 to change the state in more complicated situations - e.g., when advancing
6852 state on a single insn is not enough.
6855 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6856 This hook controls better choosing an insn from the ready insn queue
6857 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6858 chooses the first insn from the queue. If the hook returns a positive
6859 value, an additional scheduler code tries all permutations of
6860 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6861 subsequent ready insns to choose an insn whose issue will result in
6862 maximal number of issued insns on the same cycle. For the
6863 @acronym{VLIW} processor, the code could actually solve the problem of
6864 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6865 rules of @acronym{VLIW} packing are described in the automaton.
6867 This code also could be used for superscalar @acronym{RISC}
6868 processors. Let us consider a superscalar @acronym{RISC} processor
6869 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6870 @var{B}, some insns can be executed only in pipelines @var{B} or
6871 @var{C}, and one insn can be executed in pipeline @var{B}. The
6872 processor may issue the 1st insn into @var{A} and the 2nd one into
6873 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6874 until the next cycle. If the scheduler issues the 3rd insn the first,
6875 the processor could issue all 3 insns per cycle.
6877 Actually this code demonstrates advantages of the automaton based
6878 pipeline hazard recognizer. We try quickly and easy many insn
6879 schedules to choose the best one.
6881 The default is no multipass scheduling.
6884 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx_insn *@var{insn}, int @var{ready_index})
6886 This hook controls what insns from the ready insn queue will be
6887 considered for the multipass insn scheduling. If the hook returns
6888 zero for @var{insn}, the insn will be considered in multipass scheduling.
6889 Positive return values will remove @var{insn} from consideration on
6890 the current round of multipass scheduling.
6891 Negative return values will remove @var{insn} from consideration for given
6893 Backends should be careful about returning non-zero for highest priority
6894 instruction at position 0 in the ready list. @var{ready_index} is passed
6895 to allow backends make correct judgements.
6897 The default is that any ready insns can be chosen to be issued.
6900 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN (void *@var{data}, signed char *@var{ready_try}, int @var{n_ready}, bool @var{first_cycle_insn_p})
6901 This hook prepares the target backend for a new round of multipass
6905 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE (void *@var{data}, signed char *@var{ready_try}, int @var{n_ready}, rtx_insn *@var{insn}, const void *@var{prev_data})
6906 This hook is called when multipass scheduling evaluates instruction INSN.
6909 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, signed char *@var{ready_try}, int @var{n_ready})
6910 This is called when multipass scheduling backtracks from evaluation of
6914 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
6915 This hook notifies the target about the result of the concluded current
6916 round of multipass scheduling.
6919 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
6920 This hook initializes target-specific data used in multipass scheduling.
6923 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
6924 This hook finalizes target-specific data used in multipass scheduling.
6927 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *@var{dump}, int @var{verbose}, rtx_insn *@var{insn}, int @var{last_clock}, int @var{clock}, int *@var{sort_p})
6928 This hook is called by the insn scheduler before issuing @var{insn}
6929 on cycle @var{clock}. If the hook returns nonzero,
6930 @var{insn} is not issued on this processor cycle. Instead,
6931 the processor cycle is advanced. If *@var{sort_p}
6932 is zero, the insn ready queue is not sorted on the new cycle
6933 start as usually. @var{dump} and @var{verbose} specify the file and
6934 verbosity level to use for debugging output.
6935 @var{last_clock} and @var{clock} are, respectively, the
6936 processor cycle on which the previous insn has been issued,
6937 and the current processor cycle.
6940 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
6941 This hook is used to define which dependences are considered costly by
6942 the target, so costly that it is not advisable to schedule the insns that
6943 are involved in the dependence too close to one another. The parameters
6944 to this hook are as follows: The first parameter @var{_dep} is the dependence
6945 being evaluated. The second parameter @var{cost} is the cost of the
6946 dependence as estimated by the scheduler, and the third
6947 parameter @var{distance} is the distance in cycles between the two insns.
6948 The hook returns @code{true} if considering the distance between the two
6949 insns the dependence between them is considered costly by the target,
6950 and @code{false} otherwise.
6952 Defining this hook can be useful in multiple-issue out-of-order machines,
6953 where (a) it's practically hopeless to predict the actual data/resource
6954 delays, however: (b) there's a better chance to predict the actual grouping
6955 that will be formed, and (c) correctly emulating the grouping can be very
6956 important. In such targets one may want to allow issuing dependent insns
6957 closer to one another---i.e., closer than the dependence distance; however,
6958 not in cases of ``costly dependences'', which this hooks allows to define.
6961 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6962 This hook is called by the insn scheduler after emitting a new instruction to
6963 the instruction stream. The hook notifies a target backend to extend its
6964 per instruction data structures.
6967 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6968 Return a pointer to a store large enough to hold target scheduling context.
6971 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6972 Initialize store pointed to by @var{tc} to hold target scheduling context.
6973 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6974 beginning of the block. Otherwise, copy the current context into @var{tc}.
6977 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6978 Copy target scheduling context pointed to by @var{tc} to the current context.
6981 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6982 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6985 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6986 Deallocate a store for target scheduling context pointed to by @var{tc}.
6989 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx_insn *@var{insn}, unsigned int @var{dep_status}, rtx *@var{new_pat})
6990 This hook is called by the insn scheduler when @var{insn} has only
6991 speculative dependencies and therefore can be scheduled speculatively.
6992 The hook is used to check if the pattern of @var{insn} has a speculative
6993 version and, in case of successful check, to generate that speculative
6994 pattern. The hook should return 1, if the instruction has a speculative form,
6995 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6996 speculation. If the return value equals 1 then @var{new_pat} is assigned
6997 the generated speculative pattern.
7000 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (unsigned int @var{dep_status})
7001 This hook is called by the insn scheduler during generation of recovery code
7002 for @var{insn}. It should return @code{true}, if the corresponding check
7003 instruction should branch to recovery code, or @code{false} otherwise.
7006 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx_insn *@var{insn}, rtx_insn *@var{label}, unsigned int @var{ds})
7007 This hook is called by the insn scheduler to generate a pattern for recovery
7008 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
7009 speculative instruction for which the check should be generated.
7010 @var{label} is either a label of a basic block, where recovery code should
7011 be emitted, or a null pointer, when requested check doesn't branch to
7012 recovery code (a simple check). If @var{mutate_p} is nonzero, then
7013 a pattern for a branchy check corresponding to a simple check denoted by
7014 @var{insn} should be generated. In this case @var{label} can't be null.
7017 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
7018 This hook is used by the insn scheduler to find out what features should be
7020 The structure *@var{spec_info} should be filled in by the target.
7021 The structure describes speculation types that can be used in the scheduler.
7024 @deftypefn {Target Hook} bool TARGET_SCHED_CAN_SPECULATE_INSN (rtx_insn *@var{insn})
7025 Some instructions should never be speculated by the schedulers, usually
7026 because the instruction is too expensive to get this wrong. Often such
7027 instructions have long latency, and often they are not fully modeled in the
7028 pipeline descriptions. This hook should return @code{false} if @var{insn}
7029 should not be speculated.
7032 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
7033 This hook is called by the swing modulo scheduler to calculate a
7034 resource-based lower bound which is based on the resources available in
7035 the machine and the resources required by each instruction. The target
7036 backend can use @var{g} to calculate such bound. A very simple lower
7037 bound will be used in case this hook is not implemented: the total number
7038 of instructions divided by the issue rate.
7041 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx_insn *@var{insn}, int @var{x})
7042 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
7043 is supported in hardware and the condition specified in the parameter is true.
7046 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx_insn *@var{insn}, int @var{x})
7047 This hook is called by Haifa Scheduler. It performs the operation specified
7048 in its second parameter.
7051 @deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE
7052 True if the processor has an exposed pipeline, which means that not just
7053 the order of instructions is important for correctness when scheduling, but
7054 also the latencies of operations.
7057 @deftypefn {Target Hook} int TARGET_SCHED_REASSOCIATION_WIDTH (unsigned int @var{opc}, machine_mode @var{mode})
7058 This hook is called by tree reassociator to determine a level of
7059 parallelism required in output calculations chain.
7062 @deftypefn {Target Hook} void TARGET_SCHED_FUSION_PRIORITY (rtx_insn *@var{insn}, int @var{max_pri}, int *@var{fusion_pri}, int *@var{pri})
7063 This hook is called by scheduling fusion pass. It calculates fusion
7064 priorities for each instruction passed in by parameter. The priorities
7065 are returned via pointer parameters.
7067 @var{insn} is the instruction whose priorities need to be calculated.
7068 @var{max_pri} is the maximum priority can be returned in any cases.
7069 @var{fusion_pri} is the pointer parameter through which @var{insn}'s
7070 fusion priority should be calculated and returned.
7071 @var{pri} is the pointer parameter through which @var{insn}'s priority
7072 should be calculated and returned.
7074 Same @var{fusion_pri} should be returned for instructions which should
7075 be scheduled together. Different @var{pri} should be returned for
7076 instructions with same @var{fusion_pri}. @var{fusion_pri} is the major
7077 sort key, @var{pri} is the minor sort key. All instructions will be
7078 scheduled according to the two priorities. All priorities calculated
7079 should be between 0 (exclusive) and @var{max_pri} (inclusive). To avoid
7080 false dependencies, @var{fusion_pri} of instructions which need to be
7081 scheduled together should be smaller than @var{fusion_pri} of irrelevant
7084 Given below example:
7097 On targets like ARM/AArch64, the two pairs of consecutive loads should be
7098 merged. Since peephole2 pass can't help in this case unless consecutive
7099 loads are actually next to each other in instruction flow. That's where
7100 this scheduling fusion pass works. This hook calculates priority for each
7101 instruction based on its fustion type, like:
7104 ldr r10, [r1, 4] ; fusion_pri=99, pri=96
7105 add r4, r4, r10 ; fusion_pri=100, pri=100
7106 ldr r15, [r2, 8] ; fusion_pri=98, pri=92
7107 sub r5, r5, r15 ; fusion_pri=100, pri=100
7108 ldr r11, [r1, 0] ; fusion_pri=99, pri=100
7109 add r4, r4, r11 ; fusion_pri=100, pri=100
7110 ldr r16, [r2, 12] ; fusion_pri=98, pri=88
7111 sub r5, r5, r16 ; fusion_pri=100, pri=100
7114 Scheduling fusion pass then sorts all ready to issue instructions according
7115 to the priorities. As a result, instructions of same fusion type will be
7116 pushed together in instruction flow, like:
7129 Now peephole2 pass can simply merge the two pairs of loads.
7131 Since scheduling fusion pass relies on peephole2 to do real fusion
7132 work, it is only enabled by default when peephole2 is in effect.
7134 This is firstly introduced on ARM/AArch64 targets, please refer to
7135 the hook implementation for how different fusion types are supported.
7138 @deftypefn {Target Hook} void TARGET_EXPAND_DIVMOD_LIBFUNC (rtx @var{libfunc}, machine_mode @var{mode}, rtx @var{op0}, rtx @var{op1}, rtx *@var{quot}, rtx *@var{rem})
7139 Define this hook for enabling divmod transform if the port does not have
7140 hardware divmod insn but defines target-specific divmod libfuncs.
7144 @section Dividing the Output into Sections (Texts, Data, @dots{})
7145 @c the above section title is WAY too long. maybe cut the part between
7146 @c the (...)? --mew 10feb93
7148 An object file is divided into sections containing different types of
7149 data. In the most common case, there are three sections: the @dfn{text
7150 section}, which holds instructions and read-only data; the @dfn{data
7151 section}, which holds initialized writable data; and the @dfn{bss
7152 section}, which holds uninitialized data. Some systems have other kinds
7155 @file{varasm.c} provides several well-known sections, such as
7156 @code{text_section}, @code{data_section} and @code{bss_section}.
7157 The normal way of controlling a @code{@var{foo}_section} variable
7158 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
7159 as described below. The macros are only read once, when @file{varasm.c}
7160 initializes itself, so their values must be run-time constants.
7161 They may however depend on command-line flags.
7163 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
7164 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
7165 to be string literals.
7167 Some assemblers require a different string to be written every time a
7168 section is selected. If your assembler falls into this category, you
7169 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
7170 @code{get_unnamed_section} to set up the sections.
7172 You must always create a @code{text_section}, either by defining
7173 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
7174 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
7175 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
7176 create a distinct @code{readonly_data_section}, the default is to
7177 reuse @code{text_section}.
7179 All the other @file{varasm.c} sections are optional, and are null
7180 if the target does not provide them.
7182 @defmac TEXT_SECTION_ASM_OP
7183 A C expression whose value is a string, including spacing, containing the
7184 assembler operation that should precede instructions and read-only data.
7185 Normally @code{"\t.text"} is right.
7188 @defmac HOT_TEXT_SECTION_NAME
7189 If defined, a C string constant for the name of the section containing most
7190 frequently executed functions of the program. If not defined, GCC will provide
7191 a default definition if the target supports named sections.
7194 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
7195 If defined, a C string constant for the name of the section containing unlikely
7196 executed functions in the program.
7199 @defmac DATA_SECTION_ASM_OP
7200 A C expression whose value is a string, including spacing, containing the
7201 assembler operation to identify the following data as writable initialized
7202 data. Normally @code{"\t.data"} is right.
7205 @defmac SDATA_SECTION_ASM_OP
7206 If defined, a C expression whose value is a string, including spacing,
7207 containing the assembler operation to identify the following data as
7208 initialized, writable small data.
7211 @defmac READONLY_DATA_SECTION_ASM_OP
7212 A C expression whose value is a string, including spacing, containing the
7213 assembler operation to identify the following data as read-only initialized
7217 @defmac BSS_SECTION_ASM_OP
7218 If defined, a C expression whose value is a string, including spacing,
7219 containing the assembler operation to identify the following data as
7220 uninitialized global data. If not defined, and
7221 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
7222 uninitialized global data will be output in the data section if
7223 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
7227 @defmac SBSS_SECTION_ASM_OP
7228 If defined, a C expression whose value is a string, including spacing,
7229 containing the assembler operation to identify the following data as
7230 uninitialized, writable small data.
7233 @defmac TLS_COMMON_ASM_OP
7234 If defined, a C expression whose value is a string containing the
7235 assembler operation to identify the following data as thread-local
7236 common data. The default is @code{".tls_common"}.
7239 @defmac TLS_SECTION_ASM_FLAG
7240 If defined, a C expression whose value is a character constant
7241 containing the flag used to mark a section as a TLS section. The
7242 default is @code{'T'}.
7245 @defmac INIT_SECTION_ASM_OP
7246 If defined, a C expression whose value is a string, including spacing,
7247 containing the assembler operation to identify the following data as
7248 initialization code. If not defined, GCC will assume such a section does
7249 not exist. This section has no corresponding @code{init_section}
7250 variable; it is used entirely in runtime code.
7253 @defmac FINI_SECTION_ASM_OP
7254 If defined, a C expression whose value is a string, including spacing,
7255 containing the assembler operation to identify the following data as
7256 finalization code. If not defined, GCC will assume such a section does
7257 not exist. This section has no corresponding @code{fini_section}
7258 variable; it is used entirely in runtime code.
7261 @defmac INIT_ARRAY_SECTION_ASM_OP
7262 If defined, a C expression whose value is a string, including spacing,
7263 containing the assembler operation to identify the following data as
7264 part of the @code{.init_array} (or equivalent) section. If not
7265 defined, GCC will assume such a section does not exist. Do not define
7266 both this macro and @code{INIT_SECTION_ASM_OP}.
7269 @defmac FINI_ARRAY_SECTION_ASM_OP
7270 If defined, a C expression whose value is a string, including spacing,
7271 containing the assembler operation to identify the following data as
7272 part of the @code{.fini_array} (or equivalent) section. If not
7273 defined, GCC will assume such a section does not exist. Do not define
7274 both this macro and @code{FINI_SECTION_ASM_OP}.
7277 @defmac MACH_DEP_SECTION_ASM_FLAG
7278 If defined, a C expression whose value is a character constant
7279 containing the flag used to mark a machine-dependent section. This
7280 corresponds to the @code{SECTION_MACH_DEP} section flag.
7283 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
7284 If defined, an ASM statement that switches to a different section
7285 via @var{section_op}, calls @var{function}, and switches back to
7286 the text section. This is used in @file{crtstuff.c} if
7287 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
7288 to initialization and finalization functions from the init and fini
7289 sections. By default, this macro uses a simple function call. Some
7290 ports need hand-crafted assembly code to avoid dependencies on
7291 registers initialized in the function prologue or to ensure that
7292 constant pools don't end up too far way in the text section.
7295 @defmac TARGET_LIBGCC_SDATA_SECTION
7296 If defined, a string which names the section into which small
7297 variables defined in crtstuff and libgcc should go. This is useful
7298 when the target has options for optimizing access to small data, and
7299 you want the crtstuff and libgcc routines to be conservative in what
7300 they expect of your application yet liberal in what your application
7301 expects. For example, for targets with a @code{.sdata} section (like
7302 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
7303 require small data support from your application, but use this macro
7304 to put small data into @code{.sdata} so that your application can
7305 access these variables whether it uses small data or not.
7308 @defmac FORCE_CODE_SECTION_ALIGN
7309 If defined, an ASM statement that aligns a code section to some
7310 arbitrary boundary. This is used to force all fragments of the
7311 @code{.init} and @code{.fini} sections to have to same alignment
7312 and thus prevent the linker from having to add any padding.
7315 @defmac JUMP_TABLES_IN_TEXT_SECTION
7316 Define this macro to be an expression with a nonzero value if jump
7317 tables (for @code{tablejump} insns) should be output in the text
7318 section, along with the assembler instructions. Otherwise, the
7319 readonly data section is used.
7321 This macro is irrelevant if there is no separate readonly data section.
7324 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
7325 Define this hook if you need to do something special to set up the
7326 @file{varasm.c} sections, or if your target has some special sections
7327 of its own that you need to create.
7329 GCC calls this hook after processing the command line, but before writing
7330 any assembly code, and before calling any of the section-returning hooks
7334 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
7335 Return a mask describing how relocations should be treated when
7336 selecting sections. Bit 1 should be set if global relocations
7337 should be placed in a read-write section; bit 0 should be set if
7338 local relocations should be placed in a read-write section.
7340 The default version of this function returns 3 when @option{-fpic}
7341 is in effect, and 0 otherwise. The hook is typically redefined
7342 when the target cannot support (some kinds of) dynamic relocations
7343 in read-only sections even in executables.
7346 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
7347 Return the section into which @var{exp} should be placed. You can
7348 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
7349 some sort. @var{reloc} indicates whether the initial value of @var{exp}
7350 requires link-time relocations. Bit 0 is set when variable contains
7351 local relocations only, while bit 1 is set for global relocations.
7352 @var{align} is the constant alignment in bits.
7354 The default version of this function takes care of putting read-only
7355 variables in @code{readonly_data_section}.
7357 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7360 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7361 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7362 for @code{FUNCTION_DECL}s as well as for variables and constants.
7364 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7365 function has been determined to be likely to be called, and nonzero if
7366 it is unlikely to be called.
7369 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7370 Build up a unique section name, expressed as a @code{STRING_CST} node,
7371 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7372 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7373 the initial value of @var{exp} requires link-time relocations.
7375 The default version of this function appends the symbol name to the
7376 ELF section name that would normally be used for the symbol. For
7377 example, the function @code{foo} would be placed in @code{.text.foo}.
7378 Whatever the actual target object format, this is often good enough.
7381 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7382 Return the readonly data section associated with
7383 @samp{DECL_SECTION_NAME (@var{decl})}.
7384 The default version of this function selects @code{.gnu.linkonce.r.name} if
7385 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7386 if function is in @code{.text.name}, and the normal readonly-data section
7390 @deftypevr {Target Hook} {const char *} TARGET_ASM_MERGEABLE_RODATA_PREFIX
7391 Usually, the compiler uses the prefix @code{".rodata"} to construct
7392 section names for mergeable constant data. Define this macro to override
7393 the string if a different section name should be used.
7396 @deftypefn {Target Hook} {section *} TARGET_ASM_TM_CLONE_TABLE_SECTION (void)
7397 Return the section that should be used for transactional memory clone tables.
7400 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7401 Return the section into which a constant @var{x}, of mode @var{mode},
7402 should be placed. You can assume that @var{x} is some kind of
7403 constant in RTL@. The argument @var{mode} is redundant except in the
7404 case of a @code{const_int} rtx. @var{align} is the constant alignment
7407 The default version of this function takes care of putting symbolic
7408 constants in @code{flag_pic} mode in @code{data_section} and everything
7409 else in @code{readonly_data_section}.
7412 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7413 Define this hook if you need to postprocess the assembler name generated
7414 by target-independent code. The @var{id} provided to this hook will be
7415 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7416 or the mangled name of the @var{decl} in C++). The return value of the
7417 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7418 your target system. The default implementation of this hook just
7419 returns the @var{id} provided.
7422 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7423 Define this hook if references to a symbol or a constant must be
7424 treated differently depending on something about the variable or
7425 function named by the symbol (such as what section it is in).
7427 The hook is executed immediately after rtl has been created for
7428 @var{decl}, which may be a variable or function declaration or
7429 an entry in the constant pool. In either case, @var{rtl} is the
7430 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7431 in this hook; that field may not have been initialized yet.
7433 In the case of a constant, it is safe to assume that the rtl is
7434 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7435 will also have this form, but that is not guaranteed. Global
7436 register variables, for instance, will have a @code{reg} for their
7437 rtl. (Normally the right thing to do with such unusual rtl is
7440 The @var{new_decl_p} argument will be true if this is the first time
7441 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7442 be false for subsequent invocations, which will happen for duplicate
7443 declarations. Whether or not anything must be done for the duplicate
7444 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7445 @var{new_decl_p} is always true when the hook is called for a constant.
7447 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7448 The usual thing for this hook to do is to record flags in the
7449 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7450 Historically, the name string was modified if it was necessary to
7451 encode more than one bit of information, but this practice is now
7452 discouraged; use @code{SYMBOL_REF_FLAGS}.
7454 The default definition of this hook, @code{default_encode_section_info}
7455 in @file{varasm.c}, sets a number of commonly-useful bits in
7456 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7457 before overriding it.
7460 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7461 Decode @var{name} and return the real name part, sans
7462 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7466 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7467 Returns true if @var{exp} should be placed into a ``small data'' section.
7468 The default version of this hook always returns false.
7471 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7472 Contains the value true if the target places read-only
7473 ``small data'' into a separate section. The default value is false.
7476 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7477 It returns true if target wants profile code emitted before prologue.
7479 The default version of this hook use the target macro
7480 @code{PROFILE_BEFORE_PROLOGUE}.
7483 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7484 Returns true if @var{exp} names an object for which name resolution
7485 rules must resolve to the current ``module'' (dynamic shared library
7486 or executable image).
7488 The default version of this hook implements the name resolution rules
7489 for ELF, which has a looser model of global name binding than other
7490 currently supported object file formats.
7493 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7494 Contains the value true if the target supports thread-local storage.
7495 The default value is false.
7500 @section Position Independent Code
7501 @cindex position independent code
7504 This section describes macros that help implement generation of position
7505 independent code. Simply defining these macros is not enough to
7506 generate valid PIC; you must also add support to the hook
7507 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7508 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7509 must modify the definition of @samp{movsi} to do something appropriate
7510 when the source operand contains a symbolic address. You may also
7511 need to alter the handling of switch statements so that they use
7513 @c i rearranged the order of the macros above to try to force one of
7514 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7516 @defmac PIC_OFFSET_TABLE_REGNUM
7517 The register number of the register used to address a table of static
7518 data addresses in memory. In some cases this register is defined by a
7519 processor's ``application binary interface'' (ABI)@. When this macro
7520 is defined, RTL is generated for this register once, as with the stack
7521 pointer and frame pointer registers. If this macro is not defined, it
7522 is up to the machine-dependent files to allocate such a register (if
7523 necessary). Note that this register must be fixed when in use (e.g.@:
7524 when @code{flag_pic} is true).
7527 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7528 A C expression that is nonzero if the register defined by
7529 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7530 the default is zero. Do not define
7531 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7534 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7535 A C expression that is nonzero if @var{x} is a legitimate immediate
7536 operand on the target machine when generating position independent code.
7537 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7538 check this. You can also assume @var{flag_pic} is true, so you need not
7539 check it either. You need not define this macro if all constants
7540 (including @code{SYMBOL_REF}) can be immediate operands when generating
7541 position independent code.
7544 @node Assembler Format
7545 @section Defining the Output Assembler Language
7547 This section describes macros whose principal purpose is to describe how
7548 to write instructions in assembler language---rather than what the
7552 * File Framework:: Structural information for the assembler file.
7553 * Data Output:: Output of constants (numbers, strings, addresses).
7554 * Uninitialized Data:: Output of uninitialized variables.
7555 * Label Output:: Output and generation of labels.
7556 * Initialization:: General principles of initialization
7557 and termination routines.
7558 * Macros for Initialization::
7559 Specific macros that control the handling of
7560 initialization and termination routines.
7561 * Instruction Output:: Output of actual instructions.
7562 * Dispatch Tables:: Output of jump tables.
7563 * Exception Region Output:: Output of exception region code.
7564 * Alignment Output:: Pseudo ops for alignment and skipping data.
7567 @node File Framework
7568 @subsection The Overall Framework of an Assembler File
7569 @cindex assembler format
7570 @cindex output of assembler code
7572 @c prevent bad page break with this line
7573 This describes the overall framework of an assembly file.
7575 @findex default_file_start
7576 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7577 Output to @code{asm_out_file} any text which the assembler expects to
7578 find at the beginning of a file. The default behavior is controlled
7579 by two flags, documented below. Unless your target's assembler is
7580 quite unusual, if you override the default, you should call
7581 @code{default_file_start} at some point in your target hook. This
7582 lets other target files rely on these variables.
7585 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7586 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7587 printed as the very first line in the assembly file, unless
7588 @option{-fverbose-asm} is in effect. (If that macro has been defined
7589 to the empty string, this variable has no effect.) With the normal
7590 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7591 assembler that it need not bother stripping comments or extra
7592 whitespace from its input. This allows it to work a bit faster.
7594 The default is false. You should not set it to true unless you have
7595 verified that your port does not generate any extra whitespace or
7596 comments that will cause GAS to issue errors in NO_APP mode.
7599 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7600 If this flag is true, @code{output_file_directive} will be called
7601 for the primary source file, immediately after printing
7602 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7603 this to be done. The default is false.
7606 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7607 Output to @code{asm_out_file} any text which the assembler expects
7608 to find at the end of a file. The default is to output nothing.
7611 @deftypefun void file_end_indicate_exec_stack ()
7612 Some systems use a common convention, the @samp{.note.GNU-stack}
7613 special section, to indicate whether or not an object file relies on
7614 the stack being executable. If your system uses this convention, you
7615 should define @code{TARGET_ASM_FILE_END} to this function. If you
7616 need to do other things in that hook, have your hook function call
7620 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7621 Output to @code{asm_out_file} any text which the assembler expects
7622 to find at the start of an LTO section. The default is to output
7626 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7627 Output to @code{asm_out_file} any text which the assembler expects
7628 to find at the end of an LTO section. The default is to output
7632 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7633 Output to @code{asm_out_file} any text which is needed before emitting
7634 unwind info and debug info at the end of a file. Some targets emit
7635 here PIC setup thunks that cannot be emitted at the end of file,
7636 because they couldn't have unwind info then. The default is to output
7640 @defmac ASM_COMMENT_START
7641 A C string constant describing how to begin a comment in the target
7642 assembler language. The compiler assumes that the comment will end at
7643 the end of the line.
7647 A C string constant for text to be output before each @code{asm}
7648 statement or group of consecutive ones. Normally this is
7649 @code{"#APP"}, which is a comment that has no effect on most
7650 assemblers but tells the GNU assembler that it must check the lines
7651 that follow for all valid assembler constructs.
7655 A C string constant for text to be output after each @code{asm}
7656 statement or group of consecutive ones. Normally this is
7657 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7658 time-saving assumptions that are valid for ordinary compiler output.
7661 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7662 A C statement to output COFF information or DWARF debugging information
7663 which indicates that filename @var{name} is the current source file to
7664 the stdio stream @var{stream}.
7666 This macro need not be defined if the standard form of output
7667 for the file format in use is appropriate.
7670 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7671 Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7673 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7676 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_IDENT (const char *@var{name})
7677 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.
7680 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7681 A C statement to output the string @var{string} to the stdio stream
7682 @var{stream}. If you do not call the function @code{output_quoted_string}
7683 in your config files, GCC will only call it to output filenames to
7684 the assembler source. So you can use it to canonicalize the format
7685 of the filename using this macro.
7688 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7689 Output assembly directives to switch to section @var{name}. The section
7690 should have attributes as specified by @var{flags}, which is a bit mask
7691 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7692 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7693 this section is associated.
7696 @deftypefn {Target Hook} bool TARGET_ASM_ELF_FLAGS_NUMERIC (unsigned int @var{flags}, unsigned int *@var{num})
7697 This hook can be used to encode ELF section flags for which no letter
7698 code has been defined in the assembler. It is called by
7699 @code{default_asm_named_section} whenever the section flags need to be
7700 emitted in the assembler output. If the hook returns true, then the
7701 numerical value for ELF section flags should be calculated from
7702 @var{flags} and saved in @var{*num}; the value is printed out instead of the
7703 normal sequence of letter codes. If the hook is not defined, or if it
7704 returns false, then @var{num} is ignored and the traditional letter sequence
7708 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
7709 Return preferred text (sub)section for function @var{decl}.
7710 Main purpose of this function is to separate cold, normal and hot
7711 functions. @var{startup} is true when function is known to be used only
7712 at startup (from static constructors or it is @code{main()}).
7713 @var{exit} is true when function is known to be used only at exit
7714 (from static destructors).
7715 Return NULL if function should go to default text section.
7718 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
7719 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}.
7722 @deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7723 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7724 It must not be modified by command-line option processing.
7727 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7728 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7729 This flag is true if we can create zeroed data by switching to a BSS
7730 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7731 This is true on most ELF targets.
7734 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7735 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7736 based on a variable or function decl, a section name, and whether or not the
7737 declaration's initializer may contain runtime relocations. @var{decl} may be
7738 null, in which case read-write data should be assumed.
7740 The default version of this function handles choosing code vs data,
7741 read-only vs read-write data, and @code{flag_pic}. You should only
7742 need to override this if your target has special flags that might be
7743 set via @code{__attribute__}.
7746 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
7747 Provides the target with the ability to record the gcc command line
7748 switches that have been passed to the compiler, and options that are
7749 enabled. The @var{type} argument specifies what is being recorded.
7750 It can take the following values:
7753 @item SWITCH_TYPE_PASSED
7754 @var{text} is a command line switch that has been set by the user.
7756 @item SWITCH_TYPE_ENABLED
7757 @var{text} is an option which has been enabled. This might be as a
7758 direct result of a command line switch, or because it is enabled by
7759 default or because it has been enabled as a side effect of a different
7760 command line switch. For example, the @option{-O2} switch enables
7761 various different individual optimization passes.
7763 @item SWITCH_TYPE_DESCRIPTIVE
7764 @var{text} is either NULL or some descriptive text which should be
7765 ignored. If @var{text} is NULL then it is being used to warn the
7766 target hook that either recording is starting or ending. The first
7767 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7768 warning is for start up and the second time the warning is for
7769 wind down. This feature is to allow the target hook to make any
7770 necessary preparations before it starts to record switches and to
7771 perform any necessary tidying up after it has finished recording
7774 @item SWITCH_TYPE_LINE_START
7775 This option can be ignored by this target hook.
7777 @item SWITCH_TYPE_LINE_END
7778 This option can be ignored by this target hook.
7781 The hook's return value must be zero. Other return values may be
7782 supported in the future.
7784 By default this hook is set to NULL, but an example implementation is
7785 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7786 it records the switches as ASCII text inside a new, string mergeable
7787 section in the assembler output file. The name of the new section is
7788 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7792 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7793 This is the name of the section that will be created by the example
7794 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7800 @subsection Output of Data
7803 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7804 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7805 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7806 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7807 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7808 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7809 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7810 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7811 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7812 These hooks specify assembly directives for creating certain kinds
7813 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7814 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7815 aligned two-byte object, and so on. Any of the hooks may be
7816 @code{NULL}, indicating that no suitable directive is available.
7818 The compiler will print these strings at the start of a new line,
7819 followed immediately by the object's initial value. In most cases,
7820 the string should contain a tab, a pseudo-op, and then another tab.
7823 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7824 The @code{assemble_integer} function uses this hook to output an
7825 integer object. @var{x} is the object's value, @var{size} is its size
7826 in bytes and @var{aligned_p} indicates whether it is aligned. The
7827 function should return @code{true} if it was able to output the
7828 object. If it returns false, @code{assemble_integer} will try to
7829 split the object into smaller parts.
7831 The default implementation of this hook will use the
7832 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7833 when the relevant string is @code{NULL}.
7836 @deftypefn {Target Hook} void TARGET_ASM_DECL_END (void)
7837 Define this hook if the target assembler requires a special marker to
7838 terminate an initialized variable declaration.
7841 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
7842 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7843 can't deal with, and output assembly code to @var{file} corresponding to
7844 the pattern @var{x}. This may be used to allow machine-dependent
7845 @code{UNSPEC}s to appear within constants.
7847 If target hook fails to recognize a pattern, it must return @code{false},
7848 so that a standard error message is printed. If it prints an error message
7849 itself, by calling, for example, @code{output_operand_lossage}, it may just
7853 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7854 A C statement to output to the stdio stream @var{stream} an assembler
7855 instruction to assemble a string constant containing the @var{len}
7856 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7857 @code{char *} and @var{len} a C expression of type @code{int}.
7859 If the assembler has a @code{.ascii} pseudo-op as found in the
7860 Berkeley Unix assembler, do not define the macro
7861 @code{ASM_OUTPUT_ASCII}.
7864 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7865 A C statement to output word @var{n} of a function descriptor for
7866 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7867 is defined, and is otherwise unused.
7870 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7871 You may define this macro as a C expression. You should define the
7872 expression to have a nonzero value if GCC should output the constant
7873 pool for a function before the code for the function, or a zero value if
7874 GCC should output the constant pool after the function. If you do
7875 not define this macro, the usual case, GCC will output the constant
7876 pool before the function.
7879 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7880 A C statement to output assembler commands to define the start of the
7881 constant pool for a function. @var{funname} is a string giving
7882 the name of the function. Should the return type of the function
7883 be required, it can be obtained via @var{fundecl}. @var{size}
7884 is the size, in bytes, of the constant pool that will be written
7885 immediately after this call.
7887 If no constant-pool prefix is required, the usual case, this macro need
7891 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7892 A C statement (with or without semicolon) to output a constant in the
7893 constant pool, if it needs special treatment. (This macro need not do
7894 anything for RTL expressions that can be output normally.)
7896 The argument @var{file} is the standard I/O stream to output the
7897 assembler code on. @var{x} is the RTL expression for the constant to
7898 output, and @var{mode} is the machine mode (in case @var{x} is a
7899 @samp{const_int}). @var{align} is the required alignment for the value
7900 @var{x}; you should output an assembler directive to force this much
7903 The argument @var{labelno} is a number to use in an internal label for
7904 the address of this pool entry. The definition of this macro is
7905 responsible for outputting the label definition at the proper place.
7906 Here is how to do this:
7909 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7912 When you output a pool entry specially, you should end with a
7913 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7914 entry from being output a second time in the usual manner.
7916 You need not define this macro if it would do nothing.
7919 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7920 A C statement to output assembler commands to at the end of the constant
7921 pool for a function. @var{funname} is a string giving the name of the
7922 function. Should the return type of the function be required, you can
7923 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7924 constant pool that GCC wrote immediately before this call.
7926 If no constant-pool epilogue is required, the usual case, you need not
7930 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7931 Define this macro as a C expression which is nonzero if @var{C} is
7932 used as a logical line separator by the assembler. @var{STR} points
7933 to the position in the string where @var{C} was found; this can be used if
7934 a line separator uses multiple characters.
7936 If you do not define this macro, the default is that only
7937 the character @samp{;} is treated as a logical line separator.
7940 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7941 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7942 These target hooks are C string constants, describing the syntax in the
7943 assembler for grouping arithmetic expressions. If not overridden, they
7944 default to normal parentheses, which is correct for most assemblers.
7947 These macros are provided by @file{real.h} for writing the definitions
7948 of @code{ASM_OUTPUT_DOUBLE} and the like:
7950 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7951 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7952 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7953 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7954 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7955 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7956 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7957 target's floating point representation, and store its bit pattern in
7958 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7959 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7960 simple @code{long int}. For the others, it should be an array of
7961 @code{long int}. The number of elements in this array is determined
7962 by the size of the desired target floating point data type: 32 bits of
7963 it go in each @code{long int} array element. Each array element holds
7964 32 bits of the result, even if @code{long int} is wider than 32 bits
7965 on the host machine.
7967 The array element values are designed so that you can print them out
7968 using @code{fprintf} in the order they should appear in the target
7972 @node Uninitialized Data
7973 @subsection Output of Uninitialized Variables
7975 Each of the macros in this section is used to do the whole job of
7976 outputting a single uninitialized variable.
7978 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7979 A C statement (sans semicolon) to output to the stdio stream
7980 @var{stream} the assembler definition of a common-label named
7981 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7982 is the size rounded up to whatever alignment the caller wants. It is
7983 possible that @var{size} may be zero, for instance if a struct with no
7984 other member than a zero-length array is defined. In this case, the
7985 backend must output a symbol definition that allocates at least one
7986 byte, both so that the address of the resulting object does not compare
7987 equal to any other, and because some object formats cannot even express
7988 the concept of a zero-sized common symbol, as that is how they represent
7989 an ordinary undefined external.
7991 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7992 output the name itself; before and after that, output the additional
7993 assembler syntax for defining the name, and a newline.
7995 This macro controls how the assembler definitions of uninitialized
7996 common global variables are output.
7999 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
8000 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
8001 separate, explicit argument. If you define this macro, it is used in
8002 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
8003 handling the required alignment of the variable. The alignment is specified
8004 as the number of bits.
8007 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
8008 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
8009 variable to be output, if there is one, or @code{NULL_TREE} if there
8010 is no corresponding variable. If you define this macro, GCC will use it
8011 in place of both @code{ASM_OUTPUT_COMMON} and
8012 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
8013 the variable's decl in order to chose what to output.
8016 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
8017 A C statement (sans semicolon) to output to the stdio stream
8018 @var{stream} the assembler definition of uninitialized global @var{decl} named
8019 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
8020 is the alignment specified as the number of bits.
8022 Try to use function @code{asm_output_aligned_bss} defined in file
8023 @file{varasm.c} when defining this macro. If unable, use the expression
8024 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
8025 before and after that, output the additional assembler syntax for defining
8026 the name, and a newline.
8028 There are two ways of handling global BSS@. One is to define this macro.
8029 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
8030 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
8031 You do not need to do both.
8033 Some languages do not have @code{common} data, and require a
8034 non-common form of global BSS in order to handle uninitialized globals
8035 efficiently. C++ is one example of this. However, if the target does
8036 not support global BSS, the front end may choose to make globals
8037 common in order to save space in the object file.
8040 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
8041 A C statement (sans semicolon) to output to the stdio stream
8042 @var{stream} the assembler definition of a local-common-label named
8043 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
8044 is the size rounded up to whatever alignment the caller wants.
8046 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8047 output the name itself; before and after that, output the additional
8048 assembler syntax for defining the name, and a newline.
8050 This macro controls how the assembler definitions of uninitialized
8051 static variables are output.
8054 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
8055 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
8056 separate, explicit argument. If you define this macro, it is used in
8057 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
8058 handling the required alignment of the variable. The alignment is specified
8059 as the number of bits.
8062 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
8063 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
8064 variable to be output, if there is one, or @code{NULL_TREE} if there
8065 is no corresponding variable. If you define this macro, GCC will use it
8066 in place of both @code{ASM_OUTPUT_DECL} and
8067 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
8068 the variable's decl in order to chose what to output.
8072 @subsection Output and Generation of Labels
8074 @c prevent bad page break with this line
8075 This is about outputting labels.
8077 @findex assemble_name
8078 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
8079 A C statement (sans semicolon) to output to the stdio stream
8080 @var{stream} the assembler definition of a label named @var{name}.
8081 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8082 output the name itself; before and after that, output the additional
8083 assembler syntax for defining the name, and a newline. A default
8084 definition of this macro is provided which is correct for most systems.
8087 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
8088 A C statement (sans semicolon) to output to the stdio stream
8089 @var{stream} the assembler definition of a label named @var{name} of
8091 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8092 output the name itself; before and after that, output the additional
8093 assembler syntax for defining the name, and a newline. A default
8094 definition of this macro is provided which is correct for most systems.
8096 If this macro is not defined, then the function name is defined in the
8097 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8100 @findex assemble_name_raw
8101 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
8102 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
8103 to refer to a compiler-generated label. The default definition uses
8104 @code{assemble_name_raw}, which is like @code{assemble_name} except
8105 that it is more efficient.
8109 A C string containing the appropriate assembler directive to specify the
8110 size of a symbol, without any arguments. On systems that use ELF, the
8111 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
8112 systems, the default is not to define this macro.
8114 Define this macro only if it is correct to use the default definitions
8115 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
8116 for your system. If you need your own custom definitions of those
8117 macros, or if you do not need explicit symbol sizes at all, do not
8121 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
8122 A C statement (sans semicolon) to output to the stdio stream
8123 @var{stream} a directive telling the assembler that the size of the
8124 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
8125 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
8129 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
8130 A C statement (sans semicolon) to output to the stdio stream
8131 @var{stream} a directive telling the assembler to calculate the size of
8132 the symbol @var{name} by subtracting its address from the current
8135 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
8136 provided. The default assumes that the assembler recognizes a special
8137 @samp{.} symbol as referring to the current address, and can calculate
8138 the difference between this and another symbol. If your assembler does
8139 not recognize @samp{.} or cannot do calculations with it, you will need
8140 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
8143 @defmac NO_DOLLAR_IN_LABEL
8144 Define this macro if the assembler does not accept the character
8145 @samp{$} in label names. By default constructors and destructors in
8146 G++ have @samp{$} in the identifiers. If this macro is defined,
8147 @samp{.} is used instead.
8150 @defmac NO_DOT_IN_LABEL
8151 Define this macro if the assembler does not accept the character
8152 @samp{.} in label names. By default constructors and destructors in G++
8153 have names that use @samp{.}. If this macro is defined, these names
8154 are rewritten to avoid @samp{.}.
8158 A C string containing the appropriate assembler directive to specify the
8159 type of a symbol, without any arguments. On systems that use ELF, the
8160 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
8161 systems, the default is not to define this macro.
8163 Define this macro only if it is correct to use the default definition of
8164 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
8165 custom definition of this macro, or if you do not need explicit symbol
8166 types at all, do not define this macro.
8169 @defmac TYPE_OPERAND_FMT
8170 A C string which specifies (using @code{printf} syntax) the format of
8171 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
8172 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
8173 the default is not to define this macro.
8175 Define this macro only if it is correct to use the default definition of
8176 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
8177 custom definition of this macro, or if you do not need explicit symbol
8178 types at all, do not define this macro.
8181 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
8182 A C statement (sans semicolon) to output to the stdio stream
8183 @var{stream} a directive telling the assembler that the type of the
8184 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
8185 that string is always either @samp{"function"} or @samp{"object"}, but
8186 you should not count on this.
8188 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
8189 definition of this macro is provided.
8192 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
8193 A C statement (sans semicolon) to output to the stdio stream
8194 @var{stream} any text necessary for declaring the name @var{name} of a
8195 function which is being defined. This macro is responsible for
8196 outputting the label definition (perhaps using
8197 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
8198 @code{FUNCTION_DECL} tree node representing the function.
8200 If this macro is not defined, then the function name is defined in the
8201 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
8203 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
8207 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
8208 A C statement (sans semicolon) to output to the stdio stream
8209 @var{stream} any text necessary for declaring the size of a function
8210 which is being defined. The argument @var{name} is the name of the
8211 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
8212 representing the function.
8214 If this macro is not defined, then the function size is not defined.
8216 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
8220 @defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
8221 A C statement (sans semicolon) to output to the stdio stream
8222 @var{stream} any text necessary for declaring the name @var{name} of a
8223 cold function partition which is being defined. This macro is responsible
8224 for outputting the label definition (perhaps using
8225 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
8226 @code{FUNCTION_DECL} tree node representing the function.
8228 If this macro is not defined, then the cold partition name is defined in the
8229 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8231 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
8235 @defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
8236 A C statement (sans semicolon) to output to the stdio stream
8237 @var{stream} any text necessary for declaring the size of a cold function
8238 partition which is being defined. The argument @var{name} is the name of the
8239 cold partition of the function. The argument @var{decl} is the
8240 @code{FUNCTION_DECL} tree node representing the function.
8242 If this macro is not defined, then the partition size is not defined.
8244 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
8248 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
8249 A C statement (sans semicolon) to output to the stdio stream
8250 @var{stream} any text necessary for declaring the name @var{name} of an
8251 initialized variable which is being defined. This macro must output the
8252 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
8253 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
8255 If this macro is not defined, then the variable name is defined in the
8256 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8258 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
8259 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
8262 @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})
8263 A target hook to output to the stdio stream @var{file} any text necessary
8264 for declaring the name @var{name} of a constant which is being defined. This
8265 target hook is responsible for outputting the label definition (perhaps using
8266 @code{assemble_label}). The argument @var{exp} is the value of the constant,
8267 and @var{size} is the size of the constant in bytes. The @var{name}
8268 will be an internal label.
8270 The default version of this target hook, define the @var{name} in the
8271 usual manner as a label (by means of @code{assemble_label}).
8273 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
8276 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
8277 A C statement (sans semicolon) to output to the stdio stream
8278 @var{stream} any text necessary for claiming a register @var{regno}
8279 for a global variable @var{decl} with name @var{name}.
8281 If you don't define this macro, that is equivalent to defining it to do
8285 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
8286 A C statement (sans semicolon) to finish up declaring a variable name
8287 once the compiler has processed its initializer fully and thus has had a
8288 chance to determine the size of an array when controlled by an
8289 initializer. This is used on systems where it's necessary to declare
8290 something about the size of the object.
8292 If you don't define this macro, that is equivalent to defining it to do
8295 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
8296 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
8299 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
8300 This target hook is a function to output to the stdio stream
8301 @var{stream} some commands that will make the label @var{name} global;
8302 that is, available for reference from other files.
8304 The default implementation relies on a proper definition of
8305 @code{GLOBAL_ASM_OP}.
8308 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
8309 This target hook is a function to output to the stdio stream
8310 @var{stream} some commands that will make the name associated with @var{decl}
8311 global; that is, available for reference from other files.
8313 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
8316 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_UNDEFINED_DECL (FILE *@var{stream}, const char *@var{name}, const_tree @var{decl})
8317 This target hook is a function to output to the stdio stream
8318 @var{stream} some commands that will declare the name associated with
8319 @var{decl} which is not defined in the current translation unit. Most
8320 assemblers do not require anything to be output in this case.
8323 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
8324 A C statement (sans semicolon) to output to the stdio stream
8325 @var{stream} some commands that will make the label @var{name} weak;
8326 that is, available for reference from other files but only used if
8327 no other definition is available. Use the expression
8328 @code{assemble_name (@var{stream}, @var{name})} to output the name
8329 itself; before and after that, output the additional assembler syntax
8330 for making that name weak, and a newline.
8332 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
8333 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
8337 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
8338 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
8339 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
8340 or variable decl. If @var{value} is not @code{NULL}, this C statement
8341 should output to the stdio stream @var{stream} assembler code which
8342 defines (equates) the weak symbol @var{name} to have the value
8343 @var{value}. If @var{value} is @code{NULL}, it should output commands
8344 to make @var{name} weak.
8347 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
8348 Outputs a directive that enables @var{name} to be used to refer to
8349 symbol @var{value} with weak-symbol semantics. @code{decl} is the
8350 declaration of @code{name}.
8353 @defmac SUPPORTS_WEAK
8354 A preprocessor constant expression which evaluates to true if the target
8355 supports weak symbols.
8357 If you don't define this macro, @file{defaults.h} provides a default
8358 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
8359 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
8362 @defmac TARGET_SUPPORTS_WEAK
8363 A C expression which evaluates to true if the target supports weak symbols.
8365 If you don't define this macro, @file{defaults.h} provides a default
8366 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
8367 this macro if you want to control weak symbol support with a compiler
8368 flag such as @option{-melf}.
8371 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
8372 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
8373 public symbol such that extra copies in multiple translation units will
8374 be discarded by the linker. Define this macro if your object file
8375 format provides support for this concept, such as the @samp{COMDAT}
8376 section flags in the Microsoft Windows PE/COFF format, and this support
8377 requires changes to @var{decl}, such as putting it in a separate section.
8380 @defmac SUPPORTS_ONE_ONLY
8381 A C expression which evaluates to true if the target supports one-only
8384 If you don't define this macro, @file{varasm.c} provides a default
8385 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
8386 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
8387 you want to control one-only symbol support with a compiler flag, or if
8388 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
8389 be emitted as one-only.
8392 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
8393 This target hook is a function to output to @var{asm_out_file} some
8394 commands that will make the symbol(s) associated with @var{decl} have
8395 hidden, protected or internal visibility as specified by @var{visibility}.
8398 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
8399 A C expression that evaluates to true if the target's linker expects
8400 that weak symbols do not appear in a static archive's table of contents.
8401 The default is @code{0}.
8403 Leaving weak symbols out of an archive's table of contents means that,
8404 if a symbol will only have a definition in one translation unit and
8405 will have undefined references from other translation units, that
8406 symbol should not be weak. Defining this macro to be nonzero will
8407 thus have the effect that certain symbols that would normally be weak
8408 (explicit template instantiations, and vtables for polymorphic classes
8409 with noninline key methods) will instead be nonweak.
8411 The C++ ABI requires this macro to be zero. Define this macro for
8412 targets where full C++ ABI compliance is impossible and where linker
8413 restrictions require weak symbols to be left out of a static archive's
8417 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8418 A C statement (sans semicolon) to output to the stdio stream
8419 @var{stream} any text necessary for declaring the name of an external
8420 symbol named @var{name} which is referenced in this compilation but
8421 not defined. The value of @var{decl} is the tree node for the
8424 This macro need not be defined if it does not need to output anything.
8425 The GNU assembler and most Unix assemblers don't require anything.
8428 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8429 This target hook is a function to output to @var{asm_out_file} an assembler
8430 pseudo-op to declare a library function name external. The name of the
8431 library function is given by @var{symref}, which is a @code{symbol_ref}.
8434 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8435 This target hook is a function to output to @var{asm_out_file} an assembler
8436 directive to annotate @var{symbol} as used. The Darwin target uses the
8437 .no_dead_code_strip directive.
8440 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8441 A C statement (sans semicolon) to output to the stdio stream
8442 @var{stream} a reference in assembler syntax to a label named
8443 @var{name}. This should add @samp{_} to the front of the name, if that
8444 is customary on your operating system, as it is in most Berkeley Unix
8445 systems. This macro is used in @code{assemble_name}.
8448 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8449 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.
8452 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8453 A C statement (sans semicolon) to output a reference to
8454 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8455 will be used to output the name of the symbol. This macro may be used
8456 to modify the way a symbol is referenced depending on information
8457 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8460 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8461 A C statement (sans semicolon) to output a reference to @var{buf}, the
8462 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8463 @code{assemble_name} will be used to output the name of the symbol.
8464 This macro is not used by @code{output_asm_label}, or the @code{%l}
8465 specifier that calls it; the intention is that this macro should be set
8466 when it is necessary to output a label differently when its address is
8470 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8471 A function to output to the stdio stream @var{stream} a label whose
8472 name is made from the string @var{prefix} and the number @var{labelno}.
8474 It is absolutely essential that these labels be distinct from the labels
8475 used for user-level functions and variables. Otherwise, certain programs
8476 will have name conflicts with internal labels.
8478 It is desirable to exclude internal labels from the symbol table of the
8479 object file. Most assemblers have a naming convention for labels that
8480 should be excluded; on many systems, the letter @samp{L} at the
8481 beginning of a label has this effect. You should find out what
8482 convention your system uses, and follow it.
8484 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8487 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8488 A C statement to output to the stdio stream @var{stream} a debug info
8489 label whose name is made from the string @var{prefix} and the number
8490 @var{num}. This is useful for VLIW targets, where debug info labels
8491 may need to be treated differently than branch target labels. On some
8492 systems, branch target labels must be at the beginning of instruction
8493 bundles, but debug info labels can occur in the middle of instruction
8496 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8500 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8501 A C statement to store into the string @var{string} a label whose name
8502 is made from the string @var{prefix} and the number @var{num}.
8504 This string, when output subsequently by @code{assemble_name}, should
8505 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8506 with the same @var{prefix} and @var{num}.
8508 If the string begins with @samp{*}, then @code{assemble_name} will
8509 output the rest of the string unchanged. It is often convenient for
8510 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8511 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8512 to output the string, and may change it. (Of course,
8513 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8514 you should know what it does on your machine.)
8517 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8518 A C expression to assign to @var{outvar} (which is a variable of type
8519 @code{char *}) a newly allocated string made from the string
8520 @var{name} and the number @var{number}, with some suitable punctuation
8521 added. Use @code{alloca} to get space for the string.
8523 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8524 produce an assembler label for an internal static variable whose name is
8525 @var{name}. Therefore, the string must be such as to result in valid
8526 assembler code. The argument @var{number} is different each time this
8527 macro is executed; it prevents conflicts between similarly-named
8528 internal static variables in different scopes.
8530 Ideally this string should not be a valid C identifier, to prevent any
8531 conflict with the user's own symbols. Most assemblers allow periods
8532 or percent signs in assembler symbols; putting at least one of these
8533 between the name and the number will suffice.
8535 If this macro is not defined, a default definition will be provided
8536 which is correct for most systems.
8539 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8540 A C statement to output to the stdio stream @var{stream} assembler code
8541 which defines (equates) the symbol @var{name} to have the value @var{value}.
8544 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8545 correct for most systems.
8548 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8549 A C statement to output to the stdio stream @var{stream} assembler code
8550 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8551 to have the value of the tree node @var{decl_of_value}. This macro will
8552 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8553 the tree nodes are available.
8556 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8557 correct for most systems.
8560 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8561 A C statement that evaluates to true if the assembler code which defines
8562 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8563 of the tree node @var{decl_of_value} should be emitted near the end of the
8564 current compilation unit. The default is to not defer output of defines.
8565 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8566 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8569 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8570 A C statement to output to the stdio stream @var{stream} assembler code
8571 which defines (equates) the weak symbol @var{name} to have the value
8572 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8573 an undefined weak symbol.
8575 Define this macro if the target only supports weak aliases; define
8576 @code{ASM_OUTPUT_DEF} instead if possible.
8579 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8580 Define this macro to override the default assembler names used for
8581 Objective-C methods.
8583 The default name is a unique method number followed by the name of the
8584 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8585 the category is also included in the assembler name (e.g.@:
8588 These names are safe on most systems, but make debugging difficult since
8589 the method's selector is not present in the name. Therefore, particular
8590 systems define other ways of computing names.
8592 @var{buf} is an expression of type @code{char *} which gives you a
8593 buffer in which to store the name; its length is as long as
8594 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8595 50 characters extra.
8597 The argument @var{is_inst} specifies whether the method is an instance
8598 method or a class method; @var{class_name} is the name of the class;
8599 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8600 in a category); and @var{sel_name} is the name of the selector.
8602 On systems where the assembler can handle quoted names, you can use this
8603 macro to provide more human-readable names.
8606 @node Initialization
8607 @subsection How Initialization Functions Are Handled
8608 @cindex initialization routines
8609 @cindex termination routines
8610 @cindex constructors, output of
8611 @cindex destructors, output of
8613 The compiled code for certain languages includes @dfn{constructors}
8614 (also called @dfn{initialization routines})---functions to initialize
8615 data in the program when the program is started. These functions need
8616 to be called before the program is ``started''---that is to say, before
8617 @code{main} is called.
8619 Compiling some languages generates @dfn{destructors} (also called
8620 @dfn{termination routines}) that should be called when the program
8623 To make the initialization and termination functions work, the compiler
8624 must output something in the assembler code to cause those functions to
8625 be called at the appropriate time. When you port the compiler to a new
8626 system, you need to specify how to do this.
8628 There are two major ways that GCC currently supports the execution of
8629 initialization and termination functions. Each way has two variants.
8630 Much of the structure is common to all four variations.
8632 @findex __CTOR_LIST__
8633 @findex __DTOR_LIST__
8634 The linker must build two lists of these functions---a list of
8635 initialization functions, called @code{__CTOR_LIST__}, and a list of
8636 termination functions, called @code{__DTOR_LIST__}.
8638 Each list always begins with an ignored function pointer (which may hold
8639 0, @minus{}1, or a count of the function pointers after it, depending on
8640 the environment). This is followed by a series of zero or more function
8641 pointers to constructors (or destructors), followed by a function
8642 pointer containing zero.
8644 Depending on the operating system and its executable file format, either
8645 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8646 time and exit time. Constructors are called in reverse order of the
8647 list; destructors in forward order.
8649 The best way to handle static constructors works only for object file
8650 formats which provide arbitrarily-named sections. A section is set
8651 aside for a list of constructors, and another for a list of destructors.
8652 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8653 object file that defines an initialization function also puts a word in
8654 the constructor section to point to that function. The linker
8655 accumulates all these words into one contiguous @samp{.ctors} section.
8656 Termination functions are handled similarly.
8658 This method will be chosen as the default by @file{target-def.h} if
8659 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8660 support arbitrary sections, but does support special designated
8661 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8662 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8664 When arbitrary sections are available, there are two variants, depending
8665 upon how the code in @file{crtstuff.c} is called. On systems that
8666 support a @dfn{.init} section which is executed at program startup,
8667 parts of @file{crtstuff.c} are compiled into that section. The
8668 program is linked by the @command{gcc} driver like this:
8671 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8674 The prologue of a function (@code{__init}) appears in the @code{.init}
8675 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8676 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8677 files are provided by the operating system or by the GNU C library, but
8678 are provided by GCC for a few targets.
8680 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8681 compiled from @file{crtstuff.c}. They contain, among other things, code
8682 fragments within the @code{.init} and @code{.fini} sections that branch
8683 to routines in the @code{.text} section. The linker will pull all parts
8684 of a section together, which results in a complete @code{__init} function
8685 that invokes the routines we need at startup.
8687 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8690 If no init section is available, when GCC compiles any function called
8691 @code{main} (or more accurately, any function designated as a program
8692 entry point by the language front end calling @code{expand_main_function}),
8693 it inserts a procedure call to @code{__main} as the first executable code
8694 after the function prologue. The @code{__main} function is defined
8695 in @file{libgcc2.c} and runs the global constructors.
8697 In file formats that don't support arbitrary sections, there are again
8698 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8699 and an `a.out' format must be used. In this case,
8700 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8701 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8702 and with the address of the void function containing the initialization
8703 code as its value. The GNU linker recognizes this as a request to add
8704 the value to a @dfn{set}; the values are accumulated, and are eventually
8705 placed in the executable as a vector in the format described above, with
8706 a leading (ignored) count and a trailing zero element.
8707 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8708 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8709 the compilation of @code{main} to call @code{__main} as above, starting
8710 the initialization process.
8712 The last variant uses neither arbitrary sections nor the GNU linker.
8713 This is preferable when you want to do dynamic linking and when using
8714 file formats which the GNU linker does not support, such as `ECOFF'@. In
8715 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8716 termination functions are recognized simply by their names. This requires
8717 an extra program in the linkage step, called @command{collect2}. This program
8718 pretends to be the linker, for use with GCC; it does its job by running
8719 the ordinary linker, but also arranges to include the vectors of
8720 initialization and termination functions. These functions are called
8721 via @code{__main} as described above. In order to use this method,
8722 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8725 The following section describes the specific macros that control and
8726 customize the handling of initialization and termination functions.
8729 @node Macros for Initialization
8730 @subsection Macros Controlling Initialization Routines
8732 Here are the macros that control how the compiler handles initialization
8733 and termination functions:
8735 @defmac INIT_SECTION_ASM_OP
8736 If defined, a C string constant, including spacing, for the assembler
8737 operation to identify the following data as initialization code. If not
8738 defined, GCC will assume such a section does not exist. When you are
8739 using special sections for initialization and termination functions, this
8740 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8741 run the initialization functions.
8744 @defmac HAS_INIT_SECTION
8745 If defined, @code{main} will not call @code{__main} as described above.
8746 This macro should be defined for systems that control start-up code
8747 on a symbol-by-symbol basis, such as OSF/1, and should not
8748 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8751 @defmac LD_INIT_SWITCH
8752 If defined, a C string constant for a switch that tells the linker that
8753 the following symbol is an initialization routine.
8756 @defmac LD_FINI_SWITCH
8757 If defined, a C string constant for a switch that tells the linker that
8758 the following symbol is a finalization routine.
8761 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8762 If defined, a C statement that will write a function that can be
8763 automatically called when a shared library is loaded. The function
8764 should call @var{func}, which takes no arguments. If not defined, and
8765 the object format requires an explicit initialization function, then a
8766 function called @code{_GLOBAL__DI} will be generated.
8768 This function and the following one are used by collect2 when linking a
8769 shared library that needs constructors or destructors, or has DWARF2
8770 exception tables embedded in the code.
8773 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8774 If defined, a C statement that will write a function that can be
8775 automatically called when a shared library is unloaded. The function
8776 should call @var{func}, which takes no arguments. If not defined, and
8777 the object format requires an explicit finalization function, then a
8778 function called @code{_GLOBAL__DD} will be generated.
8781 @defmac INVOKE__main
8782 If defined, @code{main} will call @code{__main} despite the presence of
8783 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8784 where the init section is not actually run automatically, but is still
8785 useful for collecting the lists of constructors and destructors.
8788 @defmac SUPPORTS_INIT_PRIORITY
8789 If nonzero, the C++ @code{init_priority} attribute is supported and the
8790 compiler should emit instructions to control the order of initialization
8791 of objects. If zero, the compiler will issue an error message upon
8792 encountering an @code{init_priority} attribute.
8795 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8796 This value is true if the target supports some ``native'' method of
8797 collecting constructors and destructors to be run at startup and exit.
8798 It is false if we must use @command{collect2}.
8801 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8802 If defined, a function that outputs assembler code to arrange to call
8803 the function referenced by @var{symbol} at initialization time.
8805 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8806 no arguments and with no return value. If the target supports initialization
8807 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8808 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8810 If this macro is not defined by the target, a suitable default will
8811 be chosen if (1) the target supports arbitrary section names, (2) the
8812 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8816 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8817 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8818 functions rather than initialization functions.
8821 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8822 generated for the generated object file will have static linkage.
8824 If your system uses @command{collect2} as the means of processing
8825 constructors, then that program normally uses @command{nm} to scan
8826 an object file for constructor functions to be called.
8828 On certain kinds of systems, you can define this macro to make
8829 @command{collect2} work faster (and, in some cases, make it work at all):
8831 @defmac OBJECT_FORMAT_COFF
8832 Define this macro if the system uses COFF (Common Object File Format)
8833 object files, so that @command{collect2} can assume this format and scan
8834 object files directly for dynamic constructor/destructor functions.
8836 This macro is effective only in a native compiler; @command{collect2} as
8837 part of a cross compiler always uses @command{nm} for the target machine.
8840 @defmac REAL_NM_FILE_NAME
8841 Define this macro as a C string constant containing the file name to use
8842 to execute @command{nm}. The default is to search the path normally for
8847 @command{collect2} calls @command{nm} to scan object files for static
8848 constructors and destructors and LTO info. By default, @option{-n} is
8849 passed. Define @code{NM_FLAGS} to a C string constant if other options
8850 are needed to get the same output format as GNU @command{nm -n}
8854 If your system supports shared libraries and has a program to list the
8855 dynamic dependencies of a given library or executable, you can define
8856 these macros to enable support for running initialization and
8857 termination functions in shared libraries:
8860 Define this macro to a C string constant containing the name of the program
8861 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8864 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8865 Define this macro to be C code that extracts filenames from the output
8866 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8867 of type @code{char *} that points to the beginning of a line of output
8868 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8869 code must advance @var{ptr} to the beginning of the filename on that
8870 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8873 @defmac SHLIB_SUFFIX
8874 Define this macro to a C string constant containing the default shared
8875 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8876 strips version information after this suffix when generating global
8877 constructor and destructor names. This define is only needed on targets
8878 that use @command{collect2} to process constructors and destructors.
8881 @node Instruction Output
8882 @subsection Output of Assembler Instructions
8884 @c prevent bad page break with this line
8885 This describes assembler instruction output.
8887 @defmac REGISTER_NAMES
8888 A C initializer containing the assembler's names for the machine
8889 registers, each one as a C string constant. This is what translates
8890 register numbers in the compiler into assembler language.
8893 @defmac ADDITIONAL_REGISTER_NAMES
8894 If defined, a C initializer for an array of structures containing a name
8895 and a register number. This macro defines additional names for hard
8896 registers, thus allowing the @code{asm} option in declarations to refer
8897 to registers using alternate names.
8900 @defmac OVERLAPPING_REGISTER_NAMES
8901 If defined, a C initializer for an array of structures containing a
8902 name, a register number and a count of the number of consecutive
8903 machine registers the name overlaps. This macro defines additional
8904 names for hard registers, thus allowing the @code{asm} option in
8905 declarations to refer to registers using alternate names. Unlike
8906 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8907 register name implies multiple underlying registers.
8909 This macro should be used when it is important that a clobber in an
8910 @code{asm} statement clobbers all the underlying values implied by the
8911 register name. For example, on ARM, clobbering the double-precision
8912 VFP register ``d0'' implies clobbering both single-precision registers
8916 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8917 Define this macro if you are using an unusual assembler that
8918 requires different names for the machine instructions.
8920 The definition is a C statement or statements which output an
8921 assembler instruction opcode to the stdio stream @var{stream}. The
8922 macro-operand @var{ptr} is a variable of type @code{char *} which
8923 points to the opcode name in its ``internal'' form---the form that is
8924 written in the machine description. The definition should output the
8925 opcode name to @var{stream}, performing any translation you desire, and
8926 increment the variable @var{ptr} to point at the end of the opcode
8927 so that it will not be output twice.
8929 In fact, your macro definition may process less than the entire opcode
8930 name, or more than the opcode name; but if you want to process text
8931 that includes @samp{%}-sequences to substitute operands, you must take
8932 care of the substitution yourself. Just be sure to increment
8933 @var{ptr} over whatever text should not be output normally.
8935 @findex recog_data.operand
8936 If you need to look at the operand values, they can be found as the
8937 elements of @code{recog_data.operand}.
8939 If the macro definition does nothing, the instruction is output
8943 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8944 If defined, a C statement to be executed just prior to the output of
8945 assembler code for @var{insn}, to modify the extracted operands so
8946 they will be output differently.
8948 Here the argument @var{opvec} is the vector containing the operands
8949 extracted from @var{insn}, and @var{noperands} is the number of
8950 elements of the vector which contain meaningful data for this insn.
8951 The contents of this vector are what will be used to convert the insn
8952 template into assembler code, so you can change the assembler output
8953 by changing the contents of the vector.
8955 This macro is useful when various assembler syntaxes share a single
8956 file of instruction patterns; by defining this macro differently, you
8957 can cause a large class of instructions to be output differently (such
8958 as with rearranged operands). Naturally, variations in assembler
8959 syntax affecting individual insn patterns ought to be handled by
8960 writing conditional output routines in those patterns.
8962 If this macro is not defined, it is equivalent to a null statement.
8965 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx_insn *@var{insn}, rtx *@var{opvec}, int @var{noperands})
8966 If defined, this target hook is a function which is executed just after the
8967 output of assembler code for @var{insn}, to change the mode of the assembler
8970 Here the argument @var{opvec} is the vector containing the operands
8971 extracted from @var{insn}, and @var{noperands} is the number of
8972 elements of the vector which contain meaningful data for this insn.
8973 The contents of this vector are what was used to convert the insn
8974 template into assembler code, so you can change the assembler mode
8975 by checking the contents of the vector.
8978 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8979 A C compound statement to output to stdio stream @var{stream} the
8980 assembler syntax for an instruction operand @var{x}. @var{x} is an
8983 @var{code} is a value that can be used to specify one of several ways
8984 of printing the operand. It is used when identical operands must be
8985 printed differently depending on the context. @var{code} comes from
8986 the @samp{%} specification that was used to request printing of the
8987 operand. If the specification was just @samp{%@var{digit}} then
8988 @var{code} is 0; if the specification was @samp{%@var{ltr}
8989 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8992 If @var{x} is a register, this macro should print the register's name.
8993 The names can be found in an array @code{reg_names} whose type is
8994 @code{char *[]}. @code{reg_names} is initialized from
8995 @code{REGISTER_NAMES}.
8997 When the machine description has a specification @samp{%@var{punct}}
8998 (a @samp{%} followed by a punctuation character), this macro is called
8999 with a null pointer for @var{x} and the punctuation character for
9003 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
9004 A C expression which evaluates to true if @var{code} is a valid
9005 punctuation character for use in the @code{PRINT_OPERAND} macro. If
9006 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
9007 punctuation characters (except for the standard one, @samp{%}) are used
9011 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
9012 A C compound statement to output to stdio stream @var{stream} the
9013 assembler syntax for an instruction operand that is a memory reference
9014 whose address is @var{x}. @var{x} is an RTL expression.
9016 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
9017 On some machines, the syntax for a symbolic address depends on the
9018 section that the address refers to. On these machines, define the hook
9019 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
9020 @code{symbol_ref}, and then check for it here. @xref{Assembler
9024 @findex dbr_sequence_length
9025 @defmac DBR_OUTPUT_SEQEND (@var{file})
9026 A C statement, to be executed after all slot-filler instructions have
9027 been output. If necessary, call @code{dbr_sequence_length} to
9028 determine the number of slots filled in a sequence (zero if not
9029 currently outputting a sequence), to decide how many no-ops to output,
9032 Don't define this macro if it has nothing to do, but it is helpful in
9033 reading assembly output if the extent of the delay sequence is made
9034 explicit (e.g.@: with white space).
9037 @findex final_sequence
9038 Note that output routines for instructions with delay slots must be
9039 prepared to deal with not being output as part of a sequence
9040 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
9041 found.) The variable @code{final_sequence} is null when not
9042 processing a sequence, otherwise it contains the @code{sequence} rtx
9046 @defmac REGISTER_PREFIX
9047 @defmacx LOCAL_LABEL_PREFIX
9048 @defmacx USER_LABEL_PREFIX
9049 @defmacx IMMEDIATE_PREFIX
9050 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
9051 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
9052 @file{final.c}). These are useful when a single @file{md} file must
9053 support multiple assembler formats. In that case, the various @file{tm.h}
9054 files can define these macros differently.
9057 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
9058 If defined this macro should expand to a series of @code{case}
9059 statements which will be parsed inside the @code{switch} statement of
9060 the @code{asm_fprintf} function. This allows targets to define extra
9061 printf formats which may useful when generating their assembler
9062 statements. Note that uppercase letters are reserved for future
9063 generic extensions to asm_fprintf, and so are not available to target
9064 specific code. The output file is given by the parameter @var{file}.
9065 The varargs input pointer is @var{argptr} and the rest of the format
9066 string, starting the character after the one that is being switched
9067 upon, is pointed to by @var{format}.
9070 @defmac ASSEMBLER_DIALECT
9071 If your target supports multiple dialects of assembler language (such as
9072 different opcodes), define this macro as a C expression that gives the
9073 numeric index of the assembler language dialect to use, with zero as the
9076 If this macro is defined, you may use constructs of the form
9078 @samp{@{option0|option1|option2@dots{}@}}
9081 in the output templates of patterns (@pxref{Output Template}) or in the
9082 first argument of @code{asm_fprintf}. This construct outputs
9083 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
9084 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
9085 within these strings retain their usual meaning. If there are fewer
9086 alternatives within the braces than the value of
9087 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
9088 to print curly braces or @samp{|} character in assembler output directly,
9089 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
9091 If you do not define this macro, the characters @samp{@{}, @samp{|} and
9092 @samp{@}} do not have any special meaning when used in templates or
9093 operands to @code{asm_fprintf}.
9095 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
9096 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
9097 the variations in assembler language syntax with that mechanism. Define
9098 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
9099 if the syntax variant are larger and involve such things as different
9100 opcodes or operand order.
9103 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
9104 A C expression to output to @var{stream} some assembler code
9105 which will push hard register number @var{regno} onto the stack.
9106 The code need not be optimal, since this macro is used only when
9110 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
9111 A C expression to output to @var{stream} some assembler code
9112 which will pop hard register number @var{regno} off of the stack.
9113 The code need not be optimal, since this macro is used only when
9117 @node Dispatch Tables
9118 @subsection Output of Dispatch Tables
9120 @c prevent bad page break with this line
9121 This concerns dispatch tables.
9123 @cindex dispatch table
9124 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
9125 A C statement to output to the stdio stream @var{stream} an assembler
9126 pseudo-instruction to generate a difference between two labels.
9127 @var{value} and @var{rel} are the numbers of two internal labels. The
9128 definitions of these labels are output using
9129 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
9130 way here. For example,
9133 fprintf (@var{stream}, "\t.word L%d-L%d\n",
9134 @var{value}, @var{rel})
9137 You must provide this macro on machines where the addresses in a
9138 dispatch table are relative to the table's own address. If defined, GCC
9139 will also use this macro on all machines when producing PIC@.
9140 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
9141 mode and flags can be read.
9144 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
9145 This macro should be provided on machines where the addresses
9146 in a dispatch table are absolute.
9148 The definition should be a C statement to output to the stdio stream
9149 @var{stream} an assembler pseudo-instruction to generate a reference to
9150 a label. @var{value} is the number of an internal label whose
9151 definition is output using @code{(*targetm.asm_out.internal_label)}.
9155 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
9159 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
9160 Define this if the label before a jump-table needs to be output
9161 specially. The first three arguments are the same as for
9162 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
9163 jump-table which follows (a @code{jump_table_data} containing an
9164 @code{addr_vec} or @code{addr_diff_vec}).
9166 This feature is used on system V to output a @code{swbeg} statement
9169 If this macro is not defined, these labels are output with
9170 @code{(*targetm.asm_out.internal_label)}.
9173 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
9174 Define this if something special must be output at the end of a
9175 jump-table. The definition should be a C statement to be executed
9176 after the assembler code for the table is written. It should write
9177 the appropriate code to stdio stream @var{stream}. The argument
9178 @var{table} is the jump-table insn, and @var{num} is the label-number
9179 of the preceding label.
9181 If this macro is not defined, nothing special is output at the end of
9185 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
9186 This target hook emits a label at the beginning of each FDE@. It
9187 should be defined on targets where FDEs need special labels, and it
9188 should write the appropriate label, for the FDE associated with the
9189 function declaration @var{decl}, to the stdio stream @var{stream}.
9190 The third argument, @var{for_eh}, is a boolean: true if this is for an
9191 exception table. The fourth argument, @var{empty}, is a boolean:
9192 true if this is a placeholder label for an omitted FDE@.
9194 The default is that FDEs are not given nonlocal labels.
9197 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
9198 This target hook emits a label at the beginning of the exception table.
9199 It should be defined on targets where it is desirable for the table
9200 to be broken up according to function.
9202 The default is that no label is emitted.
9205 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
9206 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.
9209 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx_insn *@var{insn})
9210 This target hook emits assembly directives required to unwind the
9211 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
9212 returns @code{UI_TARGET}.
9215 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
9216 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.
9219 @node Exception Region Output
9220 @subsection Assembler Commands for Exception Regions
9222 @c prevent bad page break with this line
9224 This describes commands marking the start and the end of an exception
9227 @defmac EH_FRAME_SECTION_NAME
9228 If defined, a C string constant for the name of the section containing
9229 exception handling frame unwind information. If not defined, GCC will
9230 provide a default definition if the target supports named sections.
9231 @file{crtstuff.c} uses this macro to switch to the appropriate section.
9233 You should define this symbol if your target supports DWARF 2 frame
9234 unwind information and the default definition does not work.
9237 @defmac EH_FRAME_THROUGH_COLLECT2
9238 If defined, DWARF 2 frame unwind information will identified by
9239 specially named labels. The collect2 process will locate these
9240 labels and generate code to register the frames.
9242 This might be necessary, for instance, if the system linker will not
9243 place the eh_frames in-between the sentinals from @file{crtstuff.c},
9244 or if the system linker does garbage collection and sections cannot
9245 be marked as not to be collected.
9248 @defmac EH_TABLES_CAN_BE_READ_ONLY
9249 Define this macro to 1 if your target is such that no frame unwind
9250 information encoding used with non-PIC code will ever require a
9251 runtime relocation, but the linker may not support merging read-only
9252 and read-write sections into a single read-write section.
9255 @defmac MASK_RETURN_ADDR
9256 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
9257 that it does not contain any extraneous set bits in it.
9260 @defmac DWARF2_UNWIND_INFO
9261 Define this macro to 0 if your target supports DWARF 2 frame unwind
9262 information, but it does not yet work with exception handling.
9263 Otherwise, if your target supports this information (if it defines
9264 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
9265 GCC will provide a default definition of 1.
9268 @deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
9269 This hook defines the mechanism that will be used for exception handling
9270 by the target. If the target has ABI specified unwind tables, the hook
9271 should return @code{UI_TARGET}. If the target is to use the
9272 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
9273 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
9274 information, the hook should return @code{UI_DWARF2}.
9276 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
9277 This may end up simplifying other parts of target-specific code. The
9278 default implementation of this hook never returns @code{UI_NONE}.
9280 Note that the value returned by this hook should be constant. It should
9281 not depend on anything except the command-line switches described by
9282 @var{opts}. In particular, the
9283 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
9284 macros and builtin functions related to exception handling are set up
9285 depending on this setting.
9287 The default implementation of the hook first honors the
9288 @option{--enable-sjlj-exceptions} configure option, then
9289 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
9290 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
9291 must define this hook so that @var{opts} is used correctly.
9294 @deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
9295 This variable should be set to @code{true} if the target ABI requires unwinding
9296 tables even when exceptions are not used. It must not be modified by
9297 command-line option processing.
9300 @defmac DONT_USE_BUILTIN_SETJMP
9301 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
9302 should use the @code{setjmp}/@code{longjmp} functions from the C library
9303 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
9306 @defmac JMP_BUF_SIZE
9307 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
9308 defined. Define this macro if the default size of @code{jmp_buf} buffer
9309 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
9310 is not large enough, or if it is much too large.
9311 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
9314 @defmac DWARF_CIE_DATA_ALIGNMENT
9315 This macro need only be defined if the target might save registers in the
9316 function prologue at an offset to the stack pointer that is not aligned to
9317 @code{UNITS_PER_WORD}. The definition should be the negative minimum
9318 alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive
9319 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
9320 the target supports DWARF 2 frame unwind information.
9323 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
9324 Contains the value true if the target should add a zero word onto the
9325 end of a Dwarf-2 frame info section when used for exception handling.
9326 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
9330 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
9331 Given a register, this hook should return a parallel of registers to
9332 represent where to find the register pieces. Define this hook if the
9333 register and its mode are represented in Dwarf in non-contiguous
9334 locations, or if the register should be represented in more than one
9335 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
9336 If not defined, the default is to return @code{NULL_RTX}.
9339 @deftypefn {Target Hook} machine_mode TARGET_DWARF_FRAME_REG_MODE (int @var{regno})
9340 Given a register, this hook should return the mode which the
9341 corresponding Dwarf frame register should have. This is normally
9342 used to return a smaller mode than the raw mode to prevent call
9343 clobbered parts of a register altering the frame register size
9346 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
9347 If some registers are represented in Dwarf-2 unwind information in
9348 multiple pieces, define this hook to fill in information about the
9349 sizes of those pieces in the table used by the unwinder at runtime.
9350 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
9351 filling in a single size corresponding to each hard register;
9352 @var{address} is the address of the table.
9355 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
9356 This hook is used to output a reference from a frame unwinding table to
9357 the type_info object identified by @var{sym}. It should return @code{true}
9358 if the reference was output. Returning @code{false} will cause the
9359 reference to be output using the normal Dwarf2 routines.
9362 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
9363 This flag should be set to @code{true} on targets that use an ARM EABI
9364 based unwinding library, and @code{false} on other targets. This effects
9365 the format of unwinding tables, and how the unwinder in entered after
9366 running a cleanup. The default is @code{false}.
9369 @node Alignment Output
9370 @subsection Assembler Commands for Alignment
9372 @c prevent bad page break with this line
9373 This describes commands for alignment.
9375 @defmac JUMP_ALIGN (@var{label})
9376 The alignment (log base 2) to put in front of @var{label}, which is
9377 a common destination of jumps and has no fallthru incoming edge.
9379 This macro need not be defined if you don't want any special alignment
9380 to be done at such a time. Most machine descriptions do not currently
9383 Unless it's necessary to inspect the @var{label} parameter, it is better
9384 to set the variable @var{align_jumps} in the target's
9385 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9386 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
9389 @deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx_insn *@var{label})
9390 The maximum number of bytes to skip before @var{label} when applying
9391 @code{JUMP_ALIGN}. This works only if
9392 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9395 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
9396 The alignment (log base 2) to put in front of @var{label}, which follows
9399 This macro need not be defined if you don't want any special alignment
9400 to be done at such a time. Most machine descriptions do not currently
9404 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx_insn *@var{label})
9405 The maximum number of bytes to skip before @var{label} when applying
9406 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
9407 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9410 @defmac LOOP_ALIGN (@var{label})
9411 The alignment (log base 2) to put in front of @var{label} that heads
9412 a frequently executed basic block (usually the header of a loop).
9414 This macro need not be defined if you don't want any special alignment
9415 to be done at such a time. Most machine descriptions do not currently
9418 Unless it's necessary to inspect the @var{label} parameter, it is better
9419 to set the variable @code{align_loops} in the target's
9420 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9421 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
9424 @deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx_insn *@var{label})
9425 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
9426 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
9430 @defmac LABEL_ALIGN (@var{label})
9431 The alignment (log base 2) to put in front of @var{label}.
9432 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
9433 the maximum of the specified values is used.
9435 Unless it's necessary to inspect the @var{label} parameter, it is better
9436 to set the variable @code{align_labels} in the target's
9437 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9438 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9441 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx_insn *@var{label})
9442 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
9443 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
9447 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9448 A C statement to output to the stdio stream @var{stream} an assembler
9449 instruction to advance the location counter by @var{nbytes} bytes.
9450 Those bytes should be zero when loaded. @var{nbytes} will be a C
9451 expression of type @code{unsigned HOST_WIDE_INT}.
9454 @defmac ASM_NO_SKIP_IN_TEXT
9455 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9456 text section because it fails to put zeros in the bytes that are skipped.
9457 This is true on many Unix systems, where the pseudo--op to skip bytes
9458 produces no-op instructions rather than zeros when used in the text
9462 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9463 A C statement to output to the stdio stream @var{stream} an assembler
9464 command to advance the location counter to a multiple of 2 to the
9465 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9468 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9469 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9470 for padding, if necessary.
9473 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9474 A C statement to output to the stdio stream @var{stream} an assembler
9475 command to advance the location counter to a multiple of 2 to the
9476 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9477 satisfy the alignment request. @var{power} and @var{max_skip} will be
9478 a C expression of type @code{int}.
9482 @node Debugging Info
9483 @section Controlling Debugging Information Format
9485 @c prevent bad page break with this line
9486 This describes how to specify debugging information.
9489 * All Debuggers:: Macros that affect all debugging formats uniformly.
9490 * DBX Options:: Macros enabling specific options in DBX format.
9491 * DBX Hooks:: Hook macros for varying DBX format.
9492 * File Names and DBX:: Macros controlling output of file names in DBX format.
9493 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9494 * VMS Debug:: Macros for VMS debug format.
9498 @subsection Macros Affecting All Debugging Formats
9500 @c prevent bad page break with this line
9501 These macros affect all debugging formats.
9503 @defmac DBX_REGISTER_NUMBER (@var{regno})
9504 A C expression that returns the DBX register number for the compiler
9505 register number @var{regno}. In the default macro provided, the value
9506 of this expression will be @var{regno} itself. But sometimes there are
9507 some registers that the compiler knows about and DBX does not, or vice
9508 versa. In such cases, some register may need to have one number in the
9509 compiler and another for DBX@.
9511 If two registers have consecutive numbers inside GCC, and they can be
9512 used as a pair to hold a multiword value, then they @emph{must} have
9513 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9514 Otherwise, debuggers will be unable to access such a pair, because they
9515 expect register pairs to be consecutive in their own numbering scheme.
9517 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9518 does not preserve register pairs, then what you must do instead is
9519 redefine the actual register numbering scheme.
9522 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9523 A C expression that returns the integer offset value for an automatic
9524 variable having address @var{x} (an RTL expression). The default
9525 computation assumes that @var{x} is based on the frame-pointer and
9526 gives the offset from the frame-pointer. This is required for targets
9527 that produce debugging output for DBX or COFF-style debugging output
9528 for SDB and allow the frame-pointer to be eliminated when the
9529 @option{-g} options is used.
9532 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9533 A C expression that returns the integer offset value for an argument
9534 having address @var{x} (an RTL expression). The nominal offset is
9538 @defmac PREFERRED_DEBUGGING_TYPE
9539 A C expression that returns the type of debugging output GCC should
9540 produce when the user specifies just @option{-g}. Define
9541 this if you have arranged for GCC to support more than one format of
9542 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9543 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9544 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9546 When the user specifies @option{-ggdb}, GCC normally also uses the
9547 value of this macro to select the debugging output format, but with two
9548 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9549 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9550 defined, GCC uses @code{DBX_DEBUG}.
9552 The value of this macro only affects the default debugging output; the
9553 user can always get a specific type of output by using @option{-gstabs},
9554 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9558 @subsection Specific Options for DBX Output
9560 @c prevent bad page break with this line
9561 These are specific options for DBX output.
9563 @defmac DBX_DEBUGGING_INFO
9564 Define this macro if GCC should produce debugging output for DBX
9565 in response to the @option{-g} option.
9568 @defmac XCOFF_DEBUGGING_INFO
9569 Define this macro if GCC should produce XCOFF format debugging output
9570 in response to the @option{-g} option. This is a variant of DBX format.
9573 @defmac DEFAULT_GDB_EXTENSIONS
9574 Define this macro to control whether GCC should by default generate
9575 GDB's extended version of DBX debugging information (assuming DBX-format
9576 debugging information is enabled at all). If you don't define the
9577 macro, the default is 1: always generate the extended information
9578 if there is any occasion to.
9581 @defmac DEBUG_SYMS_TEXT
9582 Define this macro if all @code{.stabs} commands should be output while
9583 in the text section.
9586 @defmac ASM_STABS_OP
9587 A C string constant, including spacing, naming the assembler pseudo op to
9588 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9589 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9590 applies only to DBX debugging information format.
9593 @defmac ASM_STABD_OP
9594 A C string constant, including spacing, naming the assembler pseudo op to
9595 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9596 value is the current location. If you don't define this macro,
9597 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9601 @defmac ASM_STABN_OP
9602 A C string constant, including spacing, naming the assembler pseudo op to
9603 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9604 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9605 macro applies only to DBX debugging information format.
9608 @defmac DBX_NO_XREFS
9609 Define this macro if DBX on your system does not support the construct
9610 @samp{xs@var{tagname}}. On some systems, this construct is used to
9611 describe a forward reference to a structure named @var{tagname}.
9612 On other systems, this construct is not supported at all.
9615 @defmac DBX_CONTIN_LENGTH
9616 A symbol name in DBX-format debugging information is normally
9617 continued (split into two separate @code{.stabs} directives) when it
9618 exceeds a certain length (by default, 80 characters). On some
9619 operating systems, DBX requires this splitting; on others, splitting
9620 must not be done. You can inhibit splitting by defining this macro
9621 with the value zero. You can override the default splitting-length by
9622 defining this macro as an expression for the length you desire.
9625 @defmac DBX_CONTIN_CHAR
9626 Normally continuation is indicated by adding a @samp{\} character to
9627 the end of a @code{.stabs} string when a continuation follows. To use
9628 a different character instead, define this macro as a character
9629 constant for the character you want to use. Do not define this macro
9630 if backslash is correct for your system.
9633 @defmac DBX_STATIC_STAB_DATA_SECTION
9634 Define this macro if it is necessary to go to the data section before
9635 outputting the @samp{.stabs} pseudo-op for a non-global static
9639 @defmac DBX_TYPE_DECL_STABS_CODE
9640 The value to use in the ``code'' field of the @code{.stabs} directive
9641 for a typedef. The default is @code{N_LSYM}.
9644 @defmac DBX_STATIC_CONST_VAR_CODE
9645 The value to use in the ``code'' field of the @code{.stabs} directive
9646 for a static variable located in the text section. DBX format does not
9647 provide any ``right'' way to do this. The default is @code{N_FUN}.
9650 @defmac DBX_REGPARM_STABS_CODE
9651 The value to use in the ``code'' field of the @code{.stabs} directive
9652 for a parameter passed in registers. DBX format does not provide any
9653 ``right'' way to do this. The default is @code{N_RSYM}.
9656 @defmac DBX_REGPARM_STABS_LETTER
9657 The letter to use in DBX symbol data to identify a symbol as a parameter
9658 passed in registers. DBX format does not customarily provide any way to
9659 do this. The default is @code{'P'}.
9662 @defmac DBX_FUNCTION_FIRST
9663 Define this macro if the DBX information for a function and its
9664 arguments should precede the assembler code for the function. Normally,
9665 in DBX format, the debugging information entirely follows the assembler
9669 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9670 Define this macro, with value 1, if the value of a symbol describing
9671 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9672 relative to the start of the enclosing function. Normally, GCC uses
9673 an absolute address.
9676 @defmac DBX_LINES_FUNCTION_RELATIVE
9677 Define this macro, with value 1, if the value of a symbol indicating
9678 the current line number (@code{N_SLINE}) should be relative to the
9679 start of the enclosing function. Normally, GCC uses an absolute address.
9682 @defmac DBX_USE_BINCL
9683 Define this macro if GCC should generate @code{N_BINCL} and
9684 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9685 macro also directs GCC to output a type number as a pair of a file
9686 number and a type number within the file. Normally, GCC does not
9687 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9688 number for a type number.
9692 @subsection Open-Ended Hooks for DBX Format
9694 @c prevent bad page break with this line
9695 These are hooks for DBX format.
9697 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9698 A C statement to output DBX debugging information before code for line
9699 number @var{line} of the current source file to the stdio stream
9700 @var{stream}. @var{counter} is the number of time the macro was
9701 invoked, including the current invocation; it is intended to generate
9702 unique labels in the assembly output.
9704 This macro should not be defined if the default output is correct, or
9705 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9708 @defmac NO_DBX_FUNCTION_END
9709 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9710 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9711 On those machines, define this macro to turn this feature off without
9712 disturbing the rest of the gdb extensions.
9715 @defmac NO_DBX_BNSYM_ENSYM
9716 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9717 extension construct. On those machines, define this macro to turn this
9718 feature off without disturbing the rest of the gdb extensions.
9721 @node File Names and DBX
9722 @subsection File Names in DBX Format
9724 @c prevent bad page break with this line
9725 This describes file names in DBX format.
9727 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9728 A C statement to output DBX debugging information to the stdio stream
9729 @var{stream}, which indicates that file @var{name} is the main source
9730 file---the file specified as the input file for compilation.
9731 This macro is called only once, at the beginning of compilation.
9733 This macro need not be defined if the standard form of output
9734 for DBX debugging information is appropriate.
9736 It may be necessary to refer to a label equal to the beginning of the
9737 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9738 to do so. If you do this, you must also set the variable
9739 @var{used_ltext_label_name} to @code{true}.
9742 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9743 Define this macro, with value 1, if GCC should not emit an indication
9744 of the current directory for compilation and current source language at
9745 the beginning of the file.
9748 @defmac NO_DBX_GCC_MARKER
9749 Define this macro, with value 1, if GCC should not emit an indication
9750 that this object file was compiled by GCC@. The default is to emit
9751 an @code{N_OPT} stab at the beginning of every source file, with
9752 @samp{gcc2_compiled.} for the string and value 0.
9755 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9756 A C statement to output DBX debugging information at the end of
9757 compilation of the main source file @var{name}. Output should be
9758 written to the stdio stream @var{stream}.
9760 If you don't define this macro, nothing special is output at the end
9761 of compilation, which is correct for most machines.
9764 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9765 Define this macro @emph{instead of} defining
9766 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9767 the end of compilation is an @code{N_SO} stab with an empty string,
9768 whose value is the highest absolute text address in the file.
9773 @subsection Macros for SDB and DWARF Output
9775 @c prevent bad page break with this line
9776 Here are macros for SDB and DWARF output.
9778 @defmac SDB_DEBUGGING_INFO
9779 Define this macro to 1 if GCC should produce COFF-style debugging output
9780 for SDB in response to the @option{-g} option.
9783 @defmac DWARF2_DEBUGGING_INFO
9784 Define this macro if GCC should produce dwarf version 2 format
9785 debugging output in response to the @option{-g} option.
9787 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
9788 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9789 be emitted for each function. Instead of an integer return the enum
9790 value for the @code{DW_CC_} tag.
9793 To support optional call frame debugging information, you must also
9794 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9795 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9796 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9797 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9800 @defmac DWARF2_FRAME_INFO
9801 Define this macro to a nonzero value if GCC should always output
9802 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9803 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9804 exceptions are enabled, GCC will output this information not matter
9805 how you define @code{DWARF2_FRAME_INFO}.
9808 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
9809 This hook defines the mechanism that will be used for describing frame
9810 unwind information to the debugger. Normally the hook will return
9811 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9812 return @code{UI_NONE} otherwise.
9814 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9815 is disabled in order to always output DWARF 2 frame information.
9817 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9818 This will suppress generation of the normal debug frame unwind information.
9821 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9822 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9823 line debug info sections. This will result in much more compact line number
9824 tables, and hence is desirable if it works.
9827 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
9828 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.
9831 @deftypevr {Target Hook} bool TARGET_DELAY_SCHED2
9832 True if sched2 is not to be run at its normal place.
9833 This usually means it will be run as part of machine-specific reorg.
9836 @deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK
9837 True if vartrack is not to be run at its normal place.
9838 This usually means it will be run as part of machine-specific reorg.
9841 @deftypevr {Target Hook} bool TARGET_NO_REGISTER_ALLOCATION
9842 True if register allocation and the passes
9843 following it should not be run. Usually true only for virtual assembler
9847 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9848 A C statement to issue assembly directives that create a difference
9849 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9852 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9853 A C statement to issue assembly directives that create a difference
9854 between the two given labels in system defined units, e.g. instruction
9855 slots on IA64 VMS, using an integer of the given size.
9858 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{offset}, @var{section})
9859 A C statement to issue assembly directives that create a
9860 section-relative reference to the given @var{label} plus @var{offset}, using
9861 an integer of the given @var{size}. The label is known to be defined in the
9862 given @var{section}.
9865 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9866 A C statement to issue assembly directives that create a self-relative
9867 reference to the given @var{label}, using an integer of the given @var{size}.
9870 @defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label})
9871 A C statement to issue assembly directives that create a reference to the
9872 given @var{label} relative to the dbase, using an integer of the given @var{size}.
9875 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9876 A C statement to issue assembly directives that create a reference to
9877 the DWARF table identifier @var{label} from the current section. This
9878 is used on some systems to avoid garbage collecting a DWARF table which
9879 is referenced by a function.
9882 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
9883 If defined, this target hook is a function which outputs a DTP-relative
9884 reference to the given TLS symbol of the specified size.
9887 @defmac PUT_SDB_@dots{}
9888 Define these macros to override the assembler syntax for the special
9889 SDB assembler directives. See @file{sdbout.c} for a list of these
9890 macros and their arguments. If the standard syntax is used, you need
9891 not define them yourself.
9895 Some assemblers do not support a semicolon as a delimiter, even between
9896 SDB assembler directives. In that case, define this macro to be the
9897 delimiter to use (usually @samp{\n}). It is not necessary to define
9898 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9902 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9903 Define this macro to allow references to unknown structure,
9904 union, or enumeration tags to be emitted. Standard COFF does not
9905 allow handling of unknown references, MIPS ECOFF has support for
9909 @defmac SDB_ALLOW_FORWARD_REFERENCES
9910 Define this macro to allow references to structure, union, or
9911 enumeration tags that have not yet been seen to be handled. Some
9912 assemblers choke if forward tags are used, while some require it.
9915 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9916 A C statement to output SDB debugging information before code for line
9917 number @var{line} of the current source file to the stdio stream
9918 @var{stream}. The default is to emit an @code{.ln} directive.
9923 @subsection Macros for VMS Debug Format
9925 @c prevent bad page break with this line
9926 Here are macros for VMS debug format.
9928 @defmac VMS_DEBUGGING_INFO
9929 Define this macro if GCC should produce debugging output for VMS
9930 in response to the @option{-g} option. The default behavior for VMS
9931 is to generate minimal debug info for a traceback in the absence of
9932 @option{-g} unless explicitly overridden with @option{-g0}. This
9933 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9934 @code{TARGET_OPTION_OVERRIDE}.
9937 @node Floating Point
9938 @section Cross Compilation and Floating Point
9939 @cindex cross compilation and floating point
9940 @cindex floating point and cross compilation
9942 While all modern machines use twos-complement representation for integers,
9943 there are a variety of representations for floating point numbers. This
9944 means that in a cross-compiler the representation of floating point numbers
9945 in the compiled program may be different from that used in the machine
9946 doing the compilation.
9948 Because different representation systems may offer different amounts of
9949 range and precision, all floating point constants must be represented in
9950 the target machine's format. Therefore, the cross compiler cannot
9951 safely use the host machine's floating point arithmetic; it must emulate
9952 the target's arithmetic. To ensure consistency, GCC always uses
9953 emulation to work with floating point values, even when the host and
9954 target floating point formats are identical.
9956 The following macros are provided by @file{real.h} for the compiler to
9957 use. All parts of the compiler which generate or optimize
9958 floating-point calculations must use these macros. They may evaluate
9959 their operands more than once, so operands must not have side effects.
9961 @defmac REAL_VALUE_TYPE
9962 The C data type to be used to hold a floating point value in the target
9963 machine's format. Typically this is a @code{struct} containing an
9964 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9968 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9969 Truncates @var{x} to a signed integer, rounding toward zero.
9972 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9973 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9974 @var{x} is negative, returns zero.
9977 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode})
9978 Converts @var{string} into a floating point number in the target machine's
9979 representation for mode @var{mode}. This routine can handle both
9980 decimal and hexadecimal floating point constants, using the syntax
9981 defined by the C language for both.
9984 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9985 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9988 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9989 Determines whether @var{x} represents infinity (positive or negative).
9992 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9993 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9996 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9997 Returns the negative of the floating point value @var{x}.
10000 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
10001 Returns the absolute value of @var{x}.
10004 @node Mode Switching
10005 @section Mode Switching Instructions
10006 @cindex mode switching
10007 The following macros control mode switching optimizations:
10009 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
10010 Define this macro if the port needs extra instructions inserted for mode
10011 switching in an optimizing compilation.
10013 For an example, the SH4 can perform both single and double precision
10014 floating point operations, but to perform a single precision operation,
10015 the FPSCR PR bit has to be cleared, while for a double precision
10016 operation, this bit has to be set. Changing the PR bit requires a general
10017 purpose register as a scratch register, hence these FPSCR sets have to
10018 be inserted before reload, i.e.@: you cannot put this into instruction emitting
10019 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
10021 You can have multiple entities that are mode-switched, and select at run time
10022 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
10023 return nonzero for any @var{entity} that needs mode-switching.
10024 If you define this macro, you also have to define
10025 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
10026 @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
10027 @code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
10031 @defmac NUM_MODES_FOR_MODE_SWITCHING
10032 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
10033 initializer for an array of integers. Each initializer element
10034 N refers to an entity that needs mode switching, and specifies the number
10035 of different modes that might need to be set for this entity.
10036 The position of the initializer in the initializer---starting counting at
10037 zero---determines the integer that is used to refer to the mode-switched
10038 entity in question.
10039 In macros that take mode arguments / yield a mode result, modes are
10040 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
10041 switch is needed / supplied.
10044 @deftypefn {Target Hook} void TARGET_MODE_EMIT (int @var{entity}, int @var{mode}, int @var{prev_mode}, HARD_REG_SET @var{regs_live})
10045 Generate one or more insns to set @var{entity} to @var{mode}. @var{hard_reg_live} is the set of hard registers live at the point where the insn(s) are to be inserted. @var{prev_moxde} indicates the mode to switch from. Sets of a lower numbered entity will be emitted before sets of a higher numbered entity to a mode of the same or lower priority.
10048 @deftypefn {Target Hook} int TARGET_MODE_NEEDED (int @var{entity}, rtx_insn *@var{insn})
10049 @var{entity} is an integer specifying a mode-switched entity. If @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to return an integer value not larger than the corresponding element in @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must be switched into prior to the execution of @var{insn}.
10052 @deftypefn {Target Hook} int TARGET_MODE_AFTER (int @var{entity}, int @var{mode}, rtx_insn *@var{insn})
10053 @var{entity} is an integer specifying a mode-switched entity. If this macro is defined, it is evaluated for every @var{insn} during mode switching. It determines the mode that an insn results in (if different from the incoming mode).
10056 @deftypefn {Target Hook} int TARGET_MODE_ENTRY (int @var{entity})
10057 If this macro is defined, it is evaluated for every @var{entity} that needs mode switching. It should evaluate to an integer, which is a mode that @var{entity} is assumed to be switched to at function entry. If @code{TARGET_MODE_ENTRY} is defined then @code{TARGET_MODE_EXIT} must be defined.
10060 @deftypefn {Target Hook} int TARGET_MODE_EXIT (int @var{entity})
10061 If this macro is defined, it is evaluated for every @var{entity} that needs mode switching. It should evaluate to an integer, which is a mode that @var{entity} is assumed to be switched to at function exit. If @code{TARGET_MODE_EXIT} is defined then @code{TARGET_MODE_ENTRY} must be defined.
10064 @deftypefn {Target Hook} int TARGET_MODE_PRIORITY (int @var{entity}, int @var{n})
10065 This macro specifies the order in which modes for @var{entity} are processed. 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the lowest. The value of the macro should be an integer designating a mode for @var{entity}. For any fixed @var{entity}, @code{mode_priority} (@var{entity}, @var{n}) shall be a bijection in 0 @dots{} @code{num_modes_for_mode_switching[@var{entity}] - 1}.
10068 @node Target Attributes
10069 @section Defining target-specific uses of @code{__attribute__}
10070 @cindex target attributes
10071 @cindex machine attributes
10072 @cindex attributes, target-specific
10074 Target-specific attributes may be defined for functions, data and types.
10075 These are described using the following target hooks; they also need to
10076 be documented in @file{extend.texi}.
10078 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
10079 If defined, this target hook points to an array of @samp{struct
10080 attribute_spec} (defined in @file{tree-core.h}) specifying the machine
10081 specific attributes for this target and some of the restrictions on the
10082 entities to which these attributes are applied and the arguments they
10086 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
10087 If defined, this target hook is a function which returns true if the
10088 machine-specific attribute named @var{name} expects an identifier
10089 given as its first argument to be passed on as a plain identifier, not
10090 subjected to name lookup. If this is not defined, the default is
10091 false for all machine-specific attributes.
10094 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
10095 If defined, this target hook is a function which returns zero if the attributes on
10096 @var{type1} and @var{type2} are incompatible, one if they are compatible,
10097 and two if they are nearly compatible (which causes a warning to be
10098 generated). If this is not defined, machine-specific attributes are
10099 supposed always to be compatible.
10102 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
10103 If defined, this target hook is a function which assigns default attributes to
10104 the newly defined @var{type}.
10107 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
10108 Define this target hook if the merging of type attributes needs special
10109 handling. If defined, the result is a list of the combined
10110 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
10111 that @code{comptypes} has already been called and returned 1. This
10112 function may call @code{merge_attributes} to handle machine-independent
10116 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
10117 Define this target hook if the merging of decl attributes needs special
10118 handling. If defined, the result is a list of the combined
10119 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
10120 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
10121 when this is needed are when one attribute overrides another, or when an
10122 attribute is nullified by a subsequent definition. This function may
10123 call @code{merge_attributes} to handle machine-independent merging.
10125 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
10126 If the only target-specific handling you require is @samp{dllimport}
10127 for Microsoft Windows targets, you should define the macro
10128 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
10129 will then define a function called
10130 @code{merge_dllimport_decl_attributes} which can then be defined as
10131 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
10132 add @code{handle_dll_attribute} in the attribute table for your port
10133 to perform initial processing of the @samp{dllimport} and
10134 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
10135 @file{i386/i386.c}, for example.
10138 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
10139 @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}.
10142 @defmac TARGET_DECLSPEC
10143 Define this macro to a nonzero value if you want to treat
10144 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
10145 default, this behavior is enabled only for targets that define
10146 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
10147 of @code{__declspec} is via a built-in macro, but you should not rely
10148 on this implementation detail.
10151 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
10152 Define this target hook if you want to be able to add attributes to a decl
10153 when it is being created. This is normally useful for back ends which
10154 wish to implement a pragma by using the attributes which correspond to
10155 the pragma's effect. The @var{node} argument is the decl which is being
10156 created. The @var{attr_ptr} argument is a pointer to the attribute list
10157 for this decl. The list itself should not be modified, since it may be
10158 shared with other decls, but attributes may be chained on the head of
10159 the list and @code{*@var{attr_ptr}} modified to point to the new
10160 attributes, or a copy of the list may be made if further changes are
10164 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
10166 This target hook returns @code{true} if it is OK to inline @var{fndecl}
10167 into the current function, despite its having target-specific
10168 attributes, @code{false} otherwise. By default, if a function has a
10169 target specific attribute attached to it, it will not be inlined.
10172 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
10173 This hook is called to parse @code{attribute(target("..."))}, which
10174 allows setting target-specific options on individual functions.
10175 These function-specific options may differ
10176 from the options specified on the command line. The hook should return
10177 @code{true} if the options are valid.
10179 The hook should set the @code{DECL_FUNCTION_SPECIFIC_TARGET} field in
10180 the function declaration to hold a pointer to a target-specific
10181 @code{struct cl_target_option} structure.
10184 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr}, struct gcc_options *@var{opts})
10185 This hook is called to save any additional target-specific information
10186 in the @code{struct cl_target_option} structure for function-specific
10187 options from the @code{struct gcc_options} structure.
10188 @xref{Option file format}.
10191 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct gcc_options *@var{opts}, struct cl_target_option *@var{ptr})
10192 This hook is called to restore any additional target-specific
10193 information in the @code{struct cl_target_option} structure for
10194 function-specific options to the @code{struct gcc_options} structure.
10197 @deftypefn {Target Hook} void TARGET_OPTION_POST_STREAM_IN (struct cl_target_option *@var{ptr})
10198 This hook is called to update target-specific information in the
10199 @code{struct cl_target_option} structure after it is streamed in from
10203 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
10204 This hook is called to print any additional target-specific
10205 information in the @code{struct cl_target_option} structure for
10206 function-specific options.
10209 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
10210 This target hook parses the options for @code{#pragma GCC target}, which
10211 sets the target-specific options for functions that occur later in the
10212 input stream. The options accepted should be the same as those handled by the
10213 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
10216 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
10217 Sometimes certain combinations of command options do not make sense on
10218 a particular target machine. You can override the hook
10219 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
10220 once just after all the command options have been parsed.
10222 Don't use this hook to turn on various extra optimizations for
10223 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
10225 If you need to do something whenever the optimization level is
10226 changed via the optimize attribute or pragma, see
10227 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
10230 @deftypefn {Target Hook} bool TARGET_OPTION_FUNCTION_VERSIONS (tree @var{decl1}, tree @var{decl2})
10231 This target hook returns @code{true} if @var{DECL1} and @var{DECL2} are
10232 versions of the same function. @var{DECL1} and @var{DECL2} are function
10233 versions if and only if they have the same function signature and
10234 different target specific attributes, that is, they are compiled for
10235 different target machines.
10238 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
10239 This target hook returns @code{false} if the @var{caller} function
10240 cannot inline @var{callee}, based on target specific information. By
10241 default, inlining is not allowed if the callee function has function
10242 specific target options and the caller does not use the same options.
10245 @deftypefn {Target Hook} void TARGET_RELAYOUT_FUNCTION (tree @var{fndecl})
10246 This target hook fixes function @var{fndecl} after attributes are processed. Default does nothing. On ARM, the default function's alignment is updated with the attribute target.
10250 @section Emulating TLS
10251 @cindex Emulated TLS
10253 For targets whose psABI does not provide Thread Local Storage via
10254 specific relocations and instruction sequences, an emulation layer is
10255 used. A set of target hooks allows this emulation layer to be
10256 configured for the requirements of a particular target. For instance
10257 the psABI may in fact specify TLS support in terms of an emulation
10260 The emulation layer works by creating a control object for every TLS
10261 object. To access the TLS object, a lookup function is provided
10262 which, when given the address of the control object, will return the
10263 address of the current thread's instance of the TLS object.
10265 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
10266 Contains the name of the helper function that uses a TLS control
10267 object to locate a TLS instance. The default causes libgcc's
10268 emulated TLS helper function to be used.
10271 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
10272 Contains the name of the helper function that should be used at
10273 program startup to register TLS objects that are implicitly
10274 initialized to zero. If this is @code{NULL}, all TLS objects will
10275 have explicit initializers. The default causes libgcc's emulated TLS
10276 registration function to be used.
10279 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
10280 Contains the name of the section in which TLS control variables should
10281 be placed. The default of @code{NULL} allows these to be placed in
10285 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
10286 Contains the name of the section in which TLS initializers should be
10287 placed. The default of @code{NULL} allows these to be placed in any
10291 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
10292 Contains the prefix to be prepended to TLS control variable names.
10293 The default of @code{NULL} uses a target-specific prefix.
10296 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
10297 Contains the prefix to be prepended to TLS initializer objects. The
10298 default of @code{NULL} uses a target-specific prefix.
10301 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
10302 Specifies a function that generates the FIELD_DECLs for a TLS control
10303 object type. @var{type} is the RECORD_TYPE the fields are for and
10304 @var{name} should be filled with the structure tag, if the default of
10305 @code{__emutls_object} is unsuitable. The default creates a type suitable
10306 for libgcc's emulated TLS function.
10309 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
10310 Specifies a function that generates the CONSTRUCTOR to initialize a
10311 TLS control object. @var{var} is the TLS control object, @var{decl}
10312 is the TLS object and @var{tmpl_addr} is the address of the
10313 initializer. The default initializes libgcc's emulated TLS control object.
10316 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
10317 Specifies whether the alignment of TLS control variable objects is
10318 fixed and should not be increased as some backends may do to optimize
10319 single objects. The default is false.
10322 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
10323 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
10324 may be used to describe emulated TLS control objects.
10327 @node MIPS Coprocessors
10328 @section Defining coprocessor specifics for MIPS targets.
10329 @cindex MIPS coprocessor-definition macros
10331 The MIPS specification allows MIPS implementations to have as many as 4
10332 coprocessors, each with as many as 32 private registers. GCC supports
10333 accessing these registers and transferring values between the registers
10334 and memory using asm-ized variables. For example:
10337 register unsigned int cp0count asm ("c0r1");
10343 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
10344 names may be added as described below, or the default names may be
10345 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
10347 Coprocessor registers are assumed to be epilogue-used; sets to them will
10348 be preserved even if it does not appear that the register is used again
10349 later in the function.
10351 Another note: according to the MIPS spec, coprocessor 1 (if present) is
10352 the FPU@. One accesses COP1 registers through standard mips
10353 floating-point support; they are not included in this mechanism.
10356 @section Parameters for Precompiled Header Validity Checking
10357 @cindex parameters, precompiled headers
10359 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
10360 This hook returns a pointer to the data needed by
10361 @code{TARGET_PCH_VALID_P} and sets
10362 @samp{*@var{sz}} to the size of the data in bytes.
10365 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
10366 This hook checks whether the options used to create a PCH file are
10367 compatible with the current settings. It returns @code{NULL}
10368 if so and a suitable error message if not. Error messages will
10369 be presented to the user and must be localized using @samp{_(@var{msg})}.
10371 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
10372 when the PCH file was created and @var{sz} is the size of that data in bytes.
10373 It's safe to assume that the data was created by the same version of the
10374 compiler, so no format checking is needed.
10376 The default definition of @code{default_pch_valid_p} should be
10377 suitable for most targets.
10380 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
10381 If this hook is nonnull, the default implementation of
10382 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
10383 of @code{target_flags}. @var{pch_flags} specifies the value that
10384 @code{target_flags} had when the PCH file was created. The return
10385 value is the same as for @code{TARGET_PCH_VALID_P}.
10388 @deftypefn {Target Hook} void TARGET_PREPARE_PCH_SAVE (void)
10389 Called before writing out a PCH file. If the target has some
10390 garbage-collected data that needs to be in a particular state on PCH loads,
10391 it can use this hook to enforce that state. Very few targets need
10392 to do anything here.
10396 @section C++ ABI parameters
10397 @cindex parameters, c++ abi
10399 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10400 Define this hook to override the integer type used for guard variables.
10401 These are used to implement one-time construction of static objects. The
10402 default is long_long_integer_type_node.
10405 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10406 This hook determines how guard variables are used. It should return
10407 @code{false} (the default) if the first byte should be used. A return value of
10408 @code{true} indicates that only the least significant bit should be used.
10411 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10412 This hook returns the size of the cookie to use when allocating an array
10413 whose elements have the indicated @var{type}. Assumes that it is already
10414 known that a cookie is needed. The default is
10415 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10416 IA64/Generic C++ ABI@.
10419 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10420 This hook should return @code{true} if the element size should be stored in
10421 array cookies. The default is to return @code{false}.
10424 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10425 If defined by a backend this hook allows the decision made to export
10426 class @var{type} to be overruled. Upon entry @var{import_export}
10427 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10428 to be imported and 0 otherwise. This function should return the
10429 modified value and perform any other actions necessary to support the
10430 backend's targeted operating system.
10433 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10434 This hook should return @code{true} if constructors and destructors return
10435 the address of the object created/destroyed. The default is to return
10439 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10440 This hook returns true if the key method for a class (i.e., the method
10441 which, if defined in the current translation unit, causes the virtual
10442 table to be emitted) may be an inline function. Under the standard
10443 Itanium C++ ABI the key method may be an inline function so long as
10444 the function is not declared inline in the class definition. Under
10445 some variants of the ABI, an inline function can never be the key
10446 method. The default is to return @code{true}.
10449 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10450 @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}.
10453 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10454 This hook returns true (the default) if virtual tables and other
10455 similar implicit class data objects are always COMDAT if they have
10456 external linkage. If this hook returns false, then class data for
10457 classes whose virtual table will be emitted in only one translation
10458 unit will not be COMDAT.
10461 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10462 This hook returns true (the default) if the RTTI information for
10463 the basic types which is defined in the C++ runtime should always
10464 be COMDAT, false if it should not be COMDAT.
10467 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10468 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10469 should be used to register static destructors when @option{-fuse-cxa-atexit}
10470 is in effect. The default is to return false to use @code{__cxa_atexit}.
10473 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10474 This hook returns true if the target @code{atexit} function can be used
10475 in the same manner as @code{__cxa_atexit} to register C++ static
10476 destructors. This requires that @code{atexit}-registered functions in
10477 shared libraries are run in the correct order when the libraries are
10478 unloaded. The default is to return false.
10481 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10482 @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).
10485 @deftypefn {Target Hook} tree TARGET_CXX_DECL_MANGLING_CONTEXT (const_tree @var{decl})
10486 Return target-specific mangling context of @var{decl} or @code{NULL_TREE}.
10489 @node Named Address Spaces
10490 @section Adding support for named address spaces
10491 @cindex named address spaces
10493 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10494 standards committee, @cite{Programming Languages - C - Extensions to
10495 support embedded processors}, specifies a syntax for embedded
10496 processors to specify alternate address spaces. You can configure a
10497 GCC port to support section 5.1 of the draft report to add support for
10498 address spaces other than the default address space. These address
10499 spaces are new keywords that are similar to the @code{volatile} and
10500 @code{const} type attributes.
10502 Pointers to named address spaces can have a different size than
10503 pointers to the generic address space.
10505 For example, the SPU port uses the @code{__ea} address space to refer
10506 to memory in the host processor, rather than memory local to the SPU
10507 processor. Access to memory in the @code{__ea} address space involves
10508 issuing DMA operations to move data between the host processor and the
10509 local processor memory address space. Pointers in the @code{__ea}
10510 address space are either 32 bits or 64 bits based on the
10511 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10514 Internally, address spaces are represented as a small integer in the
10515 range 0 to 15 with address space 0 being reserved for the generic
10518 To register a named address space qualifier keyword with the C front end,
10519 the target may call the @code{c_register_addr_space} routine. For example,
10520 the SPU port uses the following to declare @code{__ea} as the keyword for
10521 named address space #1:
10523 #define ADDR_SPACE_EA 1
10524 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10527 @deftypefn {Target Hook} scalar_int_mode TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10528 Define this to return the machine mode to use for pointers to
10529 @var{address_space} if the target supports named address spaces.
10530 The default version of this hook returns @code{ptr_mode}.
10533 @deftypefn {Target Hook} scalar_int_mode TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10534 Define this to return the machine mode to use for addresses in
10535 @var{address_space} if the target supports named address spaces.
10536 The default version of this hook returns @code{Pmode}.
10539 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (scalar_int_mode @var{mode}, addr_space_t @var{as})
10540 Define this to return nonzero if the port can handle pointers
10541 with machine mode @var{mode} to address space @var{as}. This target
10542 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10543 except that it includes explicit named address space support. The default
10544 version of this hook returns true for the modes returned by either the
10545 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10546 target hooks for the given address space.
10549 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P (machine_mode @var{mode}, rtx @var{exp}, bool @var{strict}, addr_space_t @var{as})
10550 Define this to return true if @var{exp} is a valid address for mode
10551 @var{mode} in the named address space @var{as}. The @var{strict}
10552 parameter says whether strict addressing is in effect after reload has
10553 finished. This target hook is the same as the
10554 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10555 explicit named address space support.
10558 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode}, addr_space_t @var{as})
10559 Define this to modify an invalid address @var{x} to be a valid address
10560 with mode @var{mode} in the named address space @var{as}. This target
10561 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10562 except that it includes explicit named address space support.
10565 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{subset}, addr_space_t @var{superset})
10566 Define this to return whether the @var{subset} named address space is
10567 contained within the @var{superset} named address space. Pointers to
10568 a named address space that is a subset of another named address space
10569 will be converted automatically without a cast if used together in
10570 arithmetic operations. Pointers to a superset address space can be
10571 converted to pointers to a subset address space via explicit casts.
10574 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID (addr_space_t @var{as})
10575 Define this to modify the default handling of address 0 for the
10576 address space. Return true if 0 should be considered a valid address.
10579 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10580 Define this to convert the pointer expression represented by the RTL
10581 @var{op} with type @var{from_type} that points to a named address
10582 space to a new pointer expression with type @var{to_type} that points
10583 to a different named address space. When this hook it called, it is
10584 guaranteed that one of the two address spaces is a subset of the other,
10585 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10588 @deftypefn {Target Hook} int TARGET_ADDR_SPACE_DEBUG (addr_space_t @var{as})
10589 Define this to define how the address space is encoded in dwarf.
10590 The result is the value to be used with @code{DW_AT_address_class}.
10593 @deftypefn {Target Hook} void TARGET_ADDR_SPACE_DIAGNOSE_USAGE (addr_space_t @var{as}, location_t @var{loc})
10594 Define this hook if the availability of an address space depends on
10595 command line options and some diagnostics should be printed when the
10596 address space is used. This hook is called during parsing and allows
10597 to emit a better diagnostic compared to the case where the address space
10598 was not registered with @code{c_register_addr_space}. @var{as} is
10599 the address space as registered with @code{c_register_addr_space}.
10600 @var{loc} is the location of the address space qualifier token.
10601 The default implementation does nothing.
10605 @section Miscellaneous Parameters
10606 @cindex parameters, miscellaneous
10608 @c prevent bad page break with this line
10609 Here are several miscellaneous parameters.
10611 @defmac HAS_LONG_COND_BRANCH
10612 Define this boolean macro to indicate whether or not your architecture
10613 has conditional branches that can span all of memory. It is used in
10614 conjunction with an optimization that partitions hot and cold basic
10615 blocks into separate sections of the executable. If this macro is
10616 set to false, gcc will convert any conditional branches that attempt
10617 to cross between sections into unconditional branches or indirect jumps.
10620 @defmac HAS_LONG_UNCOND_BRANCH
10621 Define this boolean macro to indicate whether or not your architecture
10622 has unconditional branches that can span all of memory. It is used in
10623 conjunction with an optimization that partitions hot and cold basic
10624 blocks into separate sections of the executable. If this macro is
10625 set to false, gcc will convert any unconditional branches that attempt
10626 to cross between sections into indirect jumps.
10629 @defmac CASE_VECTOR_MODE
10630 An alias for a machine mode name. This is the machine mode that
10631 elements of a jump-table should have.
10634 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10635 Optional: return the preferred mode for an @code{addr_diff_vec}
10636 when the minimum and maximum offset are known. If you define this,
10637 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10638 To make this work, you also have to define @code{INSN_ALIGN} and
10639 make the alignment for @code{addr_diff_vec} explicit.
10640 The @var{body} argument is provided so that the offset_unsigned and scale
10641 flags can be updated.
10644 @defmac CASE_VECTOR_PC_RELATIVE
10645 Define this macro to be a C expression to indicate when jump-tables
10646 should contain relative addresses. You need not define this macro if
10647 jump-tables never contain relative addresses, or jump-tables should
10648 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10652 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10653 This function return the smallest number of different values for which it
10654 is best to use a jump-table instead of a tree of conditional branches.
10655 The default is four for machines with a @code{casesi} instruction and
10656 five otherwise. This is best for most machines.
10659 @defmac WORD_REGISTER_OPERATIONS
10660 Define this macro to 1 if operations between registers with integral mode
10661 smaller than a word are always performed on the entire register.
10662 Most RISC machines have this property and most CISC machines do not.
10665 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_ARITHMETIC_PRECISION (void)
10666 On some RISC architectures with 64-bit registers, the processor also
10667 maintains 32-bit condition codes that make it possible to do real 32-bit
10668 arithmetic, although the operations are performed on the full registers.
10670 On such architectures, defining this hook to 32 tells the compiler to try
10671 using 32-bit arithmetical operations setting the condition codes instead
10672 of doing full 64-bit arithmetic.
10674 More generally, define this hook on RISC architectures if you want the
10675 compiler to try using arithmetical operations setting the condition codes
10676 with a precision lower than the word precision.
10678 You need not define this hook if @code{WORD_REGISTER_OPERATIONS} is not
10682 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10683 Define this macro to be a C expression indicating when insns that read
10684 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10685 bits outside of @var{mem_mode} to be either the sign-extension or the
10686 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10687 of @var{mem_mode} for which the
10688 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10689 @code{UNKNOWN} for other modes.
10691 This macro is not called with @var{mem_mode} non-integral or with a width
10692 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10693 value in this case. Do not define this macro if it would always return
10694 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10695 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10697 You may return a non-@code{UNKNOWN} value even if for some hard registers
10698 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10699 of these hard registers @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false
10700 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10701 integral mode larger than this but not larger than @code{word_mode}.
10703 You must return @code{UNKNOWN} if for some hard registers that allow this
10704 mode, @code{TARGET_CAN_CHANGE_MODE_CLASS} says that they cannot change to
10705 @code{word_mode}, but that they can change to another integral mode that
10706 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10709 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10710 Define this macro to 1 if loading short immediate values into registers sign
10714 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (machine_mode @var{mode})
10715 When @option{-ffast-math} is in effect, GCC tries to optimize
10716 divisions by the same divisor, by turning them into multiplications by
10717 the reciprocal. This target hook specifies the minimum number of divisions
10718 that should be there for GCC to perform the optimization for a variable
10719 of mode @var{mode}. The default implementation returns 3 if the machine
10720 has an instruction for the division, and 2 if it does not.
10724 The maximum number of bytes that a single instruction can move quickly
10725 between memory and registers or between two memory locations.
10728 @defmac MAX_MOVE_MAX
10729 The maximum number of bytes that a single instruction can move quickly
10730 between memory and registers or between two memory locations. If this
10731 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10732 constant value that is the largest value that @code{MOVE_MAX} can have
10736 @defmac SHIFT_COUNT_TRUNCATED
10737 A C expression that is nonzero if on this machine the number of bits
10738 actually used for the count of a shift operation is equal to the number
10739 of bits needed to represent the size of the object being shifted. When
10740 this macro is nonzero, the compiler will assume that it is safe to omit
10741 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10742 truncates the count of a shift operation. On machines that have
10743 instructions that act on bit-fields at variable positions, which may
10744 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10745 also enables deletion of truncations of the values that serve as
10746 arguments to bit-field instructions.
10748 If both types of instructions truncate the count (for shifts) and
10749 position (for bit-field operations), or if no variable-position bit-field
10750 instructions exist, you should define this macro.
10752 However, on some machines, such as the 80386 and the 680x0, truncation
10753 only applies to shift operations and not the (real or pretended)
10754 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10755 such machines. Instead, add patterns to the @file{md} file that include
10756 the implied truncation of the shift instructions.
10758 You need not define this macro if it would always have the value of zero.
10761 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10762 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (machine_mode @var{mode})
10763 This function describes how the standard shift patterns for @var{mode}
10764 deal with shifts by negative amounts or by more than the width of the mode.
10765 @xref{shift patterns}.
10767 On many machines, the shift patterns will apply a mask @var{m} to the
10768 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10769 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10770 this is true for mode @var{mode}, the function should return @var{m},
10771 otherwise it should return 0. A return value of 0 indicates that no
10772 particular behavior is guaranteed.
10774 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10775 @emph{not} apply to general shift rtxes; it applies only to instructions
10776 that are generated by the named shift patterns.
10778 The default implementation of this function returns
10779 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10780 and 0 otherwise. This definition is always safe, but if
10781 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10782 nevertheless truncate the shift count, you may get better code
10786 @deftypefn {Target Hook} bool TARGET_TRULY_NOOP_TRUNCATION (unsigned int @var{outprec}, unsigned int @var{inprec})
10787 This hook returns true if it is safe to ``convert'' a value of
10788 @var{inprec} bits to one of @var{outprec} bits (where @var{outprec} is
10789 smaller than @var{inprec}) by merely operating on it as if it had only
10790 @var{outprec} bits. The default returns true unconditionally, which
10791 is correct for most machines.
10793 If @code{TARGET_MODES_TIEABLE_P} returns false for a pair of modes,
10794 suboptimal code can result if this hook returns true for the corresponding
10795 mode sizes. Making this hook return false in such cases may improve things.
10798 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (scalar_int_mode @var{mode}, scalar_int_mode @var{rep_mode})
10799 The representation of an integral mode can be such that the values
10800 are always extended to a wider integral mode. Return
10801 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10802 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10803 otherwise. (Currently, none of the targets use zero-extended
10804 representation this way so unlike @code{LOAD_EXTEND_OP},
10805 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10806 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10807 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10808 widest integral mode and currently we take advantage of this fact.)
10810 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10811 value even if the extension is not performed on certain hard registers
10812 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10813 @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false.
10815 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10816 describe two related properties. If you define
10817 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10818 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10821 In order to enforce the representation of @code{mode},
10822 @code{TARGET_TRULY_NOOP_TRUNCATION} should return false when truncating to
10826 @defmac STORE_FLAG_VALUE
10827 A C expression describing the value returned by a comparison operator
10828 with an integral mode and stored by a store-flag instruction
10829 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10830 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10831 comparison operators whose results have a @code{MODE_INT} mode.
10833 A value of 1 or @minus{}1 means that the instruction implementing the
10834 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10835 and 0 when the comparison is false. Otherwise, the value indicates
10836 which bits of the result are guaranteed to be 1 when the comparison is
10837 true. This value is interpreted in the mode of the comparison
10838 operation, which is given by the mode of the first operand in the
10839 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10840 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10843 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10844 generate code that depends only on the specified bits. It can also
10845 replace comparison operators with equivalent operations if they cause
10846 the required bits to be set, even if the remaining bits are undefined.
10847 For example, on a machine whose comparison operators return an
10848 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10849 @samp{0x80000000}, saying that just the sign bit is relevant, the
10853 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10857 can be converted to
10860 (ashift:SI @var{x} (const_int @var{n}))
10864 where @var{n} is the appropriate shift count to move the bit being
10865 tested into the sign bit.
10867 There is no way to describe a machine that always sets the low-order bit
10868 for a true value, but does not guarantee the value of any other bits,
10869 but we do not know of any machine that has such an instruction. If you
10870 are trying to port GCC to such a machine, include an instruction to
10871 perform a logical-and of the result with 1 in the pattern for the
10872 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10874 Often, a machine will have multiple instructions that obtain a value
10875 from a comparison (or the condition codes). Here are rules to guide the
10876 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10881 Use the shortest sequence that yields a valid definition for
10882 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10883 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10884 comparison operators to do so because there may be opportunities to
10885 combine the normalization with other operations.
10888 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10889 slightly preferred on machines with expensive jumps and 1 preferred on
10893 As a second choice, choose a value of @samp{0x80000001} if instructions
10894 exist that set both the sign and low-order bits but do not define the
10898 Otherwise, use a value of @samp{0x80000000}.
10901 Many machines can produce both the value chosen for
10902 @code{STORE_FLAG_VALUE} and its negation in the same number of
10903 instructions. On those machines, you should also define a pattern for
10904 those cases, e.g., one matching
10907 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10910 Some machines can also perform @code{and} or @code{plus} operations on
10911 condition code values with less instructions than the corresponding
10912 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10913 machines, define the appropriate patterns. Use the names @code{incscc}
10914 and @code{decscc}, respectively, for the patterns which perform
10915 @code{plus} or @code{minus} operations on condition code values. See
10916 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10917 find such instruction sequences on other machines.
10919 If this macro is not defined, the default value, 1, is used. You need
10920 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10921 instructions, or if the value generated by these instructions is 1.
10924 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10925 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10926 returned when comparison operators with floating-point results are true.
10927 Define this macro on machines that have comparison operations that return
10928 floating-point values. If there are no such operations, do not define
10932 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10933 A C expression that gives a rtx representing the nonzero true element
10934 for vector comparisons. The returned rtx should be valid for the inner
10935 mode of @var{mode} which is guaranteed to be a vector mode. Define
10936 this macro on machines that have vector comparison operations that
10937 return a vector result. If there are no such operations, do not define
10938 this macro. Typically, this macro is defined as @code{const1_rtx} or
10939 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10940 the compiler optimizing such vector comparison operations for the
10944 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10945 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10946 A C expression that indicates whether the architecture defines a value
10947 for @code{clz} or @code{ctz} with a zero operand.
10948 A result of @code{0} indicates the value is undefined.
10949 If the value is defined for only the RTL expression, the macro should
10950 evaluate to @code{1}; if the value applies also to the corresponding optab
10951 entry (which is normally the case if it expands directly into
10952 the corresponding RTL), then the macro should evaluate to @code{2}.
10953 In the cases where the value is defined, @var{value} should be set to
10956 If this macro is not defined, the value of @code{clz} or
10957 @code{ctz} at zero is assumed to be undefined.
10959 This macro must be defined if the target's expansion for @code{ffs}
10960 relies on a particular value to get correct results. Otherwise it
10961 is not necessary, though it may be used to optimize some corner cases, and
10962 to provide a default expansion for the @code{ffs} optab.
10964 Note that regardless of this macro the ``definedness'' of @code{clz}
10965 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10966 visible to the user. Thus one may be free to adjust the value at will
10967 to match the target expansion of these operations without fear of
10972 An alias for the machine mode for pointers. On most machines, define
10973 this to be the integer mode corresponding to the width of a hardware
10974 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10975 On some machines you must define this to be one of the partial integer
10976 modes, such as @code{PSImode}.
10978 The width of @code{Pmode} must be at least as large as the value of
10979 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10980 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10984 @defmac FUNCTION_MODE
10985 An alias for the machine mode used for memory references to functions
10986 being called, in @code{call} RTL expressions. On most CISC machines,
10987 where an instruction can begin at any byte address, this should be
10988 @code{QImode}. On most RISC machines, where all instructions have fixed
10989 size and alignment, this should be a mode with the same size and alignment
10990 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10993 @defmac STDC_0_IN_SYSTEM_HEADERS
10994 In normal operation, the preprocessor expands @code{__STDC__} to the
10995 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10996 hosts, like Solaris, the system compiler uses a different convention,
10997 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10998 strict conformance to the C Standard.
11000 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
11001 convention when processing system header files, but when processing user
11002 files @code{__STDC__} will always expand to 1.
11005 @deftypefn {C Target Hook} {const char *} TARGET_C_PREINCLUDE (void)
11006 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.
11008 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.
11011 @deftypefn {C Target Hook} bool TARGET_CXX_IMPLICIT_EXTERN_C (const char*@var{})
11012 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.
11015 @defmac NO_IMPLICIT_EXTERN_C
11016 Define this macro if the system header files support C++ as well as C@.
11017 This macro inhibits the usual method of using system header files in
11018 C++, which is to pretend that the file's contents are enclosed in
11019 @samp{extern "C" @{@dots{}@}}.
11024 @defmac REGISTER_TARGET_PRAGMAS ()
11025 Define this macro if you want to implement any target-specific pragmas.
11026 If defined, it is a C expression which makes a series of calls to
11027 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
11028 for each pragma. The macro may also do any
11029 setup required for the pragmas.
11031 The primary reason to define this macro is to provide compatibility with
11032 other compilers for the same target. In general, we discourage
11033 definition of target-specific pragmas for GCC@.
11035 If the pragma can be implemented by attributes then you should consider
11036 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
11038 Preprocessor macros that appear on pragma lines are not expanded. All
11039 @samp{#pragma} directives that do not match any registered pragma are
11040 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
11043 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
11044 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
11046 Each call to @code{c_register_pragma} or
11047 @code{c_register_pragma_with_expansion} establishes one pragma. The
11048 @var{callback} routine will be called when the preprocessor encounters a
11052 #pragma [@var{space}] @var{name} @dots{}
11055 @var{space} is the case-sensitive namespace of the pragma, or
11056 @code{NULL} to put the pragma in the global namespace. The callback
11057 routine receives @var{pfile} as its first argument, which can be passed
11058 on to cpplib's functions if necessary. You can lex tokens after the
11059 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
11060 callback will be silently ignored. The end of the line is indicated by
11061 a token of type @code{CPP_EOF}. Macro expansion occurs on the
11062 arguments of pragmas registered with
11063 @code{c_register_pragma_with_expansion} but not on the arguments of
11064 pragmas registered with @code{c_register_pragma}.
11066 Note that the use of @code{pragma_lex} is specific to the C and C++
11067 compilers. It will not work in the Java or Fortran compilers, or any
11068 other language compilers for that matter. Thus if @code{pragma_lex} is going
11069 to be called from target-specific code, it must only be done so when
11070 building the C and C++ compilers. This can be done by defining the
11071 variables @code{c_target_objs} and @code{cxx_target_objs} in the
11072 target entry in the @file{config.gcc} file. These variables should name
11073 the target-specific, language-specific object file which contains the
11074 code that uses @code{pragma_lex}. Note it will also be necessary to add a
11075 rule to the makefile fragment pointed to by @code{tmake_file} that shows
11076 how to build this object file.
11079 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
11080 Define this macro if macros should be expanded in the
11081 arguments of @samp{#pragma pack}.
11084 @defmac TARGET_DEFAULT_PACK_STRUCT
11085 If your target requires a structure packing default other than 0 (meaning
11086 the machine default), define this macro to the necessary value (in bytes).
11087 This must be a value that would also be valid to use with
11088 @samp{#pragma pack()} (that is, a small power of two).
11091 @defmac DOLLARS_IN_IDENTIFIERS
11092 Define this macro to control use of the character @samp{$} in
11093 identifier names for the C family of languages. 0 means @samp{$} is
11094 not allowed by default; 1 means it is allowed. 1 is the default;
11095 there is no need to define this macro in that case.
11098 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
11099 Define this macro as a C expression that is nonzero if it is safe for the
11100 delay slot scheduler to place instructions in the delay slot of @var{insn},
11101 even if they appear to use a resource set or clobbered in @var{insn}.
11102 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
11103 every @code{call_insn} has this behavior. On machines where some @code{insn}
11104 or @code{jump_insn} is really a function call and hence has this behavior,
11105 you should define this macro.
11107 You need not define this macro if it would always return zero.
11110 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
11111 Define this macro as a C expression that is nonzero if it is safe for the
11112 delay slot scheduler to place instructions in the delay slot of @var{insn},
11113 even if they appear to set or clobber a resource referenced in @var{insn}.
11114 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
11115 some @code{insn} or @code{jump_insn} is really a function call and its operands
11116 are registers whose use is actually in the subroutine it calls, you should
11117 define this macro. Doing so allows the delay slot scheduler to move
11118 instructions which copy arguments into the argument registers into the delay
11119 slot of @var{insn}.
11121 You need not define this macro if it would always return zero.
11124 @defmac MULTIPLE_SYMBOL_SPACES
11125 Define this macro as a C expression that is nonzero if, in some cases,
11126 global symbols from one translation unit may not be bound to undefined
11127 symbols in another translation unit without user intervention. For
11128 instance, under Microsoft Windows symbols must be explicitly imported
11129 from shared libraries (DLLs).
11131 You need not define this macro if it would always evaluate to zero.
11134 @deftypefn {Target Hook} {rtx_insn *} TARGET_MD_ASM_ADJUST (vec<rtx>& @var{outputs}, vec<rtx>& @var{inputs}, vec<const char *>& @var{constraints}, vec<rtx>& @var{clobbers}, HARD_REG_SET& @var{clobbered_regs})
11135 This target hook may add @dfn{clobbers} to @var{clobbers} and
11136 @var{clobbered_regs} for any hard regs the port wishes to automatically
11137 clobber for an asm. The @var{outputs} and @var{inputs} may be inspected
11138 to avoid clobbering a register that is already used by the asm.
11140 It may modify the @var{outputs}, @var{inputs}, and @var{constraints}
11141 as necessary for other pre-processing. In this case the return value is
11142 a sequence of insns to emit after the asm.
11145 @defmac MATH_LIBRARY
11146 Define this macro as a C string constant for the linker argument to link
11147 in the system math library, minus the initial @samp{"-l"}, or
11148 @samp{""} if the target does not have a
11149 separate math library.
11151 You need only define this macro if the default of @samp{"m"} is wrong.
11154 @defmac LIBRARY_PATH_ENV
11155 Define this macro as a C string constant for the environment variable that
11156 specifies where the linker should look for libraries.
11158 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
11162 @defmac TARGET_POSIX_IO
11163 Define this macro if the target supports the following POSIX@ file
11164 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
11165 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
11166 to use file locking when exiting a program, which avoids race conditions
11167 if the program has forked. It will also create directories at run-time
11168 for cross-profiling.
11171 @defmac MAX_CONDITIONAL_EXECUTE
11173 A C expression for the maximum number of instructions to execute via
11174 conditional execution instructions instead of a branch. A value of
11175 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
11176 1 if it does use cc0.
11179 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
11180 Used if the target needs to perform machine-dependent modifications on the
11181 conditionals used for turning basic blocks into conditionally executed code.
11182 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
11183 contains information about the currently processed blocks. @var{true_expr}
11184 and @var{false_expr} are the tests that are used for converting the
11185 then-block and the else-block, respectively. Set either @var{true_expr} or
11186 @var{false_expr} to a null pointer if the tests cannot be converted.
11189 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
11190 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
11191 if-statements into conditions combined by @code{and} and @code{or} operations.
11192 @var{bb} contains the basic block that contains the test that is currently
11193 being processed and about to be turned into a condition.
11196 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
11197 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
11198 be converted to conditional execution format. @var{ce_info} points to
11199 a data structure, @code{struct ce_if_block}, which contains information
11200 about the currently processed blocks.
11203 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
11204 A C expression to perform any final machine dependent modifications in
11205 converting code to conditional execution. The involved basic blocks
11206 can be found in the @code{struct ce_if_block} structure that is pointed
11207 to by @var{ce_info}.
11210 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
11211 A C expression to cancel any machine dependent modifications in
11212 converting code to conditional execution. The involved basic blocks
11213 can be found in the @code{struct ce_if_block} structure that is pointed
11214 to by @var{ce_info}.
11217 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
11218 A C expression to initialize any machine specific data for if-conversion
11219 of the if-block in the @code{struct ce_if_block} structure that is pointed
11220 to by @var{ce_info}.
11223 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
11224 If non-null, this hook performs a target-specific pass over the
11225 instruction stream. The compiler will run it at all optimization levels,
11226 just before the point at which it normally does delayed-branch scheduling.
11228 The exact purpose of the hook varies from target to target. Some use
11229 it to do transformations that are necessary for correctness, such as
11230 laying out in-function constant pools or avoiding hardware hazards.
11231 Others use it as an opportunity to do some machine-dependent optimizations.
11233 You need not implement the hook if it has nothing to do. The default
11234 definition is null.
11237 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
11238 Define this hook if you have any machine-specific built-in functions
11239 that need to be defined. It should be a function that performs the
11242 Machine specific built-in functions can be useful to expand special machine
11243 instructions that would otherwise not normally be generated because
11244 they have no equivalent in the source language (for example, SIMD vector
11245 instructions or prefetch instructions).
11247 To create a built-in function, call the function
11248 @code{lang_hooks.builtin_function}
11249 which is defined by the language front end. You can use any type nodes set
11250 up by @code{build_common_tree_nodes};
11251 only language front ends that use those two functions will call
11252 @samp{TARGET_INIT_BUILTINS}.
11255 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
11256 Define this hook if you have any machine-specific built-in functions
11257 that need to be defined. It should be a function that returns the
11258 builtin function declaration for the builtin function code @var{code}.
11259 If there is no such builtin and it cannot be initialized at this time
11260 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
11261 If @var{code} is out of range the function should return
11262 @code{error_mark_node}.
11265 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, machine_mode @var{mode}, int @var{ignore})
11267 Expand a call to a machine specific built-in function that was set up by
11268 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
11269 function call; the result should go to @var{target} if that is
11270 convenient, and have mode @var{mode} if that is convenient.
11271 @var{subtarget} may be used as the target for computing one of
11272 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
11273 ignored. This function should return the result of the call to the
11277 @deftypefn {Target Hook} tree TARGET_BUILTIN_CHKP_FUNCTION (unsigned @var{fcode})
11278 This hook allows target to redefine built-in functions used by
11279 Pointer Bounds Checker for code instrumentation. Hook should return
11280 fndecl of function implementing generic builtin whose code is
11281 passed in @var{fcode}. Currently following built-in functions are
11282 obtained using this hook:
11283 @deftypefn {Built-in Function} __bounds_type __chkp_bndmk (const void *@var{lb}, size_t @var{size})
11284 Function code - BUILT_IN_CHKP_BNDMK. This built-in function is used
11285 by Pointer Bounds Checker to create bound values. @var{lb} holds low
11286 bound of the resulting bounds. @var{size} holds size of created bounds.
11289 @deftypefn {Built-in Function} void __chkp_bndstx (const void *@var{ptr}, __bounds_type @var{b}, const void **@var{loc})
11290 Function code - @code{BUILT_IN_CHKP_BNDSTX}. This built-in function is used
11291 by Pointer Bounds Checker to store bounds @var{b} for pointer @var{ptr}
11292 when @var{ptr} is stored by address @var{loc}.
11295 @deftypefn {Built-in Function} __bounds_type __chkp_bndldx (const void **@var{loc}, const void *@var{ptr})
11296 Function code - @code{BUILT_IN_CHKP_BNDLDX}. This built-in function is used
11297 by Pointer Bounds Checker to get bounds of pointer @var{ptr} loaded by
11301 @deftypefn {Built-in Function} void __chkp_bndcl (const void *@var{ptr}, __bounds_type @var{b})
11302 Function code - @code{BUILT_IN_CHKP_BNDCL}. This built-in function is used
11303 by Pointer Bounds Checker to perform check for pointer @var{ptr} against
11304 lower bound of bounds @var{b}.
11307 @deftypefn {Built-in Function} void __chkp_bndcu (const void *@var{ptr}, __bounds_type @var{b})
11308 Function code - @code{BUILT_IN_CHKP_BNDCU}. This built-in function is used
11309 by Pointer Bounds Checker to perform check for pointer @var{ptr} against
11310 upper bound of bounds @var{b}.
11313 @deftypefn {Built-in Function} __bounds_type __chkp_bndret (void *@var{ptr})
11314 Function code - @code{BUILT_IN_CHKP_BNDRET}. This built-in function is used
11315 by Pointer Bounds Checker to obtain bounds returned by a call statement.
11316 @var{ptr} passed to built-in is @code{SSA_NAME} returned by the call.
11319 @deftypefn {Built-in Function} __bounds_type __chkp_intersect (__bounds_type @var{b1}, __bounds_type @var{b2})
11320 Function code - @code{BUILT_IN_CHKP_INTERSECT}. This built-in function
11321 returns intersection of bounds @var{b1} and @var{b2}.
11324 @deftypefn {Built-in Function} __bounds_type __chkp_narrow (const void *@var{ptr}, __bounds_type @var{b}, size_t @var{s})
11325 Function code - @code{BUILT_IN_CHKP_NARROW}. This built-in function
11326 returns intersection of bounds @var{b} and
11327 [@var{ptr}, @var{ptr} + @var{s} - @code{1}].
11330 @deftypefn {Built-in Function} size_t __chkp_sizeof (const void *@var{ptr})
11331 Function code - @code{BUILT_IN_CHKP_SIZEOF}. This built-in function
11332 returns size of object referenced by @var{ptr}. @var{ptr} is always
11333 @code{ADDR_EXPR} of @code{VAR_DECL}. This built-in is used by
11334 Pointer Bounds Checker when bounds of object cannot be computed statically
11335 (e.g. object has incomplete type).
11338 @deftypefn {Built-in Function} const void *__chkp_extract_lower (__bounds_type @var{b})
11339 Function code - @code{BUILT_IN_CHKP_EXTRACT_LOWER}. This built-in function
11340 returns lower bound of bounds @var{b}.
11343 @deftypefn {Built-in Function} const void *__chkp_extract_upper (__bounds_type @var{b})
11344 Function code - @code{BUILT_IN_CHKP_EXTRACT_UPPER}. This built-in function
11345 returns upper bound of bounds @var{b}.
11348 @deftypefn {Target Hook} tree TARGET_CHKP_BOUND_TYPE (void)
11349 Return type to be used for bounds
11351 @deftypefn {Target Hook} machine_mode TARGET_CHKP_BOUND_MODE (void)
11352 Return mode to be used for bounds.
11354 @deftypefn {Target Hook} tree TARGET_CHKP_MAKE_BOUNDS_CONSTANT (HOST_WIDE_INT @var{lb}, HOST_WIDE_INT @var{ub})
11355 Return constant used to statically initialize constant bounds
11356 with specified lower bound @var{lb} and upper bounds @var{ub}.
11358 @deftypefn {Target Hook} int TARGET_CHKP_INITIALIZE_BOUNDS (tree @var{var}, tree @var{lb}, tree @var{ub}, tree *@var{stmts})
11359 Generate a list of statements @var{stmts} to initialize pointer
11360 bounds variable @var{var} with bounds @var{lb} and @var{ub}. Return
11361 the number of generated statements.
11364 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
11365 Select a replacement for a machine specific built-in function that
11366 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
11367 @emph{before} regular type checking, and so allows the target to
11368 implement a crude form of function overloading. @var{fndecl} is the
11369 declaration of the built-in function. @var{arglist} is the list of
11370 arguments passed to the built-in function. The result is a
11371 complete expression that implements the operation, usually
11372 another @code{CALL_EXPR}.
11373 @var{arglist} really has type @samp{VEC(tree,gc)*}
11376 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
11377 Fold a call to a machine specific built-in function that was set up by
11378 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
11379 built-in function. @var{n_args} is the number of arguments passed to
11380 the function; the arguments themselves are pointed to by @var{argp}.
11381 The result is another tree, valid for both GIMPLE and GENERIC,
11382 containing a simplified expression for the call's result. If
11383 @var{ignore} is true the value will be ignored.
11386 @deftypefn {Target Hook} bool TARGET_GIMPLE_FOLD_BUILTIN (gimple_stmt_iterator *@var{gsi})
11387 Fold a call to a machine specific built-in function that was set up
11388 by @samp{TARGET_INIT_BUILTINS}. @var{gsi} points to the gimple
11389 statement holding the function call. Returns true if any change
11390 was made to the GIMPLE stream.
11393 @deftypefn {Target Hook} int TARGET_COMPARE_VERSION_PRIORITY (tree @var{decl1}, tree @var{decl2})
11394 This hook is used to compare the target attributes in two functions to
11395 determine which function's features get higher priority. This is used
11396 during function multi-versioning to figure out the order in which two
11397 versions must be dispatched. A function version with a higher priority
11398 is checked for dispatching earlier. @var{decl1} and @var{decl2} are
11399 the two function decls that will be compared.
11402 @deftypefn {Target Hook} tree TARGET_GET_FUNCTION_VERSIONS_DISPATCHER (void *@var{decl})
11403 This hook is used to get the dispatcher function for a set of function
11404 versions. The dispatcher function is called to invoke the right function
11405 version at run-time. @var{decl} is one version from a set of semantically
11406 identical versions.
11409 @deftypefn {Target Hook} tree TARGET_GENERATE_VERSION_DISPATCHER_BODY (void *@var{arg})
11410 This hook is used to generate the dispatcher logic to invoke the right
11411 function version at run-time for a given set of function versions.
11412 @var{arg} points to the callgraph node of the dispatcher function whose
11413 body must be generated.
11416 @deftypefn {Target Hook} bool TARGET_CAN_USE_DOLOOP_P (const widest_int @var{&iterations}, const widest_int @var{&iterations_max}, unsigned int @var{loop_depth}, bool @var{entered_at_top})
11417 Return true if it is possible to use low-overhead loops (@code{doloop_end}
11418 and @code{doloop_begin}) for a particular loop. @var{iterations} gives the
11419 exact number of iterations, or 0 if not known. @var{iterations_max} gives
11420 the maximum number of iterations, or 0 if not known. @var{loop_depth} is
11421 the nesting depth of the loop, with 1 for innermost loops, 2 for loops that
11422 contain innermost loops, and so on. @var{entered_at_top} is true if the
11423 loop is only entered from the top.
11425 This hook is only used if @code{doloop_end} is available. The default
11426 implementation returns true. You can use @code{can_use_doloop_if_innermost}
11427 if the loop must be the innermost, and if there are no other restrictions.
11430 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const rtx_insn *@var{insn})
11432 Take an instruction in @var{insn} and return NULL if it is valid within a
11433 low-overhead loop, otherwise return a string explaining why doloop
11434 could not be applied.
11436 Many targets use special registers for low-overhead looping. For any
11437 instruction that clobbers these this function should return a string indicating
11438 the reason why the doloop could not be applied.
11439 By default, the RTL loop optimizer does not use a present doloop pattern for
11440 loops containing function calls or branch on table instructions.
11443 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_COMBINED_INSN (rtx_insn *@var{insn})
11444 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.
11447 @deftypefn {Target Hook} bool TARGET_CAN_FOLLOW_JUMP (const rtx_insn *@var{follower}, const rtx_insn *@var{followee})
11448 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.
11451 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
11452 This target hook returns @code{true} if @var{x} is considered to be commutative.
11453 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
11454 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
11455 of the enclosing rtl, if known, otherwise it is UNKNOWN.
11458 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
11460 When the initial value of a hard register has been copied in a pseudo
11461 register, it is often not necessary to actually allocate another register
11462 to this pseudo register, because the original hard register or a stack slot
11463 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
11464 is called at the start of register allocation once for each hard register
11465 that had its initial value copied by using
11466 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
11467 Possible values are @code{NULL_RTX}, if you don't want
11468 to do any special allocation, a @code{REG} rtx---that would typically be
11469 the hard register itself, if it is known not to be clobbered---or a
11471 If you are returning a @code{MEM}, this is only a hint for the allocator;
11472 it might decide to use another register anyways.
11473 You may use @code{current_function_is_leaf} or
11474 @code{REG_N_SETS} in the hook to determine if the hard
11475 register in question will not be clobbered.
11476 The default value of this hook is @code{NULL}, which disables any special
11480 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
11481 This target hook returns nonzero if @var{x}, an @code{unspec} or
11482 @code{unspec_volatile} operation, might cause a trap. Targets can use
11483 this hook to enhance precision of analysis for @code{unspec} and
11484 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
11485 to analyze inner elements of @var{x} in which case @var{flags} should be
11489 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
11490 The compiler invokes this hook whenever it changes its current function
11491 context (@code{cfun}). You can define this function if
11492 the back end needs to perform any initialization or reset actions on a
11493 per-function basis. For example, it may be used to implement function
11494 attributes that affect register usage or code generation patterns.
11495 The argument @var{decl} is the declaration for the new function context,
11496 and may be null to indicate that the compiler has left a function context
11497 and is returning to processing at the top level.
11498 The default hook function does nothing.
11500 GCC sets @code{cfun} to a dummy function context during initialization of
11501 some parts of the back end. The hook function is not invoked in this
11502 situation; you need not worry about the hook being invoked recursively,
11503 or when the back end is in a partially-initialized state.
11504 @code{cfun} might be @code{NULL} to indicate processing at top level,
11505 outside of any function scope.
11508 @defmac TARGET_OBJECT_SUFFIX
11509 Define this macro to be a C string representing the suffix for object
11510 files on your target machine. If you do not define this macro, GCC will
11511 use @samp{.o} as the suffix for object files.
11514 @defmac TARGET_EXECUTABLE_SUFFIX
11515 Define this macro to be a C string representing the suffix to be
11516 automatically added to executable files on your target machine. If you
11517 do not define this macro, GCC will use the null string as the suffix for
11521 @defmac COLLECT_EXPORT_LIST
11522 If defined, @code{collect2} will scan the individual object files
11523 specified on its command line and create an export list for the linker.
11524 Define this macro for systems like AIX, where the linker discards
11525 object files that are not referenced from @code{main} and uses export
11529 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11530 Define this macro to a C expression representing a variant of the
11531 method call @var{mdecl}, if Java Native Interface (JNI) methods
11532 must be invoked differently from other methods on your target.
11533 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11534 the @code{stdcall} calling convention and this macro is then
11535 defined as this expression:
11538 build_type_attribute_variant (@var{mdecl},
11540 (get_identifier ("stdcall"),
11545 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11546 This target hook returns @code{true} past the point in which new jump
11547 instructions could be created. On machines that require a register for
11548 every jump such as the SHmedia ISA of SH5, this point would typically be
11549 reload, so this target hook should be defined to a function such as:
11553 cannot_modify_jumps_past_reload_p ()
11555 return (reload_completed || reload_in_progress);
11560 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11561 This target hook returns a register class for which branch target register
11562 optimizations should be applied. All registers in this class should be
11563 usable interchangeably. After reload, registers in this class will be
11564 re-allocated and loads will be hoisted out of loops and be subjected
11565 to inter-block scheduling.
11568 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11569 Branch target register optimization will by default exclude callee-saved
11571 that are not already live during the current function; if this target hook
11572 returns true, they will be included. The target code must than make sure
11573 that all target registers in the class returned by
11574 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11575 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11576 epilogues have already been generated. Note, even if you only return
11577 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11578 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11579 to reserve space for caller-saved target registers.
11582 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11583 This target hook returns true if the target supports conditional execution.
11584 This target hook is required only when the target has several different
11585 modes and they have different conditional execution capability, such as ARM.
11588 @deftypefn {Target Hook} rtx TARGET_GEN_CCMP_FIRST (rtx_insn **@var{prep_seq}, rtx_insn **@var{gen_seq}, int @var{code}, tree @var{op0}, tree @var{op1})
11589 This function prepares to emit a comparison insn for the first compare in a
11590 sequence of conditional comparisions. It returns an appropriate comparison
11591 with @code{CC} for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11592 The insns to prepare the compare are saved in @var{prep_seq} and the compare
11593 insns are saved in @var{gen_seq}. They will be emitted when all the
11594 compares in the the conditional comparision are generated without error.
11595 @var{code} is the @code{rtx_code} of the compare for @var{op0} and @var{op1}.
11598 @deftypefn {Target Hook} rtx TARGET_GEN_CCMP_NEXT (rtx_insn **@var{prep_seq}, rtx_insn **@var{gen_seq}, rtx @var{prev}, int @var{cmp_code}, tree @var{op0}, tree @var{op1}, int @var{bit_code})
11599 This function prepares to emit a conditional comparison within a sequence
11600 of conditional comparisons. It returns an appropriate comparison with
11601 @code{CC} for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11602 The insns to prepare the compare are saved in @var{prep_seq} and the compare
11603 insns are saved in @var{gen_seq}. They will be emitted when all the
11604 compares in the conditional comparision are generated without error. The
11605 @var{prev} expression is the result of a prior call to @code{gen_ccmp_first}
11606 or @code{gen_ccmp_next}. It may return @code{NULL} if the combination of
11607 @var{prev} and this comparison is not supported, otherwise the result must
11608 be appropriate for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11609 @var{code} is the @code{rtx_code} of the compare for @var{op0} and @var{op1}.
11610 @var{bit_code} is @code{AND} or @code{IOR}, which is the op on the compares.
11613 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop})
11614 This target hook returns a new value for the number of times @var{loop}
11615 should be unrolled. The parameter @var{nunroll} is the number of times
11616 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11617 the loop, which is going to be checked for unrolling. This target hook
11618 is required only when the target has special constraints like maximum
11619 number of memory accesses.
11622 @defmac POWI_MAX_MULTS
11623 If defined, this macro is interpreted as a signed integer C expression
11624 that specifies the maximum number of floating point multiplications
11625 that should be emitted when expanding exponentiation by an integer
11626 constant inline. When this value is defined, exponentiation requiring
11627 more than this number of multiplications is implemented by calling the
11628 system library's @code{pow}, @code{powf} or @code{powl} routines.
11629 The default value places no upper bound on the multiplication count.
11632 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11633 This target hook should register any extra include files for the
11634 target. The parameter @var{stdinc} indicates if normal include files
11635 are present. The parameter @var{sysroot} is the system root directory.
11636 The parameter @var{iprefix} is the prefix for the gcc directory.
11639 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11640 This target hook should register any extra include files for the
11641 target before any standard headers. The parameter @var{stdinc}
11642 indicates if normal include files are present. The parameter
11643 @var{sysroot} is the system root directory. The parameter
11644 @var{iprefix} is the prefix for the gcc directory.
11647 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11648 This target hook should register special include paths for the target.
11649 The parameter @var{path} is the include to register. On Darwin
11650 systems, this is used for Framework includes, which have semantics
11651 that are different from @option{-I}.
11654 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11655 This target macro returns @code{true} if it is safe to use a local alias
11656 for a virtual function @var{fndecl} when constructing thunks,
11657 @code{false} otherwise. By default, the macro returns @code{true} for all
11658 functions, if a target supports aliases (i.e.@: defines
11659 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11662 @defmac TARGET_FORMAT_TYPES
11663 If defined, this macro is the name of a global variable containing
11664 target-specific format checking information for the @option{-Wformat}
11665 option. The default is to have no target-specific format checks.
11668 @defmac TARGET_N_FORMAT_TYPES
11669 If defined, this macro is the number of entries in
11670 @code{TARGET_FORMAT_TYPES}.
11673 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11674 If defined, this macro is the name of a global variable containing
11675 target-specific format overrides for the @option{-Wformat} option. The
11676 default is to have no target-specific format overrides. If defined,
11677 @code{TARGET_FORMAT_TYPES} must be defined, too.
11680 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11681 If defined, this macro specifies the number of entries in
11682 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11685 @defmac TARGET_OVERRIDES_FORMAT_INIT
11686 If defined, this macro specifies the optional initialization
11687 routine for target specific customizations of the system printf
11688 and scanf formatter settings.
11691 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11692 If defined, this macro returns the diagnostic message when it is
11693 illegal to pass argument @var{val} to function @var{funcdecl}
11694 with prototype @var{typelist}.
11697 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11698 If defined, this macro returns the diagnostic message when it is
11699 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11700 if validity should be determined by the front end.
11703 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11704 If defined, this macro returns the diagnostic message when it is
11705 invalid to apply operation @var{op} (where unary plus is denoted by
11706 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11707 if validity should be determined by the front end.
11710 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11711 If defined, this macro returns the diagnostic message when it is
11712 invalid to apply operation @var{op} to operands of types @var{type1}
11713 and @var{type2}, or @code{NULL} if validity should be determined by
11717 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11718 If defined, this target hook returns the type to which values of
11719 @var{type} should be promoted when they appear in expressions,
11720 analogous to the integer promotions, or @code{NULL_TREE} to use the
11721 front end's normal promotion rules. This hook is useful when there are
11722 target-specific types with special promotion rules.
11723 This is currently used only by the C and C++ front ends.
11726 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11727 If defined, this hook returns the result of converting @var{expr} to
11728 @var{type}. It should return the converted expression,
11729 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11730 This hook is useful when there are target-specific types with special
11732 This is currently used only by the C and C++ front ends.
11736 This macro determines the size of the objective C jump buffer for the
11737 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11740 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11741 Define this macro if any target-specific attributes need to be attached
11742 to the functions in @file{libgcc} that provide low-level support for
11743 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11744 and the associated definitions of those functions.
11747 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11748 Define this macro to update the current function stack boundary if
11752 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11753 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11754 different argument pointer register is needed to access the function's
11755 argument list due to stack realignment. Return @code{NULL} if no DRAP
11759 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11760 When optimization is disabled, this hook indicates whether or not
11761 arguments should be allocated to stack slots. Normally, GCC allocates
11762 stacks slots for arguments when not optimizing in order to make
11763 debugging easier. However, when a function is declared with
11764 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11765 cannot safely move arguments from the registers in which they are passed
11766 to the stack. Therefore, this hook should return true in general, but
11767 false for naked functions. The default implementation always returns true.
11770 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11771 On some architectures it can take multiple instructions to synthesize
11772 a constant. If there is another constant already in a register that
11773 is close enough in value then it is preferable that the new constant
11774 is computed from this register using immediate addition or
11775 subtraction. We accomplish this through CSE. Besides the value of
11776 the constant we also add a lower and an upper constant anchor to the
11777 available expressions. These are then queried when encountering new
11778 constants. The anchors are computed by rounding the constant up and
11779 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11780 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11781 accepted by immediate-add plus one. We currently assume that the
11782 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11783 MIPS, where add-immediate takes a 16-bit signed value,
11784 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11785 is zero, which disables this optimization.
11788 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_ASAN_SHADOW_OFFSET (void)
11789 Return the offset bitwise ored into shifted address to get corresponding
11790 Address Sanitizer shadow memory address. NULL if Address Sanitizer is not
11791 supported by the target.
11794 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_MEMMODEL_CHECK (unsigned HOST_WIDE_INT @var{val})
11795 Validate target specific memory model mask bits. When NULL no target specific
11796 memory model bits are allowed.
11799 @deftypevr {Target Hook} {unsigned char} TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
11800 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}.
11803 @deftypefn {Target Hook} bool TARGET_HAS_IFUNC_P (void)
11804 It returns true if the target supports GNU indirect functions.
11805 The support includes the assembler, linker and dynamic linker.
11806 The default value of this hook is based on target's libc.
11809 @deftypefn {Target Hook} {unsigned int} TARGET_ATOMIC_ALIGN_FOR_MODE (machine_mode @var{mode})
11810 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.
11813 @deftypefn {Target Hook} void TARGET_ATOMIC_ASSIGN_EXPAND_FENV (tree *@var{hold}, tree *@var{clear}, tree *@var{update})
11814 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}}.
11817 @deftypefn {Target Hook} void TARGET_RECORD_OFFLOAD_SYMBOL (tree)
11818 Used when offloaded functions are seen in the compilation unit and no named
11819 sections are available. It is called once for each symbol that must be
11820 recorded in the offload function and variable table.
11823 @deftypefn {Target Hook} {char *} TARGET_OFFLOAD_OPTIONS (void)
11824 Used when writing out the list of options into an LTO file. It should
11825 translate any relevant target-specific options (such as the ABI in use)
11826 into one of the @option{-foffload} options that exist as a common interface
11827 to express such options. It should return a string containing these options,
11828 separated by spaces, which the caller will free.
11832 @defmac TARGET_SUPPORTS_WIDE_INT
11834 On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
11835 objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
11836 to indicate that large integers are stored in
11837 @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows
11838 very large integer constants to be represented. @code{CONST_DOUBLE}
11839 is limited to twice the size of the host's @code{HOST_WIDE_INT}
11842 Converting a port mostly requires looking for the places where
11843 @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
11844 code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i
11845 const_double"} at the port level gets you to 95% of the changes that
11846 need to be made. There are a few places that require a deeper look.
11850 There is no equivalent to @code{hval} and @code{lval} for
11851 @code{CONST_WIDE_INT}s. This would be difficult to express in the md
11852 language since there are a variable number of elements.
11854 Most ports only check that @code{hval} is either 0 or -1 to see if the
11855 value is small. As mentioned above, this will no longer be necessary
11856 since small constants are always @code{CONST_INT}. Of course there
11857 are still a few exceptions, the alpha's constraint used by the zap
11858 instruction certainly requires careful examination by C code.
11859 However, all the current code does is pass the hval and lval to C
11860 code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
11861 not really a large change.
11864 Because there is no standard template that ports use to materialize
11865 constants, there is likely to be some futzing that is unique to each
11869 The rtx costs may have to be adjusted to properly account for larger
11870 constants that are represented as @code{CONST_WIDE_INT}.
11873 All and all it does not take long to convert ports that the
11874 maintainer is familiar with.
11878 @deftypefn {Target Hook} void TARGET_RUN_TARGET_SELFTESTS (void)
11879 If selftests are enabled, run any selftests for this target.