1 @c Copyright (C) 1988-2019 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 * D Language and ABI:: Controlling D ABI changes.
56 * Named Address Spaces:: Adding support for named address spaces
57 * Misc:: Everything else.
60 @node Target Structure
61 @section The Global @code{targetm} Variable
63 @cindex target functions
65 @deftypevar {struct gcc_target} targetm
66 The target @file{.c} file must define the global @code{targetm} variable
67 which contains pointers to functions and data relating to the target
68 machine. The variable is declared in @file{target.h};
69 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
70 used to initialize the variable, and macros for the default initializers
71 for elements of the structure. The @file{.c} file should override those
72 macros for which the default definition is inappropriate. For example:
75 #include "target-def.h"
77 /* @r{Initialize the GCC target structure.} */
79 #undef TARGET_COMP_TYPE_ATTRIBUTES
80 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
82 struct gcc_target targetm = TARGET_INITIALIZER;
86 Where a macro should be defined in the @file{.c} file in this manner to
87 form part of the @code{targetm} structure, it is documented below as a
88 ``Target Hook'' with a prototype. Many macros will change in future
89 from being defined in the @file{.h} file to being part of the
90 @code{targetm} structure.
92 Similarly, there is a @code{targetcm} variable for hooks that are
93 specific to front ends for C-family languages, documented as ``C
94 Target Hook''. This is declared in @file{c-family/c-target.h}, the
95 initializer @code{TARGETCM_INITIALIZER} in
96 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
97 themselves, they should set @code{target_has_targetcm=yes} in
98 @file{config.gcc}; otherwise a default definition is used.
100 Similarly, there is a @code{targetm_common} variable for hooks that
101 are shared between the compiler driver and the compilers proper,
102 documented as ``Common Target Hook''. This is declared in
103 @file{common/common-target.h}, the initializer
104 @code{TARGETM_COMMON_INITIALIZER} in
105 @file{common/common-target-def.h}. If targets initialize
106 @code{targetm_common} themselves, they should set
107 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
108 default definition is used.
110 Similarly, there is a @code{targetdm} variable for hooks that are
111 specific to the D language front end, documented as ``D Target Hook''.
112 This is declared in @file{d/d-target.h}, the initializer
113 @code{TARGETDM_INITIALIZER} in @file{d/d-target-def.h}. If targets
114 initialize @code{targetdm} themselves, they should set
115 @code{target_has_targetdm=yes} in @file{config.gcc}; otherwise a default
119 @section Controlling the Compilation Driver, @file{gcc}
121 @cindex controlling the compilation driver
123 @c prevent bad page break with this line
124 You can control the compilation driver.
126 @defmac DRIVER_SELF_SPECS
127 A list of specs for the driver itself. It should be a suitable
128 initializer for an array of strings, with no surrounding braces.
130 The driver applies these specs to its own command line between loading
131 default @file{specs} files (but not command-line specified ones) and
132 choosing the multilib directory or running any subcommands. It
133 applies them in the order given, so each spec can depend on the
134 options added by earlier ones. It is also possible to remove options
135 using @samp{%<@var{option}} in the usual way.
137 This macro can be useful when a port has several interdependent target
138 options. It provides a way of standardizing the command line so
139 that the other specs are easier to write.
141 Do not define this macro if it does not need to do anything.
144 @defmac OPTION_DEFAULT_SPECS
145 A list of specs used to support configure-time default options (i.e.@:
146 @option{--with} options) in the driver. It should be a suitable initializer
147 for an array of structures, each containing two strings, without the
148 outermost pair of surrounding braces.
150 The first item in the pair is the name of the default. This must match
151 the code in @file{config.gcc} for the target. The second item is a spec
152 to apply if a default with this name was specified. The string
153 @samp{%(VALUE)} in the spec will be replaced by the value of the default
154 everywhere it occurs.
156 The driver will apply these specs to its own command line between loading
157 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
158 the same mechanism as @code{DRIVER_SELF_SPECS}.
160 Do not define this macro if it does not need to do anything.
164 A C string constant that tells the GCC driver program options to
165 pass to CPP@. It can also specify how to translate options you
166 give to GCC into options for GCC to pass to the CPP@.
168 Do not define this macro if it does not need to do anything.
171 @defmac CPLUSPLUS_CPP_SPEC
172 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
173 than C@. If you do not define this macro, then the value of
174 @code{CPP_SPEC} (if any) will be used instead.
178 A C string constant that tells the GCC driver program options to
179 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
181 It can also specify how to translate options you give to GCC into options
182 for GCC to pass to front ends.
184 Do not define this macro if it does not need to do anything.
188 A C string constant that tells the GCC driver program options to
189 pass to @code{cc1plus}. It can also specify how to translate options you
190 give to GCC into options for GCC to pass to the @code{cc1plus}.
192 Do not define this macro if it does not need to do anything.
193 Note that everything defined in CC1_SPEC is already passed to
194 @code{cc1plus} so there is no need to duplicate the contents of
195 CC1_SPEC in CC1PLUS_SPEC@.
199 A C string constant that tells the GCC driver program options to
200 pass to the assembler. It can also specify how to translate options
201 you give to GCC into options for GCC to pass to the assembler.
202 See the file @file{sun3.h} for an example of this.
204 Do not define this macro if it does not need to do anything.
207 @defmac ASM_FINAL_SPEC
208 A C string constant that tells the GCC driver program how to
209 run any programs which cleanup after the normal assembler.
210 Normally, this is not needed. See the file @file{mips.h} for
213 Do not define this macro if it does not need to do anything.
216 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
217 Define this macro, with no value, if the driver should give the assembler
218 an argument consisting of a single dash, @option{-}, to instruct it to
219 read from its standard input (which will be a pipe connected to the
220 output of the compiler proper). This argument is given after any
221 @option{-o} option specifying the name of the output file.
223 If you do not define this macro, the assembler is assumed to read its
224 standard input if given no non-option arguments. If your assembler
225 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
226 see @file{mips.h} for instance.
230 A C string constant that tells the GCC driver program options to
231 pass to the linker. It can also specify how to translate options you
232 give to GCC into options for GCC to pass to the linker.
234 Do not define this macro if it does not need to do anything.
238 Another C string constant used much like @code{LINK_SPEC}. The difference
239 between the two is that @code{LIB_SPEC} is used at the end of the
240 command given to the linker.
242 If this macro is not defined, a default is provided that
243 loads the standard C library from the usual place. See @file{gcc.c}.
247 Another C string constant that tells the GCC driver program
248 how and when to place a reference to @file{libgcc.a} into the
249 linker command line. This constant is placed both before and after
250 the value of @code{LIB_SPEC}.
252 If this macro is not defined, the GCC driver provides a default that
253 passes the string @option{-lgcc} to the linker.
256 @defmac REAL_LIBGCC_SPEC
257 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
258 @code{LIBGCC_SPEC} is not directly used by the driver program but is
259 instead modified to refer to different versions of @file{libgcc.a}
260 depending on the values of the command line flags @option{-static},
261 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
262 targets where these modifications are inappropriate, define
263 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
264 driver how to place a reference to @file{libgcc} on the link command
265 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
268 @defmac USE_LD_AS_NEEDED
269 A macro that controls the modifications to @code{LIBGCC_SPEC}
270 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
271 generated that uses @option{--as-needed} or equivalent options and the
272 shared @file{libgcc} in place of the
273 static exception handler library, when linking without any of
274 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
278 If defined, this C string constant is added to @code{LINK_SPEC}.
279 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
280 the modifications to @code{LIBGCC_SPEC} mentioned in
281 @code{REAL_LIBGCC_SPEC}.
284 @defmac STARTFILE_SPEC
285 Another C string constant used much like @code{LINK_SPEC}. The
286 difference between the two is that @code{STARTFILE_SPEC} is used at
287 the very beginning of the command given to the linker.
289 If this macro is not defined, a default is provided that loads the
290 standard C startup file from the usual place. See @file{gcc.c}.
294 Another C string constant used much like @code{LINK_SPEC}. The
295 difference between the two is that @code{ENDFILE_SPEC} is used at
296 the very end of the command given to the linker.
298 Do not define this macro if it does not need to do anything.
301 @defmac THREAD_MODEL_SPEC
302 GCC @code{-v} will print the thread model GCC was configured to use.
303 However, this doesn't work on platforms that are multilibbed on thread
304 models, such as AIX 4.3. On such platforms, define
305 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
306 blanks that names one of the recognized thread models. @code{%*}, the
307 default value of this macro, will expand to the value of
308 @code{thread_file} set in @file{config.gcc}.
311 @defmac SYSROOT_SUFFIX_SPEC
312 Define this macro to add a suffix to the target sysroot when GCC is
313 configured with a sysroot. This will cause GCC to search for usr/lib,
314 et al, within sysroot+suffix.
317 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
318 Define this macro to add a headers_suffix to the target sysroot when
319 GCC is configured with a sysroot. This will cause GCC to pass the
320 updated sysroot+headers_suffix to CPP, causing it to search for
321 usr/include, et al, within sysroot+headers_suffix.
325 Define this macro to provide additional specifications to put in the
326 @file{specs} file that can be used in various specifications like
329 The definition should be an initializer for an array of structures,
330 containing a string constant, that defines the specification name, and a
331 string constant that provides the specification.
333 Do not define this macro if it does not need to do anything.
335 @code{EXTRA_SPECS} is useful when an architecture contains several
336 related targets, which have various @code{@dots{}_SPECS} which are similar
337 to each other, and the maintainer would like one central place to keep
340 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
341 define either @code{_CALL_SYSV} when the System V calling sequence is
342 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
345 The @file{config/rs6000/rs6000.h} target file defines:
348 #define EXTRA_SPECS \
349 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
351 #define CPP_SYS_DEFAULT ""
354 The @file{config/rs6000/sysv.h} target file defines:
358 "%@{posix: -D_POSIX_SOURCE @} \
359 %@{mcall-sysv: -D_CALL_SYSV @} \
360 %@{!mcall-sysv: %(cpp_sysv_default) @} \
361 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
363 #undef CPP_SYSV_DEFAULT
364 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
367 while the @file{config/rs6000/eabiaix.h} target file defines
368 @code{CPP_SYSV_DEFAULT} as:
371 #undef CPP_SYSV_DEFAULT
372 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
376 @defmac LINK_LIBGCC_SPECIAL_1
377 Define this macro if the driver program should find the library
378 @file{libgcc.a}. If you do not define this macro, the driver program will pass
379 the argument @option{-lgcc} to tell the linker to do the search.
382 @defmac LINK_GCC_C_SEQUENCE_SPEC
383 The sequence in which libgcc and libc are specified to the linker.
384 By default this is @code{%G %L %G}.
387 @defmac POST_LINK_SPEC
388 Define this macro to add additional steps to be executed after linker.
389 The default value of this macro is empty string.
392 @defmac LINK_COMMAND_SPEC
393 A C string constant giving the complete command line need to execute the
394 linker. When you do this, you will need to update your port each time a
395 change is made to the link command line within @file{gcc.c}. Therefore,
396 define this macro only if you need to completely redefine the command
397 line for invoking the linker and there is no other way to accomplish
398 the effect you need. Overriding this macro may be avoidable by overriding
399 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
402 @deftypevr {Common Target Hook} bool TARGET_ALWAYS_STRIP_DOTDOT
403 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.
406 @defmac MULTILIB_DEFAULTS
407 Define this macro as a C expression for the initializer of an array of
408 string to tell the driver program which options are defaults for this
409 target and thus do not need to be handled specially when using
410 @code{MULTILIB_OPTIONS}.
412 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
413 the target makefile fragment or if none of the options listed in
414 @code{MULTILIB_OPTIONS} are set by default.
415 @xref{Target Fragment}.
418 @defmac RELATIVE_PREFIX_NOT_LINKDIR
419 Define this macro to tell @command{gcc} that it should only translate
420 a @option{-B} prefix into a @option{-L} linker option if the prefix
421 indicates an absolute file name.
424 @defmac MD_EXEC_PREFIX
425 If defined, this macro is an additional prefix to try after
426 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
427 when the compiler is built as a cross
428 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
429 to the list of directories used to find the assembler in @file{configure.ac}.
432 @defmac STANDARD_STARTFILE_PREFIX
433 Define this macro as a C string constant if you wish to override the
434 standard choice of @code{libdir} as the default prefix to
435 try when searching for startup files such as @file{crt0.o}.
436 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
437 is built as a cross compiler.
440 @defmac STANDARD_STARTFILE_PREFIX_1
441 Define this macro as a C string constant if you wish to override the
442 standard choice of @code{/lib} as a prefix to try after the default prefix
443 when searching for startup files such as @file{crt0.o}.
444 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
445 is built as a cross compiler.
448 @defmac STANDARD_STARTFILE_PREFIX_2
449 Define this macro as a C string constant if you wish to override the
450 standard choice of @code{/lib} as yet another prefix to try after the
451 default prefix when searching for startup files such as @file{crt0.o}.
452 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
453 is built as a cross compiler.
456 @defmac MD_STARTFILE_PREFIX
457 If defined, this macro supplies an additional prefix to try after the
458 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
459 compiler is built as a cross compiler.
462 @defmac MD_STARTFILE_PREFIX_1
463 If defined, this macro supplies yet another prefix to try after the
464 standard prefixes. It is not searched when the compiler is built as a
468 @defmac INIT_ENVIRONMENT
469 Define this macro as a C string constant if you wish to set environment
470 variables for programs called by the driver, such as the assembler and
471 loader. The driver passes the value of this macro to @code{putenv} to
472 initialize the necessary environment variables.
475 @defmac LOCAL_INCLUDE_DIR
476 Define this macro as a C string constant if you wish to override the
477 standard choice of @file{/usr/local/include} as the default prefix to
478 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
479 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
480 @file{config.gcc}, normally @file{/usr/include}) in the search order.
482 Cross compilers do not search either @file{/usr/local/include} or its
486 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
487 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
488 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
489 If you do not define this macro, no component is used.
492 @defmac INCLUDE_DEFAULTS
493 Define this macro if you wish to override the entire default search path
494 for include files. For a native compiler, the default search path
495 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
496 @code{GPLUSPLUS_INCLUDE_DIR}, and
497 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
498 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
499 and specify private search areas for GCC@. The directory
500 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
502 The definition should be an initializer for an array of structures.
503 Each array element should have four elements: the directory name (a
504 string constant), the component name (also a string constant), a flag
505 for C++-only directories,
506 and a flag showing that the includes in the directory don't need to be
507 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
508 the array with a null element.
510 The component name denotes what GNU package the include file is part of,
511 if any, in all uppercase letters. For example, it might be @samp{GCC}
512 or @samp{BINUTILS}. If the package is part of a vendor-supplied
513 operating system, code the component name as @samp{0}.
515 For example, here is the definition used for VAX/VMS:
518 #define INCLUDE_DEFAULTS \
520 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
521 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
522 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
529 Here is the order of prefixes tried for exec files:
533 Any prefixes specified by the user with @option{-B}.
536 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
537 is not set and the compiler has not been installed in the configure-time
538 @var{prefix}, the location in which the compiler has actually been installed.
541 The directories specified by the environment variable @code{COMPILER_PATH}.
544 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
545 in the configured-time @var{prefix}.
548 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
551 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
554 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
558 Here is the order of prefixes tried for startfiles:
562 Any prefixes specified by the user with @option{-B}.
565 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
566 value based on the installed toolchain location.
569 The directories specified by the environment variable @code{LIBRARY_PATH}
570 (or port-specific name; native only, cross compilers do not use this).
573 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
574 in the configured @var{prefix} or this is a native compiler.
577 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
580 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
584 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
585 native compiler, or we have a target system root.
588 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
589 native compiler, or we have a target system root.
592 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
593 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
594 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
597 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
598 compiler, or we have a target system root. The default for this macro is
602 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
603 compiler, or we have a target system root. The default for this macro is
607 @node Run-time Target
608 @section Run-time Target Specification
609 @cindex run-time target specification
610 @cindex predefined macros
611 @cindex target specifications
613 @c prevent bad page break with this line
614 Here are run-time target specifications.
616 @defmac TARGET_CPU_CPP_BUILTINS ()
617 This function-like macro expands to a block of code that defines
618 built-in preprocessor macros and assertions for the target CPU, using
619 the functions @code{builtin_define}, @code{builtin_define_std} and
620 @code{builtin_assert}. When the front end
621 calls this macro it provides a trailing semicolon, and since it has
622 finished command line option processing your code can use those
625 @code{builtin_assert} takes a string in the form you pass to the
626 command-line option @option{-A}, such as @code{cpu=mips}, and creates
627 the assertion. @code{builtin_define} takes a string in the form
628 accepted by option @option{-D} and unconditionally defines the macro.
630 @code{builtin_define_std} takes a string representing the name of an
631 object-like macro. If it doesn't lie in the user's namespace,
632 @code{builtin_define_std} defines it unconditionally. Otherwise, it
633 defines a version with two leading underscores, and another version
634 with two leading and trailing underscores, and defines the original
635 only if an ISO standard was not requested on the command line. For
636 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
637 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
638 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
639 defines only @code{_ABI64}.
641 You can also test for the C dialect being compiled. The variable
642 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
643 or @code{clk_objective_c}. Note that if we are preprocessing
644 assembler, this variable will be @code{clk_c} but the function-like
645 macro @code{preprocessing_asm_p()} will return true, so you might want
646 to check for that first. If you need to check for strict ANSI, the
647 variable @code{flag_iso} can be used. The function-like macro
648 @code{preprocessing_trad_p()} can be used to check for traditional
652 @defmac TARGET_OS_CPP_BUILTINS ()
653 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
654 and is used for the target operating system instead.
657 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
658 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
659 and is used for the target object format. @file{elfos.h} uses this
660 macro to define @code{__ELF__}, so you probably do not need to define
664 @deftypevar {extern int} target_flags
665 This variable is declared in @file{options.h}, which is included before
666 any target-specific headers.
669 @deftypevr {Common Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
670 This variable specifies the initial value of @code{target_flags}.
671 Its default setting is 0.
674 @cindex optional hardware or system features
675 @cindex features, optional, in system conventions
677 @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})
678 This hook is called whenever the user specifies one of the
679 target-specific options described by the @file{.opt} definition files
680 (@pxref{Options}). It has the opportunity to do some option-specific
681 processing and should return true if the option is valid. The default
682 definition does nothing but return true.
684 @var{decoded} specifies the option and its arguments. @var{opts} and
685 @var{opts_set} are the @code{gcc_options} structures to be used for
686 storing option state, and @var{loc} is the location at which the
687 option was passed (@code{UNKNOWN_LOCATION} except for options passed
691 @deftypefn {C Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
692 This target hook is called whenever the user specifies one of the
693 target-specific C language family options described by the @file{.opt}
694 definition files(@pxref{Options}). It has the opportunity to do some
695 option-specific processing and should return true if the option is
696 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
697 default definition does nothing but return false.
699 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
700 options. However, if processing an option requires routines that are
701 only available in the C (and related language) front ends, then you
702 should use @code{TARGET_HANDLE_C_OPTION} instead.
705 @deftypefn {C Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING_OBJECT (tree @var{string})
706 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.
709 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE (const char *@var{classname})
710 Declare that Objective C class @var{classname} is referenced by the current TU.
713 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_CLASS_DEFINITION (const char *@var{classname})
714 Declare that Objective C class @var{classname} is defined by the current TU.
717 @deftypefn {C Target Hook} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref})
718 If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object.
721 @deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list})
722 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.
725 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
726 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
727 but is called when the optimize level is changed via an attribute or
728 pragma or when it is reset at the end of the code affected by the
729 attribute or pragma. It is not called at the beginning of compilation
730 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
731 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
732 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
735 @defmac C_COMMON_OVERRIDE_OPTIONS
736 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
737 but is only used in the C
738 language frontends (C, Objective-C, C++, Objective-C++) and so can be
739 used to alter option flag variables which only exist in those
743 @deftypevr {Common Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
744 Some machines may desire to change what optimizations are performed for
745 various optimization levels. This variable, if defined, describes
746 options to enable at particular sets of optimization levels. These
747 options are processed once
748 just after the optimization level is determined and before the remainder
749 of the command options have been parsed, so may be overridden by other
750 options passed explicitly.
752 This processing is run once at program startup and when the optimization
753 options are changed via @code{#pragma GCC optimize} or by using the
754 @code{optimize} attribute.
757 @deftypefn {Common Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
758 Set target-dependent initial values of fields in @var{opts}.
761 @deftypefn {Common Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
762 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
765 @deftypefn {Common Target Hook} bool TARGET_OPTION_VALIDATE_PARAM (int, @var{int})
766 Validate target-dependent value for @option{--param} settings, using calls to @code{set_param_value}.
769 @defmac SWITCHABLE_TARGET
770 Some targets need to switch between substantially different subtargets
771 during compilation. For example, the MIPS target has one subtarget for
772 the traditional MIPS architecture and another for MIPS16. Source code
773 can switch between these two subarchitectures using the @code{mips16}
774 and @code{nomips16} attributes.
776 Such subtargets can differ in things like the set of available
777 registers, the set of available instructions, the costs of various
778 operations, and so on. GCC caches a lot of this type of information
779 in global variables, and recomputing them for each subtarget takes a
780 significant amount of time. The compiler therefore provides a facility
781 for maintaining several versions of the global variables and quickly
782 switching between them; see @file{target-globals.h} for details.
784 Define this macro to 1 if your target needs this facility. The default
788 @deftypefn {Target Hook} bool TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P (void)
789 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.
792 @node Per-Function Data
793 @section Defining data structures for per-function information.
794 @cindex per-function data
795 @cindex data structures
797 If the target needs to store information on a per-function basis, GCC
798 provides a macro and a couple of variables to allow this. Note, just
799 using statics to store the information is a bad idea, since GCC supports
800 nested functions, so you can be halfway through encoding one function
801 when another one comes along.
803 GCC defines a data structure called @code{struct function} which
804 contains all of the data specific to an individual function. This
805 structure contains a field called @code{machine} whose type is
806 @code{struct machine_function *}, which can be used by targets to point
807 to their own specific data.
809 If a target needs per-function specific data it should define the type
810 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
811 This macro should be used to initialize the function pointer
812 @code{init_machine_status}. This pointer is explained below.
814 One typical use of per-function, target specific data is to create an
815 RTX to hold the register containing the function's return address. This
816 RTX can then be used to implement the @code{__builtin_return_address}
817 function, for level 0.
819 Note---earlier implementations of GCC used a single data area to hold
820 all of the per-function information. Thus when processing of a nested
821 function began the old per-function data had to be pushed onto a
822 stack, and when the processing was finished, it had to be popped off the
823 stack. GCC used to provide function pointers called
824 @code{save_machine_status} and @code{restore_machine_status} to handle
825 the saving and restoring of the target specific information. Since the
826 single data area approach is no longer used, these pointers are no
829 @defmac INIT_EXPANDERS
830 Macro called to initialize any target specific information. This macro
831 is called once per function, before generation of any RTL has begun.
832 The intention of this macro is to allow the initialization of the
833 function pointer @code{init_machine_status}.
836 @deftypevar {void (*)(struct function *)} init_machine_status
837 If this function pointer is non-@code{NULL} it will be called once per
838 function, before function compilation starts, in order to allow the
839 target to perform any target specific initialization of the
840 @code{struct function} structure. It is intended that this would be
841 used to initialize the @code{machine} of that structure.
843 @code{struct machine_function} structures are expected to be freed by GC@.
844 Generally, any memory that they reference must be allocated by using
845 GC allocation, including the structure itself.
849 @section Storage Layout
850 @cindex storage layout
852 Note that the definitions of the macros in this table which are sizes or
853 alignments measured in bits do not need to be constant. They can be C
854 expressions that refer to static variables, such as the @code{target_flags}.
855 @xref{Run-time Target}.
857 @defmac BITS_BIG_ENDIAN
858 Define this macro to have the value 1 if the most significant bit in a
859 byte has the lowest number; otherwise define it to have the value zero.
860 This means that bit-field instructions count from the most significant
861 bit. If the machine has no bit-field instructions, then this must still
862 be defined, but it doesn't matter which value it is defined to. This
863 macro need not be a constant.
865 This macro does not affect the way structure fields are packed into
866 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
869 @defmac BYTES_BIG_ENDIAN
870 Define this macro to have the value 1 if the most significant byte in a
871 word has the lowest number. This macro need not be a constant.
874 @defmac WORDS_BIG_ENDIAN
875 Define this macro to have the value 1 if, in a multiword object, the
876 most significant word has the lowest number. This applies to both
877 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
878 order of words in memory is not the same as the order in registers. This
879 macro need not be a constant.
882 @defmac REG_WORDS_BIG_ENDIAN
883 On some machines, the order of words in a multiword object differs between
884 registers in memory. In such a situation, define this macro to describe
885 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
886 the order of words in memory.
889 @defmac FLOAT_WORDS_BIG_ENDIAN
890 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
891 @code{TFmode} floating point numbers are stored in memory with the word
892 containing the sign bit at the lowest address; otherwise define it to
893 have the value 0. This macro need not be a constant.
895 You need not define this macro if the ordering is the same as for
899 @defmac BITS_PER_WORD
900 Number of bits in a word. If you do not define this macro, the default
901 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
904 @defmac MAX_BITS_PER_WORD
905 Maximum number of bits in a word. If this is undefined, the default is
906 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
907 largest value that @code{BITS_PER_WORD} can have at run-time.
910 @defmac UNITS_PER_WORD
911 Number of storage units in a word; normally the size of a general-purpose
912 register, a power of two from 1 or 8.
915 @defmac MIN_UNITS_PER_WORD
916 Minimum number of units in a word. If this is undefined, the default is
917 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
918 smallest value that @code{UNITS_PER_WORD} can have at run-time.
922 Width of a pointer, in bits. You must specify a value no wider than the
923 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
924 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
925 a value the default is @code{BITS_PER_WORD}.
928 @defmac POINTERS_EXTEND_UNSIGNED
929 A C expression that determines how pointers should be extended from
930 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
931 greater than zero if pointers should be zero-extended, zero if they
932 should be sign-extended, and negative if some other sort of conversion
933 is needed. In the last case, the extension is done by the target's
934 @code{ptr_extend} instruction.
936 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
937 and @code{word_mode} are all the same width.
940 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
941 A macro to update @var{m} and @var{unsignedp} when an object whose type
942 is @var{type} and which has the specified mode and signedness is to be
943 stored in a register. This macro is only called when @var{type} is a
946 On most RISC machines, which only have operations that operate on a full
947 register, define this macro to set @var{m} to @code{word_mode} if
948 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
949 cases, only integer modes should be widened because wider-precision
950 floating-point operations are usually more expensive than their narrower
953 For most machines, the macro definition does not change @var{unsignedp}.
954 However, some machines, have instructions that preferentially handle
955 either signed or unsigned quantities of certain modes. For example, on
956 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
957 sign-extend the result to 64 bits. On such machines, set
958 @var{unsignedp} according to which kind of extension is more efficient.
960 Do not define this macro if it would never modify @var{m}.
963 @deftypefn {Target Hook} {enum flt_eval_method} TARGET_C_EXCESS_PRECISION (enum excess_precision_type @var{type})
964 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}.
967 @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})
968 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
969 function return values. The target hook should return the new mode
970 and possibly change @code{*@var{punsignedp}} if the promotion should
971 change signedness. This function is called only for scalar @emph{or
974 @var{for_return} allows to distinguish the promotion of arguments and
975 return values. If it is @code{1}, a return value is being promoted and
976 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
977 If it is @code{2}, the returned mode should be that of the register in
978 which an incoming parameter is copied, or the outgoing result is computed;
979 then the hook should return the same mode as @code{promote_mode}, though
980 the signedness may be different.
982 @var{type} can be NULL when promoting function arguments of libcalls.
984 The default is to not promote arguments and return values. You can
985 also define the hook to @code{default_promote_function_mode_always_promote}
986 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
989 @defmac PARM_BOUNDARY
990 Normal alignment required for function parameters on the stack, in
991 bits. All stack parameters receive at least this much alignment
992 regardless of data type. On most machines, this is the same as the
996 @defmac STACK_BOUNDARY
997 Define this macro to the minimum alignment enforced by hardware for the
998 stack pointer on this machine. The definition is a C expression for the
999 desired alignment (measured in bits). This value is used as a default
1000 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1001 this should be the same as @code{PARM_BOUNDARY}.
1004 @defmac PREFERRED_STACK_BOUNDARY
1005 Define this macro if you wish to preserve a certain alignment for the
1006 stack pointer, greater than what the hardware enforces. The definition
1007 is a C expression for the desired alignment (measured in bits). This
1008 macro must evaluate to a value equal to or larger than
1009 @code{STACK_BOUNDARY}.
1012 @defmac INCOMING_STACK_BOUNDARY
1013 Define this macro if the incoming stack boundary may be different
1014 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
1015 to a value equal to or larger than @code{STACK_BOUNDARY}.
1018 @defmac FUNCTION_BOUNDARY
1019 Alignment required for a function entry point, in bits.
1022 @defmac BIGGEST_ALIGNMENT
1023 Biggest alignment that any data type can require on this machine, in
1024 bits. Note that this is not the biggest alignment that is supported,
1025 just the biggest alignment that, when violated, may cause a fault.
1028 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_ABSOLUTE_BIGGEST_ALIGNMENT
1029 If defined, this target hook specifies the absolute biggest alignment
1030 that a type or variable can have on this machine, otherwise,
1031 @code{BIGGEST_ALIGNMENT} is used.
1034 @defmac MALLOC_ABI_ALIGNMENT
1035 Alignment, in bits, a C conformant malloc implementation has to
1036 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1039 @defmac ATTRIBUTE_ALIGNED_VALUE
1040 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1041 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1044 @defmac MINIMUM_ATOMIC_ALIGNMENT
1045 If defined, the smallest alignment, in bits, that can be given to an
1046 object that can be referenced in one operation, without disturbing any
1047 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1048 on machines that don't have byte or half-word store operations.
1051 @defmac BIGGEST_FIELD_ALIGNMENT
1052 Biggest alignment that any structure or union field can require on this
1053 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1054 structure and union fields only, unless the field alignment has been set
1055 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1058 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{type}, @var{computed})
1059 An expression for the alignment of a structure field @var{field} of
1060 type @var{type} if the alignment computed in the usual way (including
1061 applying of @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1062 alignment) is @var{computed}. It overrides alignment only if the
1063 field alignment has not been set by the
1064 @code{__attribute__ ((aligned (@var{n})))} construct. Note that @var{field}
1065 may be @code{NULL_TREE} in case we just query for the minimum alignment
1066 of a field of type @var{type} in structure context.
1069 @defmac MAX_STACK_ALIGNMENT
1070 Biggest stack alignment guaranteed by the backend. Use this macro
1071 to specify the maximum alignment of a variable on stack.
1073 If not defined, the default value is @code{STACK_BOUNDARY}.
1075 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1076 @c But the fix for PR 32893 indicates that we can only guarantee
1077 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1078 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1081 @defmac MAX_OFILE_ALIGNMENT
1082 Biggest alignment supported by the object file format of this machine.
1083 Use this macro to limit the alignment which can be specified using the
1084 @code{__attribute__ ((aligned (@var{n})))} construct for functions and
1085 objects with static storage duration. The alignment of automatic
1086 objects may exceed the object file format maximum up to the maximum
1087 supported by GCC. If not defined, the default value is
1088 @code{BIGGEST_ALIGNMENT}.
1090 On systems that use ELF, the default (in @file{config/elfos.h}) is
1091 the largest supported 32-bit ELF section alignment representable on
1092 a 32-bit host e.g.@: @samp{(((uint64_t) 1 << 28) * 8)}.
1093 On 32-bit ELF the largest supported section alignment in bits is
1094 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1097 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_STATIC_RTX_ALIGNMENT (machine_mode @var{mode})
1098 This hook returns the preferred alignment in bits for a
1099 statically-allocated rtx, such as a constant pool entry. @var{mode}
1100 is the mode of the rtx. The default implementation returns
1101 @samp{GET_MODE_ALIGNMENT (@var{mode})}.
1104 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1105 If defined, a C expression to compute the alignment for a variable in
1106 the static store. @var{type} is the data type, and @var{basic-align} is
1107 the alignment that the object would ordinarily have. The value of this
1108 macro is used instead of that alignment to align the object.
1110 If this macro is not defined, then @var{basic-align} is used.
1113 One use of this macro is to increase alignment of medium-size data to
1114 make it all fit in fewer cache lines. Another is to cause character
1115 arrays to be word-aligned so that @code{strcpy} calls that copy
1116 constants to character arrays can be done inline.
1119 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1120 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1121 some alignment increase, instead of optimization only purposes. E.g.@
1122 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1123 must be aligned to 16 byte boundaries.
1125 If this macro is not defined, then @var{basic-align} is used.
1128 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_CONSTANT_ALIGNMENT (const_tree @var{constant}, HOST_WIDE_INT @var{basic_align})
1129 This hook returns the alignment in bits of a constant that is being
1130 placed in memory. @var{constant} is the constant and @var{basic_align}
1131 is the alignment that the object would ordinarily have.
1133 The default definition just returns @var{basic_align}.
1135 The typical use of this hook is to increase alignment for string
1136 constants to be word aligned so that @code{strcpy} calls that copy
1137 constants can be done inline. The function
1138 @code{constant_alignment_word_strings} provides such a definition.
1141 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1142 If defined, a C expression to compute the alignment for a variable in
1143 the local store. @var{type} is the data type, and @var{basic-align} is
1144 the alignment that the object would ordinarily have. The value of this
1145 macro is used instead of that alignment to align the object.
1147 If this macro is not defined, then @var{basic-align} is used.
1149 One use of this macro is to increase alignment of medium-size data to
1150 make it all fit in fewer cache lines.
1152 If the value of this macro has a type, it should be an unsigned type.
1155 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_VECTOR_ALIGNMENT (const_tree @var{type})
1156 This hook can be used to define the alignment for a vector of type
1157 @var{type}, in order to comply with a platform ABI. The default is to
1158 require natural alignment for vector types. The alignment returned by
1159 this hook must be a power-of-two multiple of the default alignment of
1160 the vector element type.
1163 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1164 If defined, a C expression to compute the alignment for stack slot.
1165 @var{type} is the data type, @var{mode} is the widest mode available,
1166 and @var{basic-align} is the alignment that the slot would ordinarily
1167 have. The value of this macro is used instead of that alignment to
1170 If this macro is not defined, then @var{basic-align} is used when
1171 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1174 This macro is to set alignment of stack slot to the maximum alignment
1175 of all possible modes which the slot may have.
1177 If the value of this macro has a type, it should be an unsigned type.
1180 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1181 If defined, a C expression to compute the alignment for a local
1182 variable @var{decl}.
1184 If this macro is not defined, then
1185 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1188 One use of this macro is to increase alignment of medium-size data to
1189 make it all fit in fewer cache lines.
1191 If the value of this macro has a type, it should be an unsigned type.
1194 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1195 If defined, a C expression to compute the minimum required alignment
1196 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1197 @var{mode}, assuming normal alignment @var{align}.
1199 If this macro is not defined, then @var{align} will be used.
1202 @defmac EMPTY_FIELD_BOUNDARY
1203 Alignment in bits to be given to a structure bit-field that follows an
1204 empty field such as @code{int : 0;}.
1206 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1209 @defmac STRUCTURE_SIZE_BOUNDARY
1210 Number of bits which any structure or union's size must be a multiple of.
1211 Each structure or union's size is rounded up to a multiple of this.
1213 If you do not define this macro, the default is the same as
1214 @code{BITS_PER_UNIT}.
1217 @defmac STRICT_ALIGNMENT
1218 Define this macro to be the value 1 if instructions will fail to work
1219 if given data not on the nominal alignment. If instructions will merely
1220 go slower in that case, define this macro as 0.
1223 @defmac PCC_BITFIELD_TYPE_MATTERS
1224 Define this if you wish to imitate the way many other C compilers handle
1225 alignment of bit-fields and the structures that contain them.
1227 The behavior is that the type written for a named bit-field (@code{int},
1228 @code{short}, or other integer type) imposes an alignment for the entire
1229 structure, as if the structure really did contain an ordinary field of
1230 that type. In addition, the bit-field is placed within the structure so
1231 that it would fit within such a field, not crossing a boundary for it.
1233 Thus, on most machines, a named bit-field whose type is written as
1234 @code{int} would not cross a four-byte boundary, and would force
1235 four-byte alignment for the whole structure. (The alignment used may
1236 not be four bytes; it is controlled by the other alignment parameters.)
1238 An unnamed bit-field will not affect the alignment of the containing
1241 If the macro is defined, its definition should be a C expression;
1242 a nonzero value for the expression enables this behavior.
1244 Note that if this macro is not defined, or its value is zero, some
1245 bit-fields may cross more than one alignment boundary. The compiler can
1246 support such references if there are @samp{insv}, @samp{extv}, and
1247 @samp{extzv} insns that can directly reference memory.
1249 The other known way of making bit-fields work is to define
1250 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1251 Then every structure can be accessed with fullwords.
1253 Unless the machine has bit-field instructions or you define
1254 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1255 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1257 If your aim is to make GCC use the same conventions for laying out
1258 bit-fields as are used by another compiler, here is how to investigate
1259 what the other compiler does. Compile and run this program:
1278 printf ("Size of foo1 is %d\n",
1279 sizeof (struct foo1));
1280 printf ("Size of foo2 is %d\n",
1281 sizeof (struct foo2));
1286 If this prints 2 and 5, then the compiler's behavior is what you would
1287 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1290 @defmac BITFIELD_NBYTES_LIMITED
1291 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1292 to aligning a bit-field within the structure.
1295 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1296 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1297 whether unnamed bitfields affect the alignment of the containing
1298 structure. The hook should return true if the structure should inherit
1299 the alignment requirements of an unnamed bitfield's type.
1302 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1303 This target hook should return @code{true} if accesses to volatile bitfields
1304 should use the narrowest mode possible. It should return @code{false} if
1305 these accesses should use the bitfield container type.
1307 The default is @code{false}.
1310 @deftypefn {Target Hook} bool TARGET_MEMBER_TYPE_FORCES_BLK (const_tree @var{field}, machine_mode @var{mode})
1311 Return true if a structure, union or array containing @var{field} should
1312 be accessed using @code{BLKMODE}.
1314 If @var{field} is the only field in the structure, @var{mode} is its
1315 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1316 case where structures of one field would require the structure's mode to
1317 retain the field's mode.
1319 Normally, this is not needed.
1322 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1323 Define this macro as an expression for the alignment of a type (given
1324 by @var{type} as a tree node) if the alignment computed in the usual
1325 way is @var{computed} and the alignment explicitly specified was
1328 The default is to use @var{specified} if it is larger; otherwise, use
1329 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1332 @defmac MAX_FIXED_MODE_SIZE
1333 An integer expression for the size in bits of the largest integer
1334 machine mode that should actually be used. All integer machine modes of
1335 this size or smaller can be used for structures and unions with the
1336 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1337 (DImode)} is assumed.
1340 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1341 If defined, an expression of type @code{machine_mode} that
1342 specifies the mode of the save area operand of a
1343 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1344 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1345 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1346 having its mode specified.
1348 You need not define this macro if it always returns @code{Pmode}. You
1349 would most commonly define this macro if the
1350 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1354 @defmac STACK_SIZE_MODE
1355 If defined, an expression of type @code{machine_mode} that
1356 specifies the mode of the size increment operand of an
1357 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1359 You need not define this macro if it always returns @code{word_mode}.
1360 You would most commonly define this macro if the @code{allocate_stack}
1361 pattern needs to support both a 32- and a 64-bit mode.
1364 @deftypefn {Target Hook} scalar_int_mode TARGET_LIBGCC_CMP_RETURN_MODE (void)
1365 This target hook should return the mode to be used for the return value
1366 of compare instructions expanded to libgcc calls. If not defined
1367 @code{word_mode} is returned which is the right choice for a majority of
1371 @deftypefn {Target Hook} scalar_int_mode TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1372 This target hook should return the mode to be used for the shift count operand
1373 of shift instructions expanded to libgcc calls. If not defined
1374 @code{word_mode} is returned which is the right choice for a majority of
1378 @deftypefn {Target Hook} scalar_int_mode TARGET_UNWIND_WORD_MODE (void)
1379 Return machine mode to be used for @code{_Unwind_Word} type.
1380 The default is to use @code{word_mode}.
1383 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1384 This target hook returns @code{true} if bit-fields in the given
1385 @var{record_type} are to be laid out following the rules of Microsoft
1386 Visual C/C++, namely: (i) a bit-field won't share the same storage
1387 unit with the previous bit-field if their underlying types have
1388 different sizes, and the bit-field will be aligned to the highest
1389 alignment of the underlying types of itself and of the previous
1390 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1391 the whole enclosing structure, even if it is unnamed; except that
1392 (iii) a zero-sized bit-field will be disregarded unless it follows
1393 another bit-field of nonzero size. If this hook returns @code{true},
1394 other macros that control bit-field layout are ignored.
1396 When a bit-field is inserted into a packed record, the whole size
1397 of the underlying type is used by one or more same-size adjacent
1398 bit-fields (that is, if its long:3, 32 bits is used in the record,
1399 and any additional adjacent long bit-fields are packed into the same
1400 chunk of 32 bits. However, if the size changes, a new field of that
1401 size is allocated). In an unpacked record, this is the same as using
1402 alignment, but not equivalent when packing.
1404 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1405 the latter will take precedence. If @samp{__attribute__((packed))} is
1406 used on a single field when MS bit-fields are in use, it will take
1407 precedence for that field, but the alignment of the rest of the structure
1408 may affect its placement.
1411 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1412 Returns true if the target supports decimal floating point.
1415 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1416 Returns true if the target supports fixed-point arithmetic.
1419 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1420 This hook is called just before expansion into rtl, allowing the target
1421 to perform additional initializations or analysis before the expansion.
1422 For example, the rs6000 port uses it to allocate a scratch stack slot
1423 for use in copying SDmode values between memory and floating point
1424 registers whenever the function being expanded has any SDmode
1428 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1429 This hook allows the backend to perform additional instantiations on rtl
1430 that are not actually in any insns yet, but will be later.
1433 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1434 If your target defines any fundamental types, or any types your target
1435 uses should be mangled differently from the default, define this hook
1436 to return the appropriate encoding for these types as part of a C++
1437 mangled name. The @var{type} argument is the tree structure representing
1438 the type to be mangled. The hook may be applied to trees which are
1439 not target-specific fundamental types; it should return @code{NULL}
1440 for all such types, as well as arguments it does not recognize. If the
1441 return value is not @code{NULL}, it must point to a statically-allocated
1444 Target-specific fundamental types might be new fundamental types or
1445 qualified versions of ordinary fundamental types. Encode new
1446 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1447 is the name used for the type in source code, and @var{n} is the
1448 length of @var{name} in decimal. Encode qualified versions of
1449 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1450 @var{name} is the name used for the type qualifier in source code,
1451 @var{n} is the length of @var{name} as above, and @var{code} is the
1452 code used to represent the unqualified version of this type. (See
1453 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1454 codes.) In both cases the spaces are for clarity; do not include any
1455 spaces in your string.
1457 This hook is applied to types prior to typedef resolution. If the mangled
1458 name for a particular type depends only on that type's main variant, you
1459 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1462 The default version of this hook always returns @code{NULL}, which is
1463 appropriate for a target that does not define any new fundamental
1468 @section Layout of Source Language Data Types
1470 These macros define the sizes and other characteristics of the standard
1471 basic data types used in programs being compiled. Unlike the macros in
1472 the previous section, these apply to specific features of C and related
1473 languages, rather than to fundamental aspects of storage layout.
1475 @defmac INT_TYPE_SIZE
1476 A C expression for the size in bits of the type @code{int} on the
1477 target machine. If you don't define this, the default is one word.
1480 @defmac SHORT_TYPE_SIZE
1481 A C expression for the size in bits of the type @code{short} on the
1482 target machine. If you don't define this, the default is half a word.
1483 (If this would be less than one storage unit, it is rounded up to one
1487 @defmac LONG_TYPE_SIZE
1488 A C expression for the size in bits of the type @code{long} on the
1489 target machine. If you don't define this, the default is one word.
1492 @defmac ADA_LONG_TYPE_SIZE
1493 On some machines, the size used for the Ada equivalent of the type
1494 @code{long} by a native Ada compiler differs from that used by C@. In
1495 that situation, define this macro to be a C expression to be used for
1496 the size of that type. If you don't define this, the default is the
1497 value of @code{LONG_TYPE_SIZE}.
1500 @defmac LONG_LONG_TYPE_SIZE
1501 A C expression for the size in bits of the type @code{long long} on the
1502 target machine. If you don't define this, the default is two
1503 words. If you want to support GNU Ada on your machine, the value of this
1504 macro must be at least 64.
1507 @defmac CHAR_TYPE_SIZE
1508 A C expression for the size in bits of the type @code{char} on the
1509 target machine. If you don't define this, the default is
1510 @code{BITS_PER_UNIT}.
1513 @defmac BOOL_TYPE_SIZE
1514 A C expression for the size in bits of the C++ type @code{bool} and
1515 C99 type @code{_Bool} on the target machine. If you don't define
1516 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1519 @defmac FLOAT_TYPE_SIZE
1520 A C expression for the size in bits of the type @code{float} on the
1521 target machine. If you don't define this, the default is one word.
1524 @defmac DOUBLE_TYPE_SIZE
1525 A C expression for the size in bits of the type @code{double} on the
1526 target machine. If you don't define this, the default is two
1530 @defmac LONG_DOUBLE_TYPE_SIZE
1531 A C expression for the size in bits of the type @code{long double} on
1532 the target machine. If you don't define this, the default is two
1536 @defmac SHORT_FRACT_TYPE_SIZE
1537 A C expression for the size in bits of the type @code{short _Fract} on
1538 the target machine. If you don't define this, the default is
1539 @code{BITS_PER_UNIT}.
1542 @defmac FRACT_TYPE_SIZE
1543 A C expression for the size in bits of the type @code{_Fract} on
1544 the target machine. If you don't define this, the default is
1545 @code{BITS_PER_UNIT * 2}.
1548 @defmac LONG_FRACT_TYPE_SIZE
1549 A C expression for the size in bits of the type @code{long _Fract} on
1550 the target machine. If you don't define this, the default is
1551 @code{BITS_PER_UNIT * 4}.
1554 @defmac LONG_LONG_FRACT_TYPE_SIZE
1555 A C expression for the size in bits of the type @code{long long _Fract} on
1556 the target machine. If you don't define this, the default is
1557 @code{BITS_PER_UNIT * 8}.
1560 @defmac SHORT_ACCUM_TYPE_SIZE
1561 A C expression for the size in bits of the type @code{short _Accum} on
1562 the target machine. If you don't define this, the default is
1563 @code{BITS_PER_UNIT * 2}.
1566 @defmac ACCUM_TYPE_SIZE
1567 A C expression for the size in bits of the type @code{_Accum} on
1568 the target machine. If you don't define this, the default is
1569 @code{BITS_PER_UNIT * 4}.
1572 @defmac LONG_ACCUM_TYPE_SIZE
1573 A C expression for the size in bits of the type @code{long _Accum} on
1574 the target machine. If you don't define this, the default is
1575 @code{BITS_PER_UNIT * 8}.
1578 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1579 A C expression for the size in bits of the type @code{long long _Accum} on
1580 the target machine. If you don't define this, the default is
1581 @code{BITS_PER_UNIT * 16}.
1584 @defmac LIBGCC2_GNU_PREFIX
1585 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1586 hook and should be defined if that hook is overriden to be true. It
1587 causes function names in libgcc to be changed to use a @code{__gnu_}
1588 prefix for their name rather than the default @code{__}. A port which
1589 uses this macro should also arrange to use @file{t-gnu-prefix} in
1590 the libgcc @file{config.host}.
1593 @defmac WIDEST_HARDWARE_FP_SIZE
1594 A C expression for the size in bits of the widest floating-point format
1595 supported by the hardware. If you define this macro, you must specify a
1596 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1597 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1601 @defmac DEFAULT_SIGNED_CHAR
1602 An expression whose value is 1 or 0, according to whether the type
1603 @code{char} should be signed or unsigned by default. The user can
1604 always override this default with the options @option{-fsigned-char}
1605 and @option{-funsigned-char}.
1608 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1609 This target hook should return true if the compiler should give an
1610 @code{enum} type only as many bytes as it takes to represent the range
1611 of possible values of that type. It should return false if all
1612 @code{enum} types should be allocated like @code{int}.
1614 The default is to return false.
1618 A C expression for a string describing the name of the data type to use
1619 for size values. The typedef name @code{size_t} is defined using the
1620 contents of the string.
1622 The string can contain more than one keyword. If so, separate them with
1623 spaces, and write first any length keyword, then @code{unsigned} if
1624 appropriate, and finally @code{int}. The string must exactly match one
1625 of the data type names defined in the function
1626 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1627 You may not omit @code{int} or change the order---that would cause the
1628 compiler to crash on startup.
1630 If you don't define this macro, the default is @code{"long unsigned
1635 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1636 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1637 dealing with size. This macro is a C expression for a string describing
1638 the name of the data type from which the precision of @code{sizetype}
1641 The string has the same restrictions as @code{SIZE_TYPE} string.
1643 If you don't define this macro, the default is @code{SIZE_TYPE}.
1646 @defmac PTRDIFF_TYPE
1647 A C expression for a string describing the name of the data type to use
1648 for the result of subtracting two pointers. The typedef name
1649 @code{ptrdiff_t} is defined using the contents of the string. See
1650 @code{SIZE_TYPE} above for more information.
1652 If you don't define this macro, the default is @code{"long int"}.
1656 A C expression for a string describing the name of the data type to use
1657 for wide characters. The typedef name @code{wchar_t} is defined using
1658 the contents of the string. See @code{SIZE_TYPE} above for more
1661 If you don't define this macro, the default is @code{"int"}.
1664 @defmac WCHAR_TYPE_SIZE
1665 A C expression for the size in bits of the data type for wide
1666 characters. This is used in @code{cpp}, which cannot make use of
1671 A C expression for a string describing the name of the data type to
1672 use for wide characters passed to @code{printf} and returned from
1673 @code{getwc}. The typedef name @code{wint_t} is defined using the
1674 contents of the string. See @code{SIZE_TYPE} above for more
1677 If you don't define this macro, the default is @code{"unsigned int"}.
1681 A C expression for a string describing the name of the data type that
1682 can represent any value of any standard or extended signed integer type.
1683 The typedef name @code{intmax_t} is defined using the contents of the
1684 string. See @code{SIZE_TYPE} above for more information.
1686 If you don't define this macro, the default is the first of
1687 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1688 much precision as @code{long long int}.
1691 @defmac UINTMAX_TYPE
1692 A C expression for a string describing the name of the data type that
1693 can represent any value of any standard or extended unsigned integer
1694 type. The typedef name @code{uintmax_t} is defined using the contents
1695 of the string. See @code{SIZE_TYPE} above for more information.
1697 If you don't define this macro, the default is the first of
1698 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1699 unsigned int"} that has as much precision as @code{long long unsigned
1703 @defmac SIG_ATOMIC_TYPE
1709 @defmacx UINT16_TYPE
1710 @defmacx UINT32_TYPE
1711 @defmacx UINT64_TYPE
1712 @defmacx INT_LEAST8_TYPE
1713 @defmacx INT_LEAST16_TYPE
1714 @defmacx INT_LEAST32_TYPE
1715 @defmacx INT_LEAST64_TYPE
1716 @defmacx UINT_LEAST8_TYPE
1717 @defmacx UINT_LEAST16_TYPE
1718 @defmacx UINT_LEAST32_TYPE
1719 @defmacx UINT_LEAST64_TYPE
1720 @defmacx INT_FAST8_TYPE
1721 @defmacx INT_FAST16_TYPE
1722 @defmacx INT_FAST32_TYPE
1723 @defmacx INT_FAST64_TYPE
1724 @defmacx UINT_FAST8_TYPE
1725 @defmacx UINT_FAST16_TYPE
1726 @defmacx UINT_FAST32_TYPE
1727 @defmacx UINT_FAST64_TYPE
1728 @defmacx INTPTR_TYPE
1729 @defmacx UINTPTR_TYPE
1730 C expressions for the standard types @code{sig_atomic_t},
1731 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1732 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1733 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1734 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1735 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1736 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1737 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1738 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1739 @code{SIZE_TYPE} above for more information.
1741 If any of these macros evaluates to a null pointer, the corresponding
1742 type is not supported; if GCC is configured to provide
1743 @code{<stdint.h>} in such a case, the header provided may not conform
1744 to C99, depending on the type in question. The defaults for all of
1745 these macros are null pointers.
1748 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1749 The C++ compiler represents a pointer-to-member-function with a struct
1756 ptrdiff_t vtable_index;
1763 The C++ compiler must use one bit to indicate whether the function that
1764 will be called through a pointer-to-member-function is virtual.
1765 Normally, we assume that the low-order bit of a function pointer must
1766 always be zero. Then, by ensuring that the vtable_index is odd, we can
1767 distinguish which variant of the union is in use. But, on some
1768 platforms function pointers can be odd, and so this doesn't work. In
1769 that case, we use the low-order bit of the @code{delta} field, and shift
1770 the remainder of the @code{delta} field to the left.
1772 GCC will automatically make the right selection about where to store
1773 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1774 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1775 set such that functions always start at even addresses, but the lowest
1776 bit of pointers to functions indicate whether the function at that
1777 address is in ARM or Thumb mode. If this is the case of your
1778 architecture, you should define this macro to
1779 @code{ptrmemfunc_vbit_in_delta}.
1781 In general, you should not have to define this macro. On architectures
1782 in which function addresses are always even, according to
1783 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1784 @code{ptrmemfunc_vbit_in_pfn}.
1787 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1788 Normally, the C++ compiler uses function pointers in vtables. This
1789 macro allows the target to change to use ``function descriptors''
1790 instead. Function descriptors are found on targets for whom a
1791 function pointer is actually a small data structure. Normally the
1792 data structure consists of the actual code address plus a data
1793 pointer to which the function's data is relative.
1795 If vtables are used, the value of this macro should be the number
1796 of words that the function descriptor occupies.
1799 @defmac TARGET_VTABLE_ENTRY_ALIGN
1800 By default, the vtable entries are void pointers, the so the alignment
1801 is the same as pointer alignment. The value of this macro specifies
1802 the alignment of the vtable entry in bits. It should be defined only
1803 when special alignment is necessary. */
1806 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1807 There are a few non-descriptor entries in the vtable at offsets below
1808 zero. If these entries must be padded (say, to preserve the alignment
1809 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1810 of words in each data entry.
1814 @section Register Usage
1815 @cindex register usage
1817 This section explains how to describe what registers the target machine
1818 has, and how (in general) they can be used.
1820 The description of which registers a specific instruction can use is
1821 done with register classes; see @ref{Register Classes}. For information
1822 on using registers to access a stack frame, see @ref{Frame Registers}.
1823 For passing values in registers, see @ref{Register Arguments}.
1824 For returning values in registers, see @ref{Scalar Return}.
1827 * Register Basics:: Number and kinds of registers.
1828 * Allocation Order:: Order in which registers are allocated.
1829 * Values in Registers:: What kinds of values each reg can hold.
1830 * Leaf Functions:: Renumbering registers for leaf functions.
1831 * Stack Registers:: Handling a register stack such as 80387.
1834 @node Register Basics
1835 @subsection Basic Characteristics of Registers
1837 @c prevent bad page break with this line
1838 Registers have various characteristics.
1840 @defmac FIRST_PSEUDO_REGISTER
1841 Number of hardware registers known to the compiler. They receive
1842 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1843 pseudo register's number really is assigned the number
1844 @code{FIRST_PSEUDO_REGISTER}.
1847 @defmac FIXED_REGISTERS
1848 @cindex fixed register
1849 An initializer that says which registers are used for fixed purposes
1850 all throughout the compiled code and are therefore not available for
1851 general allocation. These would include the stack pointer, the frame
1852 pointer (except on machines where that can be used as a general
1853 register when no frame pointer is needed), the program counter on
1854 machines where that is considered one of the addressable registers,
1855 and any other numbered register with a standard use.
1857 This information is expressed as a sequence of numbers, separated by
1858 commas and surrounded by braces. The @var{n}th number is 1 if
1859 register @var{n} is fixed, 0 otherwise.
1861 The table initialized from this macro, and the table initialized by
1862 the following one, may be overridden at run time either automatically,
1863 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1864 the user with the command options @option{-ffixed-@var{reg}},
1865 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1868 @defmac CALL_USED_REGISTERS
1869 @cindex call-used register
1870 @cindex call-clobbered register
1871 @cindex call-saved register
1872 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1873 clobbered (in general) by function calls as well as for fixed
1874 registers. This macro therefore identifies the registers that are not
1875 available for general allocation of values that must live across
1878 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1879 automatically saves it on function entry and restores it on function
1880 exit, if the register is used within the function.
1883 @defmac CALL_REALLY_USED_REGISTERS
1884 @cindex call-used register
1885 @cindex call-clobbered register
1886 @cindex call-saved register
1887 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1888 that the entire set of @code{FIXED_REGISTERS} be included.
1889 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1890 This macro is optional. If not specified, it defaults to the value
1891 of @code{CALL_USED_REGISTERS}.
1894 @cindex call-used register
1895 @cindex call-clobbered register
1896 @cindex call-saved register
1897 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_CALL_PART_CLOBBERED (rtx_insn *@var{insn}, unsigned int @var{regno}, machine_mode @var{mode})
1898 This hook should return true if @var{regno} is partly call-saved and
1899 partly call-clobbered, and if a value of mode @var{mode} would be partly
1900 clobbered by call instruction @var{insn}. If @var{insn} is NULL then it
1901 should return true if any call could partly clobber the register.
1902 For example, if the low 32 bits of @var{regno} are preserved across a call
1903 but higher bits are clobbered, this hook should return true for a 64-bit
1904 mode but false for a 32-bit mode.
1906 The default implementation returns false, which is correct
1907 for targets that don't have partly call-clobbered registers.
1910 @deftypefn {Target Hook} void TARGET_REMOVE_EXTRA_CALL_PRESERVED_REGS (rtx_insn *@var{insn}, HARD_REG_SET *@var{used_regs})
1911 This hook removes registers from the set of call-clobbered registers
1912 in @var{used_regs} if, contrary to the default rules, something guarantees
1913 that @samp{insn} preserves those registers. For example, some targets
1914 support variant ABIs in which functions preserve more registers than
1915 normal functions would. Removing those extra registers from @var{used_regs}
1916 can lead to better register allocation.
1918 The default implementation does nothing, which is always safe.
1919 Defining the hook is purely an optimization.
1922 @deftypefn {Target Hook} {rtx_insn *} TARGET_RETURN_CALL_WITH_MAX_CLOBBERS (rtx_insn *@var{call_1}, rtx_insn *@var{call_2})
1923 This hook returns a pointer to the call that partially clobbers the
1924 most registers. If a platform supports multiple ABIs where the registers
1925 that are partially clobbered may vary, this function compares two
1926 calls and returns a pointer to the one that clobbers the most registers.
1927 If both calls clobber the same registers, @var{call_1} must be returned.
1929 The registers clobbered in different ABIs must be a proper subset or
1930 superset of all other ABIs. @var{call_1} must always be a call insn,
1931 call_2 may be NULL or a call insn.
1934 @deftypefn {Target Hook} {const char *} TARGET_GET_MULTILIB_ABI_NAME (void)
1935 This hook returns name of multilib ABI name.
1939 @findex call_used_regs
1942 @findex reg_class_contents
1943 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1944 This hook may conditionally modify five variables
1945 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1946 @code{reg_names}, and @code{reg_class_contents}, to take into account
1947 any dependence of these register sets on target flags. The first three
1948 of these are of type @code{char []} (interpreted as boolean vectors).
1949 @code{global_regs} is a @code{const char *[]}, and
1950 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1951 called, @code{fixed_regs}, @code{call_used_regs},
1952 @code{reg_class_contents}, and @code{reg_names} have been initialized
1953 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1954 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1955 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1956 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1957 command options have been applied.
1959 @cindex disabling certain registers
1960 @cindex controlling register usage
1961 If the usage of an entire class of registers depends on the target
1962 flags, you may indicate this to GCC by using this macro to modify
1963 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1964 registers in the classes which should not be used by GCC@. Also make
1965 @code{define_register_constraint}s return @code{NO_REGS} for constraints
1966 that shouldn't be used.
1968 (However, if this class is not included in @code{GENERAL_REGS} and all
1969 of the insn patterns whose constraints permit this class are
1970 controlled by target switches, then GCC will automatically avoid using
1971 these registers when the target switches are opposed to them.)
1974 @defmac INCOMING_REGNO (@var{out})
1975 Define this macro if the target machine has register windows. This C
1976 expression returns the register number as seen by the called function
1977 corresponding to the register number @var{out} as seen by the calling
1978 function. Return @var{out} if register number @var{out} is not an
1982 @defmac OUTGOING_REGNO (@var{in})
1983 Define this macro if the target machine has register windows. This C
1984 expression returns the register number as seen by the calling function
1985 corresponding to the register number @var{in} as seen by the called
1986 function. Return @var{in} if register number @var{in} is not an inbound
1990 @defmac LOCAL_REGNO (@var{regno})
1991 Define this macro if the target machine has register windows. This C
1992 expression returns true if the register is call-saved but is in the
1993 register window. Unlike most call-saved registers, such registers
1994 need not be explicitly restored on function exit or during non-local
1999 If the program counter has a register number, define this as that
2000 register number. Otherwise, do not define it.
2003 @node Allocation Order
2004 @subsection Order of Allocation of Registers
2005 @cindex order of register allocation
2006 @cindex register allocation order
2008 @c prevent bad page break with this line
2009 Registers are allocated in order.
2011 @defmac REG_ALLOC_ORDER
2012 If defined, an initializer for a vector of integers, containing the
2013 numbers of hard registers in the order in which GCC should prefer
2014 to use them (from most preferred to least).
2016 If this macro is not defined, registers are used lowest numbered first
2017 (all else being equal).
2019 One use of this macro is on machines where the highest numbered
2020 registers must always be saved and the save-multiple-registers
2021 instruction supports only sequences of consecutive registers. On such
2022 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2023 the highest numbered allocable register first.
2026 @defmac ADJUST_REG_ALLOC_ORDER
2027 A C statement (sans semicolon) to choose the order in which to allocate
2028 hard registers for pseudo-registers local to a basic block.
2030 Store the desired register order in the array @code{reg_alloc_order}.
2031 Element 0 should be the register to allocate first; element 1, the next
2032 register; and so on.
2034 The macro body should not assume anything about the contents of
2035 @code{reg_alloc_order} before execution of the macro.
2037 On most machines, it is not necessary to define this macro.
2040 @defmac HONOR_REG_ALLOC_ORDER
2041 Normally, IRA tries to estimate the costs for saving a register in the
2042 prologue and restoring it in the epilogue. This discourages it from
2043 using call-saved registers. If a machine wants to ensure that IRA
2044 allocates registers in the order given by REG_ALLOC_ORDER even if some
2045 call-saved registers appear earlier than call-used ones, then define this
2046 macro as a C expression to nonzero. Default is 0.
2049 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2050 In some case register allocation order is not enough for the
2051 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2052 If this macro is defined, it should return a floating point value
2053 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2054 be increased by approximately the pseudo's usage frequency times the
2055 value returned by this macro. Not defining this macro is equivalent
2056 to having it always return @code{0.0}.
2058 On most machines, it is not necessary to define this macro.
2061 @node Values in Registers
2062 @subsection How Values Fit in Registers
2064 This section discusses the macros that describe which kinds of values
2065 (specifically, which machine modes) each register can hold, and how many
2066 consecutive registers are needed for a given mode.
2068 @deftypefn {Target Hook} {unsigned int} TARGET_HARD_REGNO_NREGS (unsigned int @var{regno}, machine_mode @var{mode})
2069 This hook returns the number of consecutive hard registers, starting
2070 at register number @var{regno}, required to hold a value of mode
2071 @var{mode}. This hook must never return zero, even if a register
2072 cannot hold the requested mode - indicate that with
2073 @code{TARGET_HARD_REGNO_MODE_OK} and/or
2074 @code{TARGET_CAN_CHANGE_MODE_CLASS} instead.
2076 The default definition returns the number of words in @var{mode}.
2079 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2080 A C expression that is nonzero if a value of mode @var{mode}, stored
2081 in memory, ends with padding that causes it to take up more space than
2082 in registers starting at register number @var{regno} (as determined by
2083 multiplying GCC's notion of the size of the register when containing
2084 this mode by the number of registers returned by
2085 @code{TARGET_HARD_REGNO_NREGS}). By default this is zero.
2087 For example, if a floating-point value is stored in three 32-bit
2088 registers but takes up 128 bits in memory, then this would be
2091 This macros only needs to be defined if there are cases where
2092 @code{subreg_get_info}
2093 would otherwise wrongly determine that a @code{subreg} can be
2094 represented by an offset to the register number, when in fact such a
2095 @code{subreg} would contain some of the padding not stored in
2096 registers and so not be representable.
2099 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2100 For values of @var{regno} and @var{mode} for which
2101 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2102 returning the greater number of registers required to hold the value
2103 including any padding. In the example above, the value would be four.
2106 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2107 Define this macro if the natural size of registers that hold values
2108 of mode @var{mode} is not the word size. It is a C expression that
2109 should give the natural size in bytes for the specified mode. It is
2110 used by the register allocator to try to optimize its results. This
2111 happens for example on SPARC 64-bit where the natural size of
2112 floating-point registers is still 32-bit.
2115 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_MODE_OK (unsigned int @var{regno}, machine_mode @var{mode})
2116 This hook returns true if it is permissible to store a value
2117 of mode @var{mode} in hard register number @var{regno} (or in several
2118 registers starting with that one). The default definition returns true
2121 You need not include code to check for the numbers of fixed registers,
2122 because the allocation mechanism considers them to be always occupied.
2124 @cindex register pairs
2125 On some machines, double-precision values must be kept in even/odd
2126 register pairs. You can implement that by defining this hook to reject
2127 odd register numbers for such modes.
2129 The minimum requirement for a mode to be OK in a register is that the
2130 @samp{mov@var{mode}} instruction pattern support moves between the
2131 register and other hard register in the same class and that moving a
2132 value into the register and back out not alter it.
2134 Since the same instruction used to move @code{word_mode} will work for
2135 all narrower integer modes, it is not necessary on any machine for
2136 this hook to distinguish between these modes, provided you define
2137 patterns @samp{movhi}, etc., to take advantage of this. This is
2138 useful because of the interaction between @code{TARGET_HARD_REGNO_MODE_OK}
2139 and @code{TARGET_MODES_TIEABLE_P}; it is very desirable for all integer
2140 modes to be tieable.
2142 Many machines have special registers for floating point arithmetic.
2143 Often people assume that floating point machine modes are allowed only
2144 in floating point registers. This is not true. Any registers that
2145 can hold integers can safely @emph{hold} a floating point machine
2146 mode, whether or not floating arithmetic can be done on it in those
2147 registers. Integer move instructions can be used to move the values.
2149 On some machines, though, the converse is true: fixed-point machine
2150 modes may not go in floating registers. This is true if the floating
2151 registers normalize any value stored in them, because storing a
2152 non-floating value there would garble it. In this case,
2153 @code{TARGET_HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2154 floating registers. But if the floating registers do not automatically
2155 normalize, if you can store any bit pattern in one and retrieve it
2156 unchanged without a trap, then any machine mode may go in a floating
2157 register, so you can define this hook to say so.
2159 The primary significance of special floating registers is rather that
2160 they are the registers acceptable in floating point arithmetic
2161 instructions. However, this is of no concern to
2162 @code{TARGET_HARD_REGNO_MODE_OK}. You handle it by writing the proper
2163 constraints for those instructions.
2165 On some machines, the floating registers are especially slow to access,
2166 so that it is better to store a value in a stack frame than in such a
2167 register if floating point arithmetic is not being done. As long as the
2168 floating registers are not in class @code{GENERAL_REGS}, they will not
2169 be used unless some pattern's constraint asks for one.
2172 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2173 A C expression that is nonzero if it is OK to rename a hard register
2174 @var{from} to another hard register @var{to}.
2176 One common use of this macro is to prevent renaming of a register to
2177 another register that is not saved by a prologue in an interrupt
2180 The default is always nonzero.
2183 @deftypefn {Target Hook} bool TARGET_MODES_TIEABLE_P (machine_mode @var{mode1}, machine_mode @var{mode2})
2184 This hook returns true if a value of mode @var{mode1} is accessible
2185 in mode @var{mode2} without copying.
2187 If @code{TARGET_HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2188 @code{TARGET_HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always
2189 the same for any @var{r}, then
2190 @code{TARGET_MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2191 should be true. If they differ for any @var{r}, you should define
2192 this hook to return false unless some other mechanism ensures the
2193 accessibility of the value in a narrower mode.
2195 You should define this hook to return true in as many cases as
2196 possible since doing so will allow GCC to perform better register
2197 allocation. The default definition returns true unconditionally.
2200 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2201 This target hook should return @code{true} if it is OK to use a hard register
2202 @var{regno} as scratch reg in peephole2.
2204 One common use of this macro is to prevent using of a register that
2205 is not saved by a prologue in an interrupt handler.
2207 The default version of this hook always returns @code{true}.
2210 @defmac AVOID_CCMODE_COPIES
2211 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2212 registers. You should only define this macro if support for copying to/from
2213 @code{CCmode} is incomplete.
2216 @node Leaf Functions
2217 @subsection Handling Leaf Functions
2219 @cindex leaf functions
2220 @cindex functions, leaf
2221 On some machines, a leaf function (i.e., one which makes no calls) can run
2222 more efficiently if it does not make its own register window. Often this
2223 means it is required to receive its arguments in the registers where they
2224 are passed by the caller, instead of the registers where they would
2227 The special treatment for leaf functions generally applies only when
2228 other conditions are met; for example, often they may use only those
2229 registers for its own variables and temporaries. We use the term ``leaf
2230 function'' to mean a function that is suitable for this special
2231 handling, so that functions with no calls are not necessarily ``leaf
2234 GCC assigns register numbers before it knows whether the function is
2235 suitable for leaf function treatment. So it needs to renumber the
2236 registers in order to output a leaf function. The following macros
2239 @defmac LEAF_REGISTERS
2240 Name of a char vector, indexed by hard register number, which
2241 contains 1 for a register that is allowable in a candidate for leaf
2244 If leaf function treatment involves renumbering the registers, then the
2245 registers marked here should be the ones before renumbering---those that
2246 GCC would ordinarily allocate. The registers which will actually be
2247 used in the assembler code, after renumbering, should not be marked with 1
2250 Define this macro only if the target machine offers a way to optimize
2251 the treatment of leaf functions.
2254 @defmac LEAF_REG_REMAP (@var{regno})
2255 A C expression whose value is the register number to which @var{regno}
2256 should be renumbered, when a function is treated as a leaf function.
2258 If @var{regno} is a register number which should not appear in a leaf
2259 function before renumbering, then the expression should yield @minus{}1, which
2260 will cause the compiler to abort.
2262 Define this macro only if the target machine offers a way to optimize the
2263 treatment of leaf functions, and registers need to be renumbered to do
2267 @findex current_function_is_leaf
2268 @findex current_function_uses_only_leaf_regs
2269 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2270 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2271 specially. They can test the C variable @code{current_function_is_leaf}
2272 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2273 set prior to local register allocation and is valid for the remaining
2274 compiler passes. They can also test the C variable
2275 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2276 functions which only use leaf registers.
2277 @code{current_function_uses_only_leaf_regs} is valid after all passes
2278 that modify the instructions have been run and is only useful if
2279 @code{LEAF_REGISTERS} is defined.
2280 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2281 @c of the next paragraph?! --mew 2feb93
2283 @node Stack Registers
2284 @subsection Registers That Form a Stack
2286 There are special features to handle computers where some of the
2287 ``registers'' form a stack. Stack registers are normally written by
2288 pushing onto the stack, and are numbered relative to the top of the
2291 Currently, GCC can only handle one group of stack-like registers, and
2292 they must be consecutively numbered. Furthermore, the existing
2293 support for stack-like registers is specific to the 80387 floating
2294 point coprocessor. If you have a new architecture that uses
2295 stack-like registers, you will need to do substantial work on
2296 @file{reg-stack.c} and write your machine description to cooperate
2297 with it, as well as defining these macros.
2300 Define this if the machine has any stack-like registers.
2303 @defmac STACK_REG_COVER_CLASS
2304 This is a cover class containing the stack registers. Define this if
2305 the machine has any stack-like registers.
2308 @defmac FIRST_STACK_REG
2309 The number of the first stack-like register. This one is the top
2313 @defmac LAST_STACK_REG
2314 The number of the last stack-like register. This one is the bottom of
2318 @node Register Classes
2319 @section Register Classes
2320 @cindex register class definitions
2321 @cindex class definitions, register
2323 On many machines, the numbered registers are not all equivalent.
2324 For example, certain registers may not be allowed for indexed addressing;
2325 certain registers may not be allowed in some instructions. These machine
2326 restrictions are described to the compiler using @dfn{register classes}.
2328 You define a number of register classes, giving each one a name and saying
2329 which of the registers belong to it. Then you can specify register classes
2330 that are allowed as operands to particular instruction patterns.
2334 In general, each register will belong to several classes. In fact, one
2335 class must be named @code{ALL_REGS} and contain all the registers. Another
2336 class must be named @code{NO_REGS} and contain no registers. Often the
2337 union of two classes will be another class; however, this is not required.
2339 @findex GENERAL_REGS
2340 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2341 terribly special about the name, but the operand constraint letters
2342 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2343 the same as @code{ALL_REGS}, just define it as a macro which expands
2346 Order the classes so that if class @var{x} is contained in class @var{y}
2347 then @var{x} has a lower class number than @var{y}.
2349 The way classes other than @code{GENERAL_REGS} are specified in operand
2350 constraints is through machine-dependent operand constraint letters.
2351 You can define such letters to correspond to various classes, then use
2352 them in operand constraints.
2354 You must define the narrowest register classes for allocatable
2355 registers, so that each class either has no subclasses, or that for
2356 some mode, the move cost between registers within the class is
2357 cheaper than moving a register in the class to or from memory
2360 You should define a class for the union of two classes whenever some
2361 instruction allows both classes. For example, if an instruction allows
2362 either a floating point (coprocessor) register or a general register for a
2363 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2364 which includes both of them. Otherwise you will get suboptimal code,
2365 or even internal compiler errors when reload cannot find a register in the
2366 class computed via @code{reg_class_subunion}.
2368 You must also specify certain redundant information about the register
2369 classes: for each class, which classes contain it and which ones are
2370 contained in it; for each pair of classes, the largest class contained
2373 When a value occupying several consecutive registers is expected in a
2374 certain class, all the registers used must belong to that class.
2375 Therefore, register classes cannot be used to enforce a requirement for
2376 a register pair to start with an even-numbered register. The way to
2377 specify this requirement is with @code{TARGET_HARD_REGNO_MODE_OK}.
2379 Register classes used for input-operands of bitwise-and or shift
2380 instructions have a special requirement: each such class must have, for
2381 each fixed-point machine mode, a subclass whose registers can transfer that
2382 mode to or from memory. For example, on some machines, the operations for
2383 single-byte values (@code{QImode}) are limited to certain registers. When
2384 this is so, each register class that is used in a bitwise-and or shift
2385 instruction must have a subclass consisting of registers from which
2386 single-byte values can be loaded or stored. This is so that
2387 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2389 @deftp {Data type} {enum reg_class}
2390 An enumerated type that must be defined with all the register class names
2391 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2392 must be the last register class, followed by one more enumerated value,
2393 @code{LIM_REG_CLASSES}, which is not a register class but rather
2394 tells how many classes there are.
2396 Each register class has a number, which is the value of casting
2397 the class name to type @code{int}. The number serves as an index
2398 in many of the tables described below.
2401 @defmac N_REG_CLASSES
2402 The number of distinct register classes, defined as follows:
2405 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2409 @defmac REG_CLASS_NAMES
2410 An initializer containing the names of the register classes as C string
2411 constants. These names are used in writing some of the debugging dumps.
2414 @defmac REG_CLASS_CONTENTS
2415 An initializer containing the contents of the register classes, as integers
2416 which are bit masks. The @var{n}th integer specifies the contents of class
2417 @var{n}. The way the integer @var{mask} is interpreted is that
2418 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2420 When the machine has more than 32 registers, an integer does not suffice.
2421 Then the integers are replaced by sub-initializers, braced groupings containing
2422 several integers. Each sub-initializer must be suitable as an initializer
2423 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2424 In this situation, the first integer in each sub-initializer corresponds to
2425 registers 0 through 31, the second integer to registers 32 through 63, and
2429 @defmac REGNO_REG_CLASS (@var{regno})
2430 A C expression whose value is a register class containing hard register
2431 @var{regno}. In general there is more than one such class; choose a class
2432 which is @dfn{minimal}, meaning that no smaller class also contains the
2436 @defmac BASE_REG_CLASS
2437 A macro whose definition is the name of the class to which a valid
2438 base register must belong. A base register is one used in an address
2439 which is the register value plus a displacement.
2442 @defmac MODE_BASE_REG_CLASS (@var{mode})
2443 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2444 the selection of a base register in a mode dependent manner. If
2445 @var{mode} is VOIDmode then it should return the same value as
2446 @code{BASE_REG_CLASS}.
2449 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2450 A C expression whose value is the register class to which a valid
2451 base register must belong in order to be used in a base plus index
2452 register address. You should define this macro if base plus index
2453 addresses have different requirements than other base register uses.
2456 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2457 A C expression whose value is the register class to which a valid
2458 base register for a memory reference in mode @var{mode} to address
2459 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2460 define the context in which the base register occurs. @var{outer_code} is
2461 the code of the immediately enclosing expression (@code{MEM} for the top level
2462 of an address, @code{ADDRESS} for something that occurs in an
2463 @code{address_operand}). @var{index_code} is the code of the corresponding
2464 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2467 @defmac INDEX_REG_CLASS
2468 A macro whose definition is the name of the class to which a valid
2469 index register must belong. An index register is one used in an
2470 address where its value is either multiplied by a scale factor or
2471 added to another register (as well as added to a displacement).
2474 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2475 A C expression which is nonzero if register number @var{num} is
2476 suitable for use as a base register in operand addresses.
2479 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2480 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2481 that expression may examine the mode of the memory reference in
2482 @var{mode}. You should define this macro if the mode of the memory
2483 reference affects whether a register may be used as a base register. If
2484 you define this macro, the compiler will use it instead of
2485 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2486 addresses that appear outside a @code{MEM}, i.e., as an
2487 @code{address_operand}.
2490 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2491 A C expression which is nonzero if register number @var{num} is suitable for
2492 use as a base register in base plus index operand addresses, accessing
2493 memory in mode @var{mode}. It may be either a suitable hard register or a
2494 pseudo register that has been allocated such a hard register. You should
2495 define this macro if base plus index addresses have different requirements
2496 than other base register uses.
2498 Use of this macro is deprecated; please use the more general
2499 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2502 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2503 A C expression which is nonzero if register number @var{num} is
2504 suitable for use as a base register in operand addresses, accessing
2505 memory in mode @var{mode} in address space @var{address_space}.
2506 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2507 that that expression may examine the context in which the register
2508 appears in the memory reference. @var{outer_code} is the code of the
2509 immediately enclosing expression (@code{MEM} if at the top level of the
2510 address, @code{ADDRESS} for something that occurs in an
2511 @code{address_operand}). @var{index_code} is the code of the
2512 corresponding index expression if @var{outer_code} is @code{PLUS};
2513 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2514 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2517 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2518 A C expression which is nonzero if register number @var{num} is
2519 suitable for use as an index register in operand addresses. It may be
2520 either a suitable hard register or a pseudo register that has been
2521 allocated such a hard register.
2523 The difference between an index register and a base register is that
2524 the index register may be scaled. If an address involves the sum of
2525 two registers, neither one of them scaled, then either one may be
2526 labeled the ``base'' and the other the ``index''; but whichever
2527 labeling is used must fit the machine's constraints of which registers
2528 may serve in each capacity. The compiler will try both labelings,
2529 looking for one that is valid, and will reload one or both registers
2530 only if neither labeling works.
2533 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2534 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.
2537 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2538 A target hook that places additional restrictions on the register class
2539 to use when it is necessary to copy value @var{x} into a register in class
2540 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2541 another, smaller class.
2543 The default version of this hook always returns value of @code{rclass} argument.
2545 Sometimes returning a more restrictive class makes better code. For
2546 example, on the 68000, when @var{x} is an integer constant that is in range
2547 for a @samp{moveq} instruction, the value of this macro is always
2548 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2549 Requiring a data register guarantees that a @samp{moveq} will be used.
2551 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2552 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2553 loaded into some register class. By returning @code{NO_REGS} you can
2554 force @var{x} into a memory location. For example, rs6000 can load
2555 immediate values into general-purpose registers, but does not have an
2556 instruction for loading an immediate value into a floating-point
2557 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2558 @var{x} is a floating-point constant. If the constant can't be loaded
2559 into any kind of register, code generation will be better if
2560 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2561 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2563 If an insn has pseudos in it after register allocation, reload will go
2564 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2565 to find the best one. Returning @code{NO_REGS}, in this case, makes
2566 reload add a @code{!} in front of the constraint: the x86 back-end uses
2567 this feature to discourage usage of 387 registers when math is done in
2568 the SSE registers (and vice versa).
2571 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2572 A C expression that places additional restrictions on the register class
2573 to use when it is necessary to copy value @var{x} into a register in class
2574 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2575 another, smaller class. On many machines, the following definition is
2579 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2582 Sometimes returning a more restrictive class makes better code. For
2583 example, on the 68000, when @var{x} is an integer constant that is in range
2584 for a @samp{moveq} instruction, the value of this macro is always
2585 @code{DATA_REGS} as long as @var{class} includes the data registers.
2586 Requiring a data register guarantees that a @samp{moveq} will be used.
2588 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2589 @var{class} is if @var{x} is a legitimate constant which cannot be
2590 loaded into some register class. By returning @code{NO_REGS} you can
2591 force @var{x} into a memory location. For example, rs6000 can load
2592 immediate values into general-purpose registers, but does not have an
2593 instruction for loading an immediate value into a floating-point
2594 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2595 @var{x} is a floating-point constant. If the constant cannot be loaded
2596 into any kind of register, code generation will be better if
2597 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2598 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2600 If an insn has pseudos in it after register allocation, reload will go
2601 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2602 to find the best one. Returning @code{NO_REGS}, in this case, makes
2603 reload add a @code{!} in front of the constraint: the x86 back-end uses
2604 this feature to discourage usage of 387 registers when math is done in
2605 the SSE registers (and vice versa).
2608 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2609 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2612 The default version of this hook always returns value of @code{rclass}
2615 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2616 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2619 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2620 A C expression that places additional restrictions on the register class
2621 to use when it is necessary to be able to hold a value of mode
2622 @var{mode} in a reload register for which class @var{class} would
2625 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2626 there are certain modes that simply cannot go in certain reload classes.
2628 The value is a register class; perhaps @var{class}, or perhaps another,
2631 Don't define this macro unless the target machine has limitations which
2632 require the macro to do something nontrivial.
2635 @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})
2636 Many machines have some registers that cannot be copied directly to or
2637 from memory or even from other types of registers. An example is the
2638 @samp{MQ} register, which on most machines, can only be copied to or
2639 from general registers, but not memory. Below, we shall be using the
2640 term 'intermediate register' when a move operation cannot be performed
2641 directly, but has to be done by copying the source into the intermediate
2642 register first, and then copying the intermediate register to the
2643 destination. An intermediate register always has the same mode as
2644 source and destination. Since it holds the actual value being copied,
2645 reload might apply optimizations to re-use an intermediate register
2646 and eliding the copy from the source when it can determine that the
2647 intermediate register still holds the required value.
2649 Another kind of secondary reload is required on some machines which
2650 allow copying all registers to and from memory, but require a scratch
2651 register for stores to some memory locations (e.g., those with symbolic
2652 address on the RT, and those with certain symbolic address on the SPARC
2653 when compiling PIC)@. Scratch registers need not have the same mode
2654 as the value being copied, and usually hold a different value than
2655 that being copied. Special patterns in the md file are needed to
2656 describe how the copy is performed with the help of the scratch register;
2657 these patterns also describe the number, register class(es) and mode(s)
2658 of the scratch register(s).
2660 In some cases, both an intermediate and a scratch register are required.
2662 For input reloads, this target hook is called with nonzero @var{in_p},
2663 and @var{x} is an rtx that needs to be copied to a register of class
2664 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2665 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2666 needs to be copied to rtx @var{x} in @var{reload_mode}.
2668 If copying a register of @var{reload_class} from/to @var{x} requires
2669 an intermediate register, the hook @code{secondary_reload} should
2670 return the register class required for this intermediate register.
2671 If no intermediate register is required, it should return NO_REGS.
2672 If more than one intermediate register is required, describe the one
2673 that is closest in the copy chain to the reload register.
2675 If scratch registers are needed, you also have to describe how to
2676 perform the copy from/to the reload register to/from this
2677 closest intermediate register. Or if no intermediate register is
2678 required, but still a scratch register is needed, describe the
2679 copy from/to the reload register to/from the reload operand @var{x}.
2681 You do this by setting @code{sri->icode} to the instruction code of a pattern
2682 in the md file which performs the move. Operands 0 and 1 are the output
2683 and input of this copy, respectively. Operands from operand 2 onward are
2684 for scratch operands. These scratch operands must have a mode, and a
2685 single-register-class
2686 @c [later: or memory]
2689 When an intermediate register is used, the @code{secondary_reload}
2690 hook will be called again to determine how to copy the intermediate
2691 register to/from the reload operand @var{x}, so your hook must also
2692 have code to handle the register class of the intermediate operand.
2694 @c [For later: maybe we'll allow multi-alternative reload patterns -
2695 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2696 @c and match the constraints of input and output to determine the required
2697 @c alternative. A restriction would be that constraints used to match
2698 @c against reloads registers would have to be written as register class
2699 @c constraints, or we need a new target macro / hook that tells us if an
2700 @c arbitrary constraint can match an unknown register of a given class.
2701 @c Such a macro / hook would also be useful in other places.]
2704 @var{x} might be a pseudo-register or a @code{subreg} of a
2705 pseudo-register, which could either be in a hard register or in memory.
2706 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2707 in memory and the hard register number if it is in a register.
2709 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2710 currently not supported. For the time being, you will have to continue
2711 to use @code{TARGET_SECONDARY_MEMORY_NEEDED} for that purpose.
2713 @code{copy_cost} also uses this target hook to find out how values are
2714 copied. If you want it to include some extra cost for the need to allocate
2715 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2716 Or if two dependent moves are supposed to have a lower cost than the sum
2717 of the individual moves due to expected fortuitous scheduling and/or special
2718 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2721 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2722 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2723 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2724 These macros are obsolete, new ports should use the target hook
2725 @code{TARGET_SECONDARY_RELOAD} instead.
2727 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2728 target hook. Older ports still define these macros to indicate to the
2729 reload phase that it may
2730 need to allocate at least one register for a reload in addition to the
2731 register to contain the data. Specifically, if copying @var{x} to a
2732 register @var{class} in @var{mode} requires an intermediate register,
2733 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2734 largest register class all of whose registers can be used as
2735 intermediate registers or scratch registers.
2737 If copying a register @var{class} in @var{mode} to @var{x} requires an
2738 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2739 was supposed to be defined be defined to return the largest register
2740 class required. If the
2741 requirements for input and output reloads were the same, the macro
2742 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2745 The values returned by these macros are often @code{GENERAL_REGS}.
2746 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2747 can be directly copied to or from a register of @var{class} in
2748 @var{mode} without requiring a scratch register. Do not define this
2749 macro if it would always return @code{NO_REGS}.
2751 If a scratch register is required (either with or without an
2752 intermediate register), you were supposed to define patterns for
2753 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2754 (@pxref{Standard Names}. These patterns, which were normally
2755 implemented with a @code{define_expand}, should be similar to the
2756 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2759 These patterns need constraints for the reload register and scratch
2761 contain a single register class. If the original reload register (whose
2762 class is @var{class}) can meet the constraint given in the pattern, the
2763 value returned by these macros is used for the class of the scratch
2764 register. Otherwise, two additional reload registers are required.
2765 Their classes are obtained from the constraints in the insn pattern.
2767 @var{x} might be a pseudo-register or a @code{subreg} of a
2768 pseudo-register, which could either be in a hard register or in memory.
2769 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2770 in memory and the hard register number if it is in a register.
2772 These macros should not be used in the case where a particular class of
2773 registers can only be copied to memory and not to another class of
2774 registers. In that case, secondary reload registers are not needed and
2775 would not be helpful. Instead, a stack location must be used to perform
2776 the copy and the @code{mov@var{m}} pattern should use memory as an
2777 intermediate storage. This case often occurs between floating-point and
2781 @deftypefn {Target Hook} bool TARGET_SECONDARY_MEMORY_NEEDED (machine_mode @var{mode}, reg_class_t @var{class1}, reg_class_t @var{class2})
2782 Certain machines have the property that some registers cannot be copied
2783 to some other registers without using memory. Define this hook on
2784 those machines to return true if objects of mode @var{m} in registers
2785 of @var{class1} can only be copied to registers of class @var{class2} by
2786 storing a register of @var{class1} into memory and loading that memory
2787 location into a register of @var{class2}. The default definition returns
2788 false for all inputs.
2791 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2792 Normally when @code{TARGET_SECONDARY_MEMORY_NEEDED} is defined, the compiler
2793 allocates a stack slot for a memory location needed for register copies.
2794 If this macro is defined, the compiler instead uses the memory location
2795 defined by this macro.
2797 Do not define this macro if you do not define
2798 @code{TARGET_SECONDARY_MEMORY_NEEDED}.
2801 @deftypefn {Target Hook} machine_mode TARGET_SECONDARY_MEMORY_NEEDED_MODE (machine_mode @var{mode})
2802 If @code{TARGET_SECONDARY_MEMORY_NEEDED} tells the compiler to use memory
2803 when moving between two particular registers of mode @var{mode},
2804 this hook specifies the mode that the memory should have.
2806 The default depends on @code{TARGET_LRA_P}. Without LRA, the default
2807 is to use a word-sized mode for integral modes that are smaller than a
2808 a word. This is right thing to do on most machines because it ensures
2809 that all bits of the register are copied and prevents accesses to the
2810 registers in a narrower mode, which some machines prohibit for
2811 floating-point registers.
2813 However, this default behavior is not correct on some machines, such as
2814 the DEC Alpha, that store short integers in floating-point registers
2815 differently than in integer registers. On those machines, the default
2816 widening will not work correctly and you must define this hook to
2817 suppress that widening in some cases. See the file @file{alpha.c} for
2820 With LRA, the default is to use @var{mode} unmodified.
2823 @deftypefn {Target Hook} void TARGET_SELECT_EARLY_REMAT_MODES (sbitmap @var{modes})
2824 On some targets, certain modes cannot be held in registers around a
2825 standard ABI call and are relatively expensive to spill to the stack.
2826 The early rematerialization pass can help in such cases by aggressively
2827 recomputing values after calls, so that they don't need to be spilled.
2829 This hook returns the set of such modes by setting the associated bits
2830 in @var{modes}. The default implementation selects no modes, which has
2831 the effect of disabling the early rematerialization pass.
2834 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2835 A target hook which returns @code{true} if pseudos that have been assigned
2836 to registers of class @var{rclass} would likely be spilled because
2837 registers of @var{rclass} are needed for spill registers.
2839 The default version of this target hook returns @code{true} if @var{rclass}
2840 has exactly one register and @code{false} otherwise. On most machines, this
2841 default should be used. For generally register-starved machines, such as
2842 i386, or machines with right register constraints, such as SH, this hook
2843 can be used to avoid excessive spilling.
2845 This hook is also used by some of the global intra-procedural code
2846 transformations to throtle code motion, to avoid increasing register
2850 @deftypefn {Target Hook} {unsigned char} TARGET_CLASS_MAX_NREGS (reg_class_t @var{rclass}, machine_mode @var{mode})
2851 A target hook returns the maximum number of consecutive registers
2852 of class @var{rclass} needed to hold a value of mode @var{mode}.
2854 This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}.
2855 In fact, the value returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass},
2856 @var{mode})} target hook should be the maximum value of
2857 @code{TARGET_HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno}
2858 values in the class @var{rclass}.
2860 This target hook helps control the handling of multiple-word values
2863 The default version of this target hook returns the size of @var{mode}
2867 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2868 A C expression for the maximum number of consecutive registers
2869 of class @var{class} needed to hold a value of mode @var{mode}.
2871 This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}. In fact,
2872 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2873 should be the maximum value of @code{TARGET_HARD_REGNO_NREGS (@var{regno},
2874 @var{mode})} for all @var{regno} values in the class @var{class}.
2876 This macro helps control the handling of multiple-word values
2880 @deftypefn {Target Hook} bool TARGET_CAN_CHANGE_MODE_CLASS (machine_mode @var{from}, machine_mode @var{to}, reg_class_t @var{rclass})
2881 This hook returns true if it is possible to bitcast values held in
2882 registers of class @var{rclass} from mode @var{from} to mode @var{to}
2883 and if doing so preserves the low-order bits that are common to both modes.
2884 The result is only meaningful if @var{rclass} has registers that can hold
2885 both @code{from} and @code{to}. The default implementation returns true.
2887 As an example of when such bitcasting is invalid, loading 32-bit integer or
2888 floating-point objects into floating-point registers on Alpha extends them
2889 to 64 bits. Therefore loading a 64-bit object and then storing it as a
2890 32-bit object does not store the low-order 32 bits, as would be the case
2891 for a normal register. Therefore, @file{alpha.h} defines
2892 @code{TARGET_CAN_CHANGE_MODE_CLASS} to return:
2895 (GET_MODE_SIZE (from) == GET_MODE_SIZE (to)
2896 || !reg_classes_intersect_p (FLOAT_REGS, rclass))
2899 Even if storing from a register in mode @var{to} would be valid,
2900 if both @var{from} and @code{raw_reg_mode} for @var{rclass} are wider
2901 than @code{word_mode}, then we must prevent @var{to} narrowing the
2902 mode. This happens when the middle-end assumes that it can load
2903 or store pieces of an @var{N}-word pseudo, and that the pseudo will
2904 eventually be allocated to @var{N} @code{word_mode} hard registers.
2905 Failure to prevent this kind of mode change will result in the
2906 entire @code{raw_reg_mode} being modified instead of the partial
2907 value that the middle-end intended.
2910 @deftypefn {Target Hook} reg_class_t TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS (int, @var{reg_class_t}, @var{reg_class_t})
2911 A target hook which can change allocno class for given pseudo from
2912 allocno and best class calculated by IRA.
2914 The default version of this target hook always returns given class.
2917 @deftypefn {Target Hook} bool TARGET_LRA_P (void)
2918 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.
2921 @deftypefn {Target Hook} int TARGET_REGISTER_PRIORITY (int)
2922 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.
2925 @deftypefn {Target Hook} bool TARGET_REGISTER_USAGE_LEVELING_P (void)
2926 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.
2929 @deftypefn {Target Hook} bool TARGET_DIFFERENT_ADDR_DISPLACEMENT_P (void)
2930 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.
2933 @deftypefn {Target Hook} bool TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P (rtx @var{subst})
2934 A target hook which returns @code{true} if @var{subst} can't
2935 substitute safely pseudos with equivalent memory values during
2936 register allocation.
2937 The default version of this target hook returns @code{false}.
2938 On most machines, this default should be used. For generally
2939 machines with non orthogonal register usage for addressing, such
2940 as SH, this hook can be used to avoid excessive spilling.
2943 @deftypefn {Target Hook} bool TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT (rtx *@var{offset1}, rtx *@var{offset2}, poly_int64 @var{orig_offset}, machine_mode @var{mode})
2944 This hook tries to split address offset @var{orig_offset} into
2945 two parts: one that should be added to the base address to create
2946 a local anchor point, and an additional offset that can be applied
2947 to the anchor to address a value of mode @var{mode}. The idea is that
2948 the local anchor could be shared by other accesses to nearby locations.
2950 The hook returns true if it succeeds, storing the offset of the
2951 anchor from the base in @var{offset1} and the offset of the final address
2952 from the anchor in @var{offset2}. The default implementation returns false.
2955 @deftypefn {Target Hook} reg_class_t TARGET_SPILL_CLASS (reg_class_t, @var{machine_mode})
2956 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.
2959 @deftypefn {Target Hook} bool TARGET_ADDITIONAL_ALLOCNO_CLASS_P (reg_class_t)
2960 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.
2963 @deftypefn {Target Hook} scalar_int_mode TARGET_CSTORE_MODE (enum insn_code @var{icode})
2964 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.
2967 @deftypefn {Target Hook} int TARGET_COMPUTE_PRESSURE_CLASSES (enum reg_class *@var{pressure_classes})
2968 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}.
2971 @node Stack and Calling
2972 @section Stack Layout and Calling Conventions
2973 @cindex calling conventions
2975 @c prevent bad page break with this line
2976 This describes the stack layout and calling conventions.
2980 * Exception Handling::
2985 * Register Arguments::
2987 * Aggregate Return::
2992 * Shrink-wrapping separate components::
2993 * Stack Smashing Protection::
2994 * Miscellaneous Register Hooks::
2998 @subsection Basic Stack Layout
2999 @cindex stack frame layout
3000 @cindex frame layout
3002 @c prevent bad page break with this line
3003 Here is the basic stack layout.
3005 @defmac STACK_GROWS_DOWNWARD
3006 Define this macro to be true if pushing a word onto the stack moves the stack
3007 pointer to a smaller address, and false otherwise.
3010 @defmac STACK_PUSH_CODE
3011 This macro defines the operation used when something is pushed
3012 on the stack. In RTL, a push operation will be
3013 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3015 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3016 and @code{POST_INC}. Which of these is correct depends on
3017 the stack direction and on whether the stack pointer points
3018 to the last item on the stack or whether it points to the
3019 space for the next item on the stack.
3021 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3022 true, which is almost always right, and @code{PRE_INC} otherwise,
3023 which is often wrong.
3026 @defmac FRAME_GROWS_DOWNWARD
3027 Define this macro to nonzero value if the addresses of local variable slots
3028 are at negative offsets from the frame pointer.
3031 @defmac ARGS_GROW_DOWNWARD
3032 Define this macro if successive arguments to a function occupy decreasing
3033 addresses on the stack.
3036 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_STARTING_FRAME_OFFSET (void)
3037 This hook returns the offset from the frame pointer to the first local
3038 variable slot to be allocated. If @code{FRAME_GROWS_DOWNWARD}, it is the
3039 offset to @emph{end} of the first slot allocated, otherwise it is the
3040 offset to @emph{beginning} of the first slot allocated. The default
3041 implementation returns 0.
3044 @defmac STACK_ALIGNMENT_NEEDED
3045 Define to zero to disable final alignment of the stack during reload.
3046 The nonzero default for this macro is suitable for most ports.
3048 On ports where @code{TARGET_STARTING_FRAME_OFFSET} is nonzero or where there
3049 is a register save block following the local block that doesn't require
3050 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3051 stack alignment and do it in the backend.
3054 @defmac STACK_POINTER_OFFSET
3055 Offset from the stack pointer register to the first location at which
3056 outgoing arguments are placed. If not specified, the default value of
3057 zero is used. This is the proper value for most machines.
3059 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3060 the first location at which outgoing arguments are placed.
3063 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3064 Offset from the argument pointer register to the first argument's
3065 address. On some machines it may depend on the data type of the
3068 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3069 the first argument's address.
3072 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3073 Offset from the stack pointer register to an item dynamically allocated
3074 on the stack, e.g., by @code{alloca}.
3076 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3077 length of the outgoing arguments. The default is correct for most
3078 machines. See @file{function.c} for details.
3081 @defmac INITIAL_FRAME_ADDRESS_RTX
3082 A C expression whose value is RTL representing the address of the initial
3083 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3084 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3085 default value will be used. Define this macro in order to make frame pointer
3086 elimination work in the presence of @code{__builtin_frame_address (count)} and
3087 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3090 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3091 A C expression whose value is RTL representing the address in a stack
3092 frame where the pointer to the caller's frame is stored. Assume that
3093 @var{frameaddr} is an RTL expression for the address of the stack frame
3096 If you don't define this macro, the default is to return the value
3097 of @var{frameaddr}---that is, the stack frame address is also the
3098 address of the stack word that points to the previous frame.
3101 @defmac SETUP_FRAME_ADDRESSES
3102 A C expression that produces the machine-specific code to
3103 setup the stack so that arbitrary frames can be accessed. For example,
3104 on the SPARC, we must flush all of the register windows to the stack
3105 before we can access arbitrary stack frames. You will seldom need to
3106 define this macro. The default is to do nothing.
3109 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3110 This target hook should return an rtx that is used to store
3111 the address of the current frame into the built in @code{setjmp} buffer.
3112 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3113 machines. One reason you may need to define this target hook is if
3114 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3117 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3118 A C expression whose value is RTL representing the value of the frame
3119 address for the current frame. @var{frameaddr} is the frame pointer
3120 of the current frame. This is used for __builtin_frame_address.
3121 You need only define this macro if the frame address is not the same
3122 as the frame pointer. Most machines do not need to define it.
3125 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3126 A C expression whose value is RTL representing the value of the return
3127 address for the frame @var{count} steps up from the current frame, after
3128 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3129 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3130 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero.
3132 The value of the expression must always be the correct address when
3133 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3134 determine the return address of other frames.
3137 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3138 Define this macro to nonzero value if the return address of a particular
3139 stack frame is accessed from the frame pointer of the previous stack
3140 frame. The zero default for this macro is suitable for most ports.
3143 @defmac INCOMING_RETURN_ADDR_RTX
3144 A C expression whose value is RTL representing the location of the
3145 incoming return address at the beginning of any function, before the
3146 prologue. This RTL is either a @code{REG}, indicating that the return
3147 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3150 You only need to define this macro if you want to support call frame
3151 debugging information like that provided by DWARF 2.
3153 If this RTL is a @code{REG}, you should also define
3154 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3157 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3158 A C expression whose value is an integer giving a DWARF 2 column
3159 number that may be used as an alternative return column. The column
3160 must not correspond to any gcc hard register (that is, it must not
3161 be in the range of @code{DWARF_FRAME_REGNUM}).
3163 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3164 general register, but an alternative column needs to be used for signal
3165 frames. Some targets have also used different frame return columns
3169 @defmac DWARF_ZERO_REG
3170 A C expression whose value is an integer giving a DWARF 2 register
3171 number that is considered to always have the value zero. This should
3172 only be defined if the target has an architected zero register, and
3173 someone decided it was a good idea to use that register number to
3174 terminate the stack backtrace. New ports should avoid this.
3177 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3178 This target hook allows the backend to emit frame-related insns that
3179 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3180 info engine will invoke it on insns of the form
3182 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3186 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3188 to let the backend emit the call frame instructions. @var{label} is
3189 the CFI label attached to the insn, @var{pattern} is the pattern of
3190 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3193 @deftypefn {Target Hook} {unsigned int} TARGET_DWARF_POLY_INDETERMINATE_VALUE (unsigned int @var{i}, unsigned int *@var{factor}, int *@var{offset})
3194 Express the value of @code{poly_int} indeterminate @var{i} as a DWARF
3195 expression, with @var{i} counting from 1. Return the number of a DWARF
3196 register @var{R} and set @samp{*@var{factor}} and @samp{*@var{offset}} such
3197 that the value of the indeterminate is:
3199 value_of(@var{R}) / @var{factor} - @var{offset}
3202 A target only needs to define this hook if it sets
3203 @samp{NUM_POLY_INT_COEFFS} to a value greater than 1.
3206 @defmac INCOMING_FRAME_SP_OFFSET
3207 A C expression whose value is an integer giving the offset, in bytes,
3208 from the value of the stack pointer register to the top of the stack
3209 frame at the beginning of any function, before the prologue. The top of
3210 the frame is defined to be the value of the stack pointer in the
3211 previous frame, just before the call instruction.
3213 You only need to define this macro if you want to support call frame
3214 debugging information like that provided by DWARF 2.
3217 @defmac DEFAULT_INCOMING_FRAME_SP_OFFSET
3218 Like @code{INCOMING_FRAME_SP_OFFSET}, but must be the same for all
3219 functions of the same ABI, and when using GAS @code{.cfi_*} directives
3220 must also agree with the default CFI GAS emits. Define this macro
3221 only if @code{INCOMING_FRAME_SP_OFFSET} can have different values
3222 between different functions of the same ABI or when
3223 @code{INCOMING_FRAME_SP_OFFSET} does not agree with GAS default CFI.
3226 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3227 A C expression whose value is an integer giving the offset, in bytes,
3228 from the argument pointer to the canonical frame address (cfa). The
3229 final value should coincide with that calculated by
3230 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3231 during virtual register instantiation.
3233 The default value for this macro is
3234 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3235 which is correct for most machines; in general, the arguments are found
3236 immediately before the stack frame. Note that this is not the case on
3237 some targets that save registers into the caller's frame, such as SPARC
3238 and rs6000, and so such targets need to define this macro.
3240 You only need to define this macro if the default is incorrect, and you
3241 want to support call frame debugging information like that provided by
3245 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3246 If defined, a C expression whose value is an integer giving the offset
3247 in bytes from the frame pointer to the canonical frame address (cfa).
3248 The final value should coincide with that calculated by
3249 @code{INCOMING_FRAME_SP_OFFSET}.
3251 Normally the CFA is calculated as an offset from the argument pointer,
3252 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3253 variable due to the ABI, this may not be possible. If this macro is
3254 defined, it implies that the virtual register instantiation should be
3255 based on the frame pointer instead of the argument pointer. Only one
3256 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3260 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3261 If defined, a C expression whose value is an integer giving the offset
3262 in bytes from the canonical frame address (cfa) to the frame base used
3263 in DWARF 2 debug information. The default is zero. A different value
3264 may reduce the size of debug information on some ports.
3267 @node Exception Handling
3268 @subsection Exception Handling Support
3269 @cindex exception handling
3271 @defmac EH_RETURN_DATA_REGNO (@var{N})
3272 A C expression whose value is the @var{N}th register number used for
3273 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3274 @var{N} registers are usable.
3276 The exception handling library routines communicate with the exception
3277 handlers via a set of agreed upon registers. Ideally these registers
3278 should be call-clobbered; it is possible to use call-saved registers,
3279 but may negatively impact code size. The target must support at least
3280 2 data registers, but should define 4 if there are enough free registers.
3282 You must define this macro if you want to support call frame exception
3283 handling like that provided by DWARF 2.
3286 @defmac EH_RETURN_STACKADJ_RTX
3287 A C expression whose value is RTL representing a location in which
3288 to store a stack adjustment to be applied before function return.
3289 This is used to unwind the stack to an exception handler's call frame.
3290 It will be assigned zero on code paths that return normally.
3292 Typically this is a call-clobbered hard register that is otherwise
3293 untouched by the epilogue, but could also be a stack slot.
3295 Do not define this macro if the stack pointer is saved and restored
3296 by the regular prolog and epilog code in the call frame itself; in
3297 this case, the exception handling library routines will update the
3298 stack location to be restored in place. Otherwise, you must define
3299 this macro if you want to support call frame exception handling like
3300 that provided by DWARF 2.
3303 @defmac EH_RETURN_HANDLER_RTX
3304 A C expression whose value is RTL representing a location in which
3305 to store the address of an exception handler to which we should
3306 return. It will not be assigned on code paths that return normally.
3308 Typically this is the location in the call frame at which the normal
3309 return address is stored. For targets that return by popping an
3310 address off the stack, this might be a memory address just below
3311 the @emph{target} call frame rather than inside the current call
3312 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3313 been assigned, so it may be used to calculate the location of the
3316 Some targets have more complex requirements than storing to an
3317 address calculable during initial code generation. In that case
3318 the @code{eh_return} instruction pattern should be used instead.
3320 If you want to support call frame exception handling, you must
3321 define either this macro or the @code{eh_return} instruction pattern.
3324 @defmac RETURN_ADDR_OFFSET
3325 If defined, an integer-valued C expression for which rtl will be generated
3326 to add it to the exception handler address before it is searched in the
3327 exception handling tables, and to subtract it again from the address before
3328 using it to return to the exception handler.
3331 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3332 This macro chooses the encoding of pointers embedded in the exception
3333 handling sections. If at all possible, this should be defined such
3334 that the exception handling section will not require dynamic relocations,
3335 and so may be read-only.
3337 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3338 @var{global} is true if the symbol may be affected by dynamic relocations.
3339 The macro should return a combination of the @code{DW_EH_PE_*} defines
3340 as found in @file{dwarf2.h}.
3342 If this macro is not defined, pointers will not be encoded but
3343 represented directly.
3346 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3347 This macro allows the target to emit whatever special magic is required
3348 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3349 Generic code takes care of pc-relative and indirect encodings; this must
3350 be defined if the target uses text-relative or data-relative encodings.
3352 This is a C statement that branches to @var{done} if the format was
3353 handled. @var{encoding} is the format chosen, @var{size} is the number
3354 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3358 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3359 This macro allows the target to add CPU and operating system specific
3360 code to the call-frame unwinder for use when there is no unwind data
3361 available. The most common reason to implement this macro is to unwind
3362 through signal frames.
3364 This macro is called from @code{uw_frame_state_for} in
3365 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3366 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3367 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3368 for the address of the code being executed and @code{context->cfa} for
3369 the stack pointer value. If the frame can be decoded, the register
3370 save addresses should be updated in @var{fs} and the macro should
3371 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3372 the macro should evaluate to @code{_URC_END_OF_STACK}.
3374 For proper signal handling in Java this macro is accompanied by
3375 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3378 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3379 This macro allows the target to add operating system specific code to the
3380 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3381 usually used for signal or interrupt frames.
3383 This macro is called from @code{uw_update_context} in libgcc's
3384 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3385 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3386 for the abi and context in the @code{.unwabi} directive. If the
3387 @code{.unwabi} directive can be handled, the register save addresses should
3388 be updated in @var{fs}.
3391 @defmac TARGET_USES_WEAK_UNWIND_INFO
3392 A C expression that evaluates to true if the target requires unwind
3393 info to be given comdat linkage. Define it to be @code{1} if comdat
3394 linkage is necessary. The default is @code{0}.
3397 @node Stack Checking
3398 @subsection Specifying How Stack Checking is Done
3400 GCC will check that stack references are within the boundaries of the
3401 stack, if the option @option{-fstack-check} is specified, in one of
3406 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3407 will assume that you have arranged for full stack checking to be done
3408 at appropriate places in the configuration files. GCC will not do
3409 other special processing.
3412 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3413 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3414 that you have arranged for static stack checking (checking of the
3415 static stack frame of functions) to be done at appropriate places
3416 in the configuration files. GCC will only emit code to do dynamic
3417 stack checking (checking on dynamic stack allocations) using the third
3421 If neither of the above are true, GCC will generate code to periodically
3422 ``probe'' the stack pointer using the values of the macros defined below.
3425 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3426 GCC will change its allocation strategy for large objects if the option
3427 @option{-fstack-check} is specified: they will always be allocated
3428 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3430 @defmac STACK_CHECK_BUILTIN
3431 A nonzero value if stack checking is done by the configuration files in a
3432 machine-dependent manner. You should define this macro if stack checking
3433 is required by the ABI of your machine or if you would like to do stack
3434 checking in some more efficient way than the generic approach. The default
3435 value of this macro is zero.
3438 @defmac STACK_CHECK_STATIC_BUILTIN
3439 A nonzero value if static stack checking is done by the configuration files
3440 in a machine-dependent manner. You should define this macro if you would
3441 like to do static stack checking in some more efficient way than the generic
3442 approach. The default value of this macro is zero.
3445 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3446 An integer specifying the interval at which GCC must generate stack probe
3447 instructions, defined as 2 raised to this integer. You will normally
3448 define this macro so that the interval be no larger than the size of
3449 the ``guard pages'' at the end of a stack area. The default value
3450 of 12 (4096-byte interval) is suitable for most systems.
3453 @defmac STACK_CHECK_MOVING_SP
3454 An integer which is nonzero if GCC should move the stack pointer page by page
3455 when doing probes. This can be necessary on systems where the stack pointer
3456 contains the bottom address of the memory area accessible to the executing
3457 thread at any point in time. In this situation an alternate signal stack
3458 is required in order to be able to recover from a stack overflow. The
3459 default value of this macro is zero.
3462 @defmac STACK_CHECK_PROTECT
3463 The number of bytes of stack needed to recover from a stack overflow, for
3464 languages where such a recovery is supported. The default value of 4KB/8KB
3465 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3466 8KB/12KB with other exception handling mechanisms should be adequate for most
3467 architectures and operating systems.
3470 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3471 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3472 in the opposite case.
3474 @defmac STACK_CHECK_MAX_FRAME_SIZE
3475 The maximum size of a stack frame, in bytes. GCC will generate probe
3476 instructions in non-leaf functions to ensure at least this many bytes of
3477 stack are available. If a stack frame is larger than this size, stack
3478 checking will not be reliable and GCC will issue a warning. The
3479 default is chosen so that GCC only generates one instruction on most
3480 systems. You should normally not change the default value of this macro.
3483 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3484 GCC uses this value to generate the above warning message. It
3485 represents the amount of fixed frame used by a function, not including
3486 space for any callee-saved registers, temporaries and user variables.
3487 You need only specify an upper bound for this amount and will normally
3488 use the default of four words.
3491 @defmac STACK_CHECK_MAX_VAR_SIZE
3492 The maximum size, in bytes, of an object that GCC will place in the
3493 fixed area of the stack frame when the user specifies
3494 @option{-fstack-check}.
3495 GCC computed the default from the values of the above macros and you will
3496 normally not need to override that default.
3499 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE (void)
3500 Some targets have an ABI defined interval for which no probing needs to be done.
3501 When a probe does need to be done this same interval is used as the probe distance up when doing stack clash protection for alloca.
3502 On such targets this value can be set to override the default probing up interval.
3503 Define this variable to return nonzero if such a probe range is required or zero otherwise. Defining this hook also requires your functions which make use of alloca to have at least 8 byesof outgoing arguments. If this is not the case the stack will be corrupted.
3504 You need not define this macro if it would always have the value zero.
3508 @node Frame Registers
3509 @subsection Registers That Address the Stack Frame
3511 @c prevent bad page break with this line
3512 This discusses registers that address the stack frame.
3514 @defmac STACK_POINTER_REGNUM
3515 The register number of the stack pointer register, which must also be a
3516 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3517 the hardware determines which register this is.
3520 @defmac FRAME_POINTER_REGNUM
3521 The register number of the frame pointer register, which is used to
3522 access automatic variables in the stack frame. On some machines, the
3523 hardware determines which register this is. On other machines, you can
3524 choose any register you wish for this purpose.
3527 @defmac HARD_FRAME_POINTER_REGNUM
3528 On some machines the offset between the frame pointer and starting
3529 offset of the automatic variables is not known until after register
3530 allocation has been done (for example, because the saved registers are
3531 between these two locations). On those machines, define
3532 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3533 be used internally until the offset is known, and define
3534 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3535 used for the frame pointer.
3537 You should define this macro only in the very rare circumstances when it
3538 is not possible to calculate the offset between the frame pointer and
3539 the automatic variables until after register allocation has been
3540 completed. When this macro is defined, you must also indicate in your
3541 definition of @code{ELIMINABLE_REGS} how to eliminate
3542 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3543 or @code{STACK_POINTER_REGNUM}.
3545 Do not define this macro if it would be the same as
3546 @code{FRAME_POINTER_REGNUM}.
3549 @defmac ARG_POINTER_REGNUM
3550 The register number of the arg pointer register, which is used to access
3551 the function's argument list. On some machines, this is the same as the
3552 frame pointer register. On some machines, the hardware determines which
3553 register this is. On other machines, you can choose any register you
3554 wish for this purpose. If this is not the same register as the frame
3555 pointer register, then you must mark it as a fixed register according to
3556 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3557 (@pxref{Elimination}).
3560 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3561 Define this to a preprocessor constant that is nonzero if
3562 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3563 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3564 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3565 definition is not suitable for use in preprocessor conditionals.
3568 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3569 Define this to a preprocessor constant that is nonzero if
3570 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3571 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3572 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3573 definition is not suitable for use in preprocessor conditionals.
3576 @defmac RETURN_ADDRESS_POINTER_REGNUM
3577 The register number of the return address pointer register, which is used to
3578 access the current function's return address from the stack. On some
3579 machines, the return address is not at a fixed offset from the frame
3580 pointer or stack pointer or argument pointer. This register can be defined
3581 to point to the return address on the stack, and then be converted by
3582 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3584 Do not define this macro unless there is no other way to get the return
3585 address from the stack.
3588 @defmac STATIC_CHAIN_REGNUM
3589 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3590 Register numbers used for passing a function's static chain pointer. If
3591 register windows are used, the register number as seen by the called
3592 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3593 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3594 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3597 The static chain register need not be a fixed register.
3599 If the static chain is passed in memory, these macros should not be
3600 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3603 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl_or_type}, bool @var{incoming_p})
3604 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3605 targets that may use different static chain locations for different
3606 nested functions. This may be required if the target has function
3607 attributes that affect the calling conventions of the function and
3608 those calling conventions use different static chain locations.
3610 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3612 If the static chain is passed in memory, this hook should be used to
3613 provide rtx giving @code{mem} expressions that denote where they are stored.
3614 Often the @code{mem} expression as seen by the caller will be at an offset
3615 from the stack pointer and the @code{mem} expression as seen by the callee
3616 will be at an offset from the frame pointer.
3617 @findex stack_pointer_rtx
3618 @findex frame_pointer_rtx
3619 @findex arg_pointer_rtx
3620 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3621 @code{arg_pointer_rtx} will have been initialized and should be used
3622 to refer to those items.
3625 @defmac DWARF_FRAME_REGISTERS
3626 This macro specifies the maximum number of hard registers that can be
3627 saved in a call frame. This is used to size data structures used in
3628 DWARF2 exception handling.
3630 Prior to GCC 3.0, this macro was needed in order to establish a stable
3631 exception handling ABI in the face of adding new hard registers for ISA
3632 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3633 in the number of hard registers. Nevertheless, this macro can still be
3634 used to reduce the runtime memory requirements of the exception handling
3635 routines, which can be substantial if the ISA contains a lot of
3636 registers that are not call-saved.
3638 If this macro is not defined, it defaults to
3639 @code{FIRST_PSEUDO_REGISTER}.
3642 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3644 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3645 for backward compatibility in pre GCC 3.0 compiled code.
3647 If this macro is not defined, it defaults to
3648 @code{DWARF_FRAME_REGISTERS}.
3651 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3653 Define this macro if the target's representation for dwarf registers
3654 is different than the internal representation for unwind column.
3655 Given a dwarf register, this macro should return the internal unwind
3656 column number to use instead.
3659 @defmac DWARF_FRAME_REGNUM (@var{regno})
3661 Define this macro if the target's representation for dwarf registers
3662 used in .eh_frame or .debug_frame is different from that used in other
3663 debug info sections. Given a GCC hard register number, this macro
3664 should return the .eh_frame register number. The default is
3665 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3669 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3671 Define this macro to map register numbers held in the call frame info
3672 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3673 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3674 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3675 return @code{@var{regno}}.
3679 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3681 Define this macro if the target stores register values as
3682 @code{_Unwind_Word} type in unwind context. It should be defined if
3683 target register size is larger than the size of @code{void *}. The
3684 default is to store register values as @code{void *} type.
3688 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3690 Define this macro to be 1 if the target always uses extended unwind
3691 context with version, args_size and by_value fields. If it is undefined,
3692 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3693 defined and 0 otherwise.
3697 @defmac DWARF_LAZY_REGISTER_VALUE (@var{regno}, @var{value})
3698 Define this macro if the target has pseudo DWARF registers whose
3699 values need to be computed lazily on demand by the unwinder (such as when
3700 referenced in a CFA expression). The macro returns true if @var{regno}
3701 is such a register and stores its value in @samp{*@var{value}} if so.
3705 @subsection Eliminating Frame Pointer and Arg Pointer
3707 @c prevent bad page break with this line
3708 This is about eliminating the frame pointer and arg pointer.
3710 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3711 This target hook should return @code{true} if a function must have and use
3712 a frame pointer. This target hook is called in the reload pass. If its return
3713 value is @code{true} the function will have a frame pointer.
3715 This target hook can in principle examine the current function and decide
3716 according to the facts, but on most machines the constant @code{false} or the
3717 constant @code{true} suffices. Use @code{false} when the machine allows code
3718 to be generated with no frame pointer, and doing so saves some time or space.
3719 Use @code{true} when there is no possible advantage to avoiding a frame
3722 In certain cases, the compiler does not know how to produce valid code
3723 without a frame pointer. The compiler recognizes those cases and
3724 automatically gives the function a frame pointer regardless of what
3725 @code{targetm.frame_pointer_required} returns. You don't need to worry about
3728 In a function that does not require a frame pointer, the frame pointer
3729 register can be allocated for ordinary usage, unless you mark it as a
3730 fixed register. See @code{FIXED_REGISTERS} for more information.
3732 Default return value is @code{false}.
3735 @defmac ELIMINABLE_REGS
3736 This macro specifies a table of register pairs used to eliminate
3737 unneeded registers that point into the stack frame.
3739 The definition of this macro is a list of structure initializations, each
3740 of which specifies an original and replacement register.
3742 On some machines, the position of the argument pointer is not known until
3743 the compilation is completed. In such a case, a separate hard register
3744 must be used for the argument pointer. This register can be eliminated by
3745 replacing it with either the frame pointer or the argument pointer,
3746 depending on whether or not the frame pointer has been eliminated.
3748 In this case, you might specify:
3750 #define ELIMINABLE_REGS \
3751 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3752 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3753 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3756 Note that the elimination of the argument pointer with the stack pointer is
3757 specified first since that is the preferred elimination.
3760 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3761 This target hook should return @code{true} if the compiler is allowed to
3762 try to replace register number @var{from_reg} with register number
3763 @var{to_reg}. This target hook will usually be @code{true}, since most of the
3764 cases preventing register elimination are things that the compiler already
3767 Default return value is @code{true}.
3770 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3771 This macro returns the initial difference between the specified pair
3772 of registers. The value would be computed from information
3773 such as the result of @code{get_frame_size ()} and the tables of
3774 registers @code{df_regs_ever_live_p} and @code{call_used_regs}.
3777 @deftypefn {Target Hook} void TARGET_COMPUTE_FRAME_LAYOUT (void)
3778 This target hook is called once each time the frame layout needs to be
3779 recalculated. The calculations can be cached by the target and can then
3780 be used by @code{INITIAL_ELIMINATION_OFFSET} instead of re-computing the
3781 layout on every invocation of that hook. This is particularly useful
3782 for targets that have an expensive frame layout function. Implementing
3783 this callback is optional.
3786 @node Stack Arguments
3787 @subsection Passing Function Arguments on the Stack
3788 @cindex arguments on stack
3789 @cindex stack arguments
3791 The macros in this section control how arguments are passed
3792 on the stack. See the following section for other macros that
3793 control passing certain arguments in registers.
3795 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3796 This target hook returns @code{true} if an argument declared in a
3797 prototype as an integral type smaller than @code{int} should actually be
3798 passed as an @code{int}. In addition to avoiding errors in certain
3799 cases of mismatch, it also makes for better code on certain machines.
3800 The default is to not promote prototypes.
3804 A C expression. If nonzero, push insns will be used to pass
3806 If the target machine does not have a push instruction, set it to zero.
3807 That directs GCC to use an alternate strategy: to
3808 allocate the entire argument block and then store the arguments into
3809 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3812 @defmac PUSH_ARGS_REVERSED
3813 A C expression. If nonzero, function arguments will be evaluated from
3814 last to first, rather than from first to last. If this macro is not
3815 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3816 and args grow in opposite directions, and 0 otherwise.
3819 @defmac PUSH_ROUNDING (@var{npushed})
3820 A C expression that is the number of bytes actually pushed onto the
3821 stack when an instruction attempts to push @var{npushed} bytes.
3823 On some machines, the definition
3826 #define PUSH_ROUNDING(BYTES) (BYTES)
3830 will suffice. But on other machines, instructions that appear
3831 to push one byte actually push two bytes in an attempt to maintain
3832 alignment. Then the definition should be
3835 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3838 If the value of this macro has a type, it should be an unsigned type.
3841 @findex outgoing_args_size
3842 @findex crtl->outgoing_args_size
3843 @defmac ACCUMULATE_OUTGOING_ARGS
3844 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3845 will be computed and placed into
3846 @code{crtl->outgoing_args_size}. No space will be pushed
3847 onto the stack for each call; instead, the function prologue should
3848 increase the stack frame size by this amount.
3850 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3854 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3855 Define this macro if functions should assume that stack space has been
3856 allocated for arguments even when their values are passed in
3859 The value of this macro is the size, in bytes, of the area reserved for
3860 arguments passed in registers for the function represented by @var{fndecl},
3861 which can be zero if GCC is calling a library function.
3862 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3865 This space can be allocated by the caller, or be a part of the
3866 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3869 @c above is overfull. not sure what to do. --mew 5feb93 did
3870 @c something, not sure if it looks good. --mew 10feb93
3872 @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3873 Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3874 Define this macro if space guaranteed when compiling a function body
3875 is different to space required when making a call, a situation that
3876 can arise with K&R style function definitions.
3879 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3880 Define this to a nonzero value if it is the responsibility of the
3881 caller to allocate the area reserved for arguments passed in registers
3882 when calling a function of @var{fntype}. @var{fntype} may be NULL
3883 if the function called is a library function.
3885 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3886 whether the space for these arguments counts in the value of
3887 @code{crtl->outgoing_args_size}.
3890 @defmac STACK_PARMS_IN_REG_PARM_AREA
3891 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3892 stack parameters don't skip the area specified by it.
3893 @c i changed this, makes more sens and it should have taken care of the
3894 @c overfull.. not as specific, tho. --mew 5feb93
3896 Normally, when a parameter is not passed in registers, it is placed on the
3897 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3898 suppresses this behavior and causes the parameter to be passed on the
3899 stack in its natural location.
3902 @deftypefn {Target Hook} poly_int64 TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, poly_int64 @var{size})
3903 This target hook returns the number of bytes of its own arguments that
3904 a function pops on returning, or 0 if the function pops no arguments
3905 and the caller must therefore pop them all after the function returns.
3907 @var{fundecl} is a C variable whose value is a tree node that describes
3908 the function in question. Normally it is a node of type
3909 @code{FUNCTION_DECL} that describes the declaration of the function.
3910 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3912 @var{funtype} is a C variable whose value is a tree node that
3913 describes the function in question. Normally it is a node of type
3914 @code{FUNCTION_TYPE} that describes the data type of the function.
3915 From this it is possible to obtain the data types of the value and
3916 arguments (if known).
3918 When a call to a library function is being considered, @var{fundecl}
3919 will contain an identifier node for the library function. Thus, if
3920 you need to distinguish among various library functions, you can do so
3921 by their names. Note that ``library function'' in this context means
3922 a function used to perform arithmetic, whose name is known specially
3923 in the compiler and was not mentioned in the C code being compiled.
3925 @var{size} is the number of bytes of arguments passed on the
3926 stack. If a variable number of bytes is passed, it is zero, and
3927 argument popping will always be the responsibility of the calling function.
3929 On the VAX, all functions always pop their arguments, so the definition
3930 of this macro is @var{size}. On the 68000, using the standard
3931 calling convention, no functions pop their arguments, so the value of
3932 the macro is always 0 in this case. But an alternative calling
3933 convention is available in which functions that take a fixed number of
3934 arguments pop them but other functions (such as @code{printf}) pop
3935 nothing (the caller pops all). When this convention is in use,
3936 @var{funtype} is examined to determine whether a function takes a fixed
3937 number of arguments.
3940 @defmac CALL_POPS_ARGS (@var{cum})
3941 A C expression that should indicate the number of bytes a call sequence
3942 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3943 when compiling a function call.
3945 @var{cum} is the variable in which all arguments to the called function
3946 have been accumulated.
3948 On certain architectures, such as the SH5, a call trampoline is used
3949 that pops certain registers off the stack, depending on the arguments
3950 that have been passed to the function. Since this is a property of the
3951 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3955 @node Register Arguments
3956 @subsection Passing Arguments in Registers
3957 @cindex arguments in registers
3958 @cindex registers arguments
3960 This section describes the macros which let you control how various
3961 types of arguments are passed in registers or how they are arranged in
3964 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, const function_arg_info @var{&arg})
3965 Return an RTX indicating whether function argument @var{arg} is passed
3966 in a register and if so, which register. Argument @var{ca} summarizes all
3967 the previous arguments.
3969 The return value is usually either a @code{reg} RTX for the hard
3970 register in which to pass the argument, or zero to pass the argument
3973 The return value can be a @code{const_int} which means argument is
3974 passed in a target specific slot with specified number. Target hooks
3975 should be used to store or load argument in such case. See
3976 @code{TARGET_STORE_BOUNDS_FOR_ARG} and @code{TARGET_LOAD_BOUNDS_FOR_ARG}
3977 for more information.
3979 The value of the expression can also be a @code{parallel} RTX@. This is
3980 used when an argument is passed in multiple locations. The mode of the
3981 @code{parallel} should be the mode of the entire argument. The
3982 @code{parallel} holds any number of @code{expr_list} pairs; each one
3983 describes where part of the argument is passed. In each
3984 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3985 register in which to pass this part of the argument, and the mode of the
3986 register RTX indicates how large this part of the argument is. The
3987 second operand of the @code{expr_list} is a @code{const_int} which gives
3988 the offset in bytes into the entire argument of where this part starts.
3989 As a special exception the first @code{expr_list} in the @code{parallel}
3990 RTX may have a first operand of zero. This indicates that the entire
3991 argument is also stored on the stack.
3993 The last time this hook is called, it is called with @code{MODE ==
3994 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3995 pattern as operands 2 and 3 respectively.
3997 @cindex @file{stdarg.h} and register arguments
3998 The usual way to make the ISO library @file{stdarg.h} work on a
3999 machine where some arguments are usually passed in registers, is to
4000 cause nameless arguments to be passed on the stack instead. This is
4001 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
4002 @var{named} is @code{false}.
4004 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
4005 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
4006 You may use the hook @code{targetm.calls.must_pass_in_stack}
4007 in the definition of this macro to determine if this argument is of a
4008 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4009 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
4010 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4011 defined, the argument will be computed in the stack and then loaded into
4015 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (const function_arg_info @var{&arg})
4016 This target hook should return @code{true} if we should not pass @var{arg}
4017 solely in registers. The file @file{expr.h} defines a
4018 definition that is usually appropriate, refer to @file{expr.h} for additional
4022 @deftypefn {Target Hook} rtx TARGET_FUNCTION_INCOMING_ARG (cumulative_args_t @var{ca}, const function_arg_info @var{&arg})
4023 Define this hook if the caller and callee on the target have different
4024 views of where arguments are passed. Also define this hook if there are
4025 functions that are never directly called, but are invoked by the hardware
4026 and which have nonstandard calling conventions.
4028 In this case @code{TARGET_FUNCTION_ARG} computes the register in
4029 which the caller passes the value, and
4030 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4031 fashion to tell the function being called where the arguments will
4034 @code{TARGET_FUNCTION_INCOMING_ARG} can also return arbitrary address
4035 computation using hard register, which can be forced into a register,
4036 so that it can be used to pass special arguments.
4038 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4039 @code{TARGET_FUNCTION_ARG} serves both purposes.
4042 @deftypefn {Target Hook} bool TARGET_USE_PSEUDO_PIC_REG (void)
4043 This hook should return 1 in case pseudo register should be created
4044 for pic_offset_table_rtx during function expand.
4047 @deftypefn {Target Hook} void TARGET_INIT_PIC_REG (void)
4048 Perform a target dependent initialization of pic_offset_table_rtx.
4049 This hook is called at the start of register allocation.
4052 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, const function_arg_info @var{&arg})
4053 This target hook returns the number of bytes at the beginning of an
4054 argument that must be put in registers. The value must be zero for
4055 arguments that are passed entirely in registers or that are entirely
4056 pushed on the stack.
4058 On some machines, certain arguments must be passed partially in
4059 registers and partially in memory. On these machines, typically the
4060 first few words of arguments are passed in registers, and the rest
4061 on the stack. If a multi-word argument (a @code{double} or a
4062 structure) crosses that boundary, its first few words must be passed
4063 in registers and the rest must be pushed. This macro tells the
4064 compiler when this occurs, and how many bytes should go in registers.
4066 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4067 register to be used by the caller for this argument; likewise
4068 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4071 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (cumulative_args_t @var{cum}, const function_arg_info @var{&arg})
4072 This target hook should return @code{true} if argument @var{arg} at the
4073 position indicated by @var{cum} should be passed by reference. This
4074 predicate is queried after target independent reasons for being
4075 passed by reference, such as @code{TREE_ADDRESSABLE (@var{arg}.type)}.
4077 If the hook returns true, a copy of that argument is made in memory and a
4078 pointer to the argument is passed instead of the argument itself.
4079 The pointer is passed in whatever way is appropriate for passing a pointer
4083 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, const function_arg_info @var{&arg})
4084 The function argument described by the parameters to this hook is
4085 known to be passed by reference. The hook should return true if the
4086 function argument should be copied by the callee instead of copied
4089 For any argument for which the hook returns true, if it can be
4090 determined that the argument is not modified, then a copy need
4093 The default version of this hook always returns false.
4096 @defmac CUMULATIVE_ARGS
4097 A C type for declaring a variable that is used as the first argument
4098 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4099 target machines, the type @code{int} suffices and can hold the number
4100 of bytes of argument so far.
4102 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4103 arguments that have been passed on the stack. The compiler has other
4104 variables to keep track of that. For target machines on which all
4105 arguments are passed on the stack, there is no need to store anything in
4106 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4107 should not be empty, so use @code{int}.
4110 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4111 If defined, this macro is called before generating any code for a
4112 function, but after the @var{cfun} descriptor for the function has been
4113 created. The back end may use this macro to update @var{cfun} to
4114 reflect an ABI other than that which would normally be used by default.
4115 If the compiler is generating code for a compiler-generated function,
4116 @var{fndecl} may be @code{NULL}.
4119 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4120 A C statement (sans semicolon) for initializing the variable
4121 @var{cum} for the state at the beginning of the argument list. The
4122 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4123 is the tree node for the data type of the function which will receive
4124 the args, or 0 if the args are to a compiler support library function.
4125 For direct calls that are not libcalls, @var{fndecl} contain the
4126 declaration node of the function. @var{fndecl} is also set when
4127 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4128 being compiled. @var{n_named_args} is set to the number of named
4129 arguments, including a structure return address if it is passed as a
4130 parameter, when making a call. When processing incoming arguments,
4131 @var{n_named_args} is set to @minus{}1.
4133 When processing a call to a compiler support library function,
4134 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4135 contains the name of the function, as a string. @var{libname} is 0 when
4136 an ordinary C function call is being processed. Thus, each time this
4137 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4138 never both of them at once.
4141 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4142 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4143 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4144 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4145 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4146 0)} is used instead.
4149 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4150 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4151 finding the arguments for the function being compiled. If this macro is
4152 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4154 The value passed for @var{libname} is always 0, since library routines
4155 with special calling conventions are never compiled with GCC@. The
4156 argument @var{libname} exists for symmetry with
4157 @code{INIT_CUMULATIVE_ARGS}.
4158 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4159 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4162 @deftypefn {Target Hook} void TARGET_FUNCTION_ARG_ADVANCE (cumulative_args_t @var{ca}, const function_arg_info @var{&arg})
4163 This hook updates the summarizer variable pointed to by @var{ca} to
4164 advance past argument @var{arg} in the argument list. Once this is done,
4165 the variable @var{cum} is suitable for analyzing the @emph{following}
4166 argument with @code{TARGET_FUNCTION_ARG}, etc.
4168 This hook need not do anything if the argument in question was passed
4169 on the stack. The compiler knows how to track the amount of stack space
4170 used for arguments without any special help.
4173 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_FUNCTION_ARG_OFFSET (machine_mode @var{mode}, const_tree @var{type})
4174 This hook returns the number of bytes to add to the offset of an
4175 argument of type @var{type} and mode @var{mode} when passed in memory.
4176 This is needed for the SPU, which passes @code{char} and @code{short}
4177 arguments in the preferred slot that is in the middle of the quad word
4178 instead of starting at the top. The default implementation returns 0.
4181 @deftypefn {Target Hook} pad_direction TARGET_FUNCTION_ARG_PADDING (machine_mode @var{mode}, const_tree @var{type})
4182 This hook determines whether, and in which direction, to pad out
4183 an argument of mode @var{mode} and type @var{type}. It returns
4184 @code{PAD_UPWARD} to insert padding above the argument, @code{PAD_DOWNWARD}
4185 to insert padding below the argument, or @code{PAD_NONE} to inhibit padding.
4187 The @emph{amount} of padding is not controlled by this hook, but by
4188 @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is always just enough
4189 to reach the next multiple of that boundary.
4191 This hook has a default definition that is right for most systems.
4192 For little-endian machines, the default is to pad upward. For
4193 big-endian machines, the default is to pad downward for an argument of
4194 constant size shorter than an @code{int}, and upward otherwise.
4197 @defmac PAD_VARARGS_DOWN
4198 If defined, a C expression which determines whether the default
4199 implementation of va_arg will attempt to pad down before reading the
4200 next argument, if that argument is smaller than its aligned space as
4201 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4202 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4205 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4206 Specify padding for the last element of a block move between registers and
4207 memory. @var{first} is nonzero if this is the only element. Defining this
4208 macro allows better control of register function parameters on big-endian
4209 machines, without using @code{PARALLEL} rtl. In particular,
4210 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4211 registers, as there is no longer a "wrong" part of a register; For example,
4212 a three byte aggregate may be passed in the high part of a register if so
4216 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (machine_mode @var{mode}, const_tree @var{type})
4217 This hook returns the alignment boundary, in bits, of an argument
4218 with the specified mode and type. The default hook returns
4219 @code{PARM_BOUNDARY} for all arguments.
4222 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_ROUND_BOUNDARY (machine_mode @var{mode}, const_tree @var{type})
4223 Normally, the size of an argument is rounded up to @code{PARM_BOUNDARY},
4224 which is the default value for this hook. You can define this hook to
4225 return a different value if an argument size must be rounded to a larger
4229 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4230 A C expression that is nonzero if @var{regno} is the number of a hard
4231 register in which function arguments are sometimes passed. This does
4232 @emph{not} include implicit arguments such as the static chain and
4233 the structure-value address. On many machines, no registers can be
4234 used for this purpose since all function arguments are pushed on the
4238 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4239 This hook should return true if parameter of type @var{type} are passed
4240 as two scalar parameters. By default, GCC will attempt to pack complex
4241 arguments into the target's word size. Some ABIs require complex arguments
4242 to be split and treated as their individual components. For example, on
4243 AIX64, complex floats should be passed in a pair of floating point
4244 registers, even though a complex float would fit in one 64-bit floating
4247 The default value of this hook is @code{NULL}, which is treated as always
4251 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4252 This hook returns a type node for @code{va_list} for the target.
4253 The default version of the hook returns @code{void*}.
4256 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4257 This target hook is used in function @code{c_common_nodes_and_builtins}
4258 to iterate through the target specific builtin types for va_list. The
4259 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4260 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4262 The arguments @var{pname} and @var{ptree} are used to store the result of
4263 this macro and are set to the name of the va_list builtin type and its
4265 If the return value of this macro is zero, then there is no more element.
4266 Otherwise the @var{IDX} should be increased for the next call of this
4267 macro to iterate through all types.
4270 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4271 This hook returns the va_list type of the calling convention specified by
4273 The default version of this hook returns @code{va_list_type_node}.
4276 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4277 This hook returns the va_list type of the calling convention specified by the
4278 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4282 @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})
4283 This hook performs target-specific gimplification of
4284 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4285 arguments to @code{va_arg}; the latter two are as in
4286 @code{gimplify.c:gimplify_expr}.
4289 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (scalar_int_mode @var{mode})
4290 Define this to return nonzero if the port can handle pointers
4291 with machine mode @var{mode}. The default version of this
4292 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4295 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (ao_ref *@var{ref})
4296 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.
4299 @deftypefn {Target Hook} machine_mode TARGET_TRANSLATE_MODE_ATTRIBUTE (machine_mode @var{mode})
4300 Define this hook if during mode attribute processing, the port should
4301 translate machine_mode @var{mode} to another mode. For example, rs6000's
4302 @code{KFmode}, when it is the same as @code{TFmode}.
4304 The default version of the hook returns that mode that was passed in.
4307 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (scalar_mode @var{mode})
4308 Define this to return nonzero if the port is prepared to handle
4309 insns involving scalar mode @var{mode}. For a scalar mode to be
4310 considered supported, all the basic arithmetic and comparisons
4313 The default version of this hook returns true for any mode
4314 required to handle the basic C types (as defined by the port).
4315 Included here are the double-word arithmetic supported by the
4316 code in @file{optabs.c}.
4319 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (machine_mode @var{mode})
4320 Define this to return nonzero if the port is prepared to handle
4321 insns involving vector mode @var{mode}. At the very least, it
4322 must have move patterns for this mode.
4325 @deftypefn {Target Hook} opt_machine_mode TARGET_ARRAY_MODE (machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4326 Return the mode that GCC should use for an array that has
4327 @var{nelems} elements, with each element having mode @var{mode}.
4328 Return no mode if the target has no special requirements. In the
4329 latter case, GCC looks for an integer mode of the appropriate size
4330 if available and uses BLKmode otherwise. Usually the search for the
4331 integer mode is limited to @code{MAX_FIXED_MODE_SIZE}, but the
4332 @code{TARGET_ARRAY_MODE_SUPPORTED_P} hook allows a larger mode to be
4333 used in specific cases.
4335 The main use of this hook is to specify that an array of vectors should
4336 also have a vector mode. The default implementation returns no mode.
4339 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4340 Return true if GCC should try to use a scalar mode to store an array
4341 of @var{nelems} elements, given that each element has mode @var{mode}.
4342 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4343 and allows GCC to use any defined integer mode.
4345 One use of this hook is to support vector load and store operations
4346 that operate on several homogeneous vectors. For example, ARM NEON
4347 has operations like:
4350 int8x8x3_t vld3_s8 (const int8_t *)
4353 where the return type is defined as:
4356 typedef struct int8x8x3_t
4362 If this hook allows @code{val} to have a scalar mode, then
4363 @code{int8x8x3_t} can have the same mode. GCC can then store
4364 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4367 @deftypefn {Target Hook} bool TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P (scalar_float_mode @var{mode})
4368 Define this to return nonzero if libgcc provides support for the
4369 floating-point mode @var{mode}, which is known to pass
4370 @code{TARGET_SCALAR_MODE_SUPPORTED_P}. The default version of this
4371 hook returns true for all of @code{SFmode}, @code{DFmode},
4372 @code{XFmode} and @code{TFmode}, if such modes exist.
4375 @deftypefn {Target Hook} opt_scalar_float_mode TARGET_FLOATN_MODE (int @var{n}, bool @var{extended})
4376 Define this to return the machine mode to use for the type
4377 @code{_Float@var{n}}, if @var{extended} is false, or the type
4378 @code{_Float@var{n}x}, if @var{extended} is true. If such a type is not
4379 supported, return @code{opt_scalar_float_mode ()}. The default version of
4380 this hook returns @code{SFmode} for @code{_Float32}, @code{DFmode} for
4381 @code{_Float64} and @code{_Float32x} and @code{TFmode} for
4382 @code{_Float128}, if those modes exist and satisfy the requirements for
4383 those types and pass @code{TARGET_SCALAR_MODE_SUPPORTED_P} and
4384 @code{TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P}; for @code{_Float64x}, it
4385 returns the first of @code{XFmode} and @code{TFmode} that exists and
4386 satisfies the same requirements; for other types, it returns
4387 @code{opt_scalar_float_mode ()}. The hook is only called for values
4388 of @var{n} and @var{extended} that are valid according to
4389 ISO/IEC TS 18661-3:2015; that is, @var{n} is one of 32, 64, 128, or,
4390 if @var{extended} is false, 16 or greater than 128 and a multiple of 32.
4393 @deftypefn {Target Hook} bool TARGET_FLOATN_BUILTIN_P (int @var{func})
4394 Define this to return true if the @code{_Float@var{n}} and
4395 @code{_Float@var{n}x} built-in functions should implicitly enable the
4396 built-in function without the @code{__builtin_} prefix in addition to the
4397 normal built-in function with the @code{__builtin_} prefix. The default is
4398 to only enable built-in functions without the @code{__builtin_} prefix for
4399 the GNU C langauge. In strict ANSI/ISO mode, the built-in function without
4400 the @code{__builtin_} prefix is not enabled. The argument @code{FUNC} is the
4401 @code{enum built_in_function} id of the function to be enabled.
4404 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (machine_mode @var{mode})
4405 Define this to return nonzero for machine modes for which the port has
4406 small register classes. If this target hook returns nonzero for a given
4407 @var{mode}, the compiler will try to minimize the lifetime of registers
4408 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4409 In this case, the hook is expected to return nonzero if it returns nonzero
4412 On some machines, it is risky to let hard registers live across arbitrary
4413 insns. Typically, these machines have instructions that require values
4414 to be in specific registers (like an accumulator), and reload will fail
4415 if the required hard register is used for another purpose across such an
4418 Passes before reload do not know which hard registers will be used
4419 in an instruction, but the machine modes of the registers set or used in
4420 the instruction are already known. And for some machines, register
4421 classes are small for, say, integer registers but not for floating point
4422 registers. For example, the AMD x86-64 architecture requires specific
4423 registers for the legacy x86 integer instructions, but there are many
4424 SSE registers for floating point operations. On such targets, a good
4425 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4426 machine modes but zero for the SSE register classes.
4428 The default version of this hook returns false for any mode. It is always
4429 safe to redefine this hook to return with a nonzero value. But if you
4430 unnecessarily define it, you will reduce the amount of optimizations
4431 that can be performed in some cases. If you do not define this hook
4432 to return a nonzero value when it is required, the compiler will run out
4433 of spill registers and print a fatal error message.
4437 @subsection How Scalar Function Values Are Returned
4438 @cindex return values in registers
4439 @cindex values, returned by functions
4440 @cindex scalars, returned as values
4442 This section discusses the macros that control returning scalars as
4443 values---values that can fit in registers.
4445 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4447 Define this to return an RTX representing the place where a function
4448 returns or receives a value of data type @var{ret_type}, a tree node
4449 representing a data type. @var{fn_decl_or_type} is a tree node
4450 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4451 function being called. If @var{outgoing} is false, the hook should
4452 compute the register in which the caller will see the return value.
4453 Otherwise, the hook should return an RTX representing the place where
4454 a function returns a value.
4456 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4457 (Actually, on most machines, scalar values are returned in the same
4458 place regardless of mode.) The value of the expression is usually a
4459 @code{reg} RTX for the hard register where the return value is stored.
4460 The value can also be a @code{parallel} RTX, if the return value is in
4461 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4462 @code{parallel} form. Note that the callee will populate every
4463 location specified in the @code{parallel}, but if the first element of
4464 the @code{parallel} contains the whole return value, callers will use
4465 that element as the canonical location and ignore the others. The m68k
4466 port uses this type of @code{parallel} to return pointers in both
4467 @samp{%a0} (the canonical location) and @samp{%d0}.
4469 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4470 the same promotion rules specified in @code{PROMOTE_MODE} if
4471 @var{valtype} is a scalar type.
4473 If the precise function being called is known, @var{func} is a tree
4474 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4475 pointer. This makes it possible to use a different value-returning
4476 convention for specific functions when all their calls are
4479 Some target machines have ``register windows'' so that the register in
4480 which a function returns its value is not the same as the one in which
4481 the caller sees the value. For such machines, you should return
4482 different RTX depending on @var{outgoing}.
4484 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4485 aggregate data types, because these are returned in another way. See
4486 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4489 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4490 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4491 a new target instead.
4494 @defmac LIBCALL_VALUE (@var{mode})
4495 A C expression to create an RTX representing the place where a library
4496 function returns a value of mode @var{mode}.
4498 Note that ``library function'' in this context means a compiler
4499 support routine, used to perform arithmetic, whose name is known
4500 specially by the compiler and was not mentioned in the C code being
4504 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (machine_mode @var{mode}, const_rtx @var{fun})
4505 Define this hook if the back-end needs to know the name of the libcall
4506 function in order to determine where the result should be returned.
4508 The mode of the result is given by @var{mode} and the name of the called
4509 library function is given by @var{fun}. The hook should return an RTX
4510 representing the place where the library function result will be returned.
4512 If this hook is not defined, then LIBCALL_VALUE will be used.
4515 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4516 A C expression that is nonzero if @var{regno} is the number of a hard
4517 register in which the values of called function may come back.
4519 A register whose use for returning values is limited to serving as the
4520 second of a pair (for a value of type @code{double}, say) need not be
4521 recognized by this macro. So for most machines, this definition
4525 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4528 If the machine has register windows, so that the caller and the called
4529 function use different registers for the return value, this macro
4530 should recognize only the caller's register numbers.
4532 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4533 for a new target instead.
4536 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4537 A target hook that return @code{true} if @var{regno} is the number of a hard
4538 register in which the values of called function may come back.
4540 A register whose use for returning values is limited to serving as the
4541 second of a pair (for a value of type @code{double}, say) need not be
4542 recognized by this target hook.
4544 If the machine has register windows, so that the caller and the called
4545 function use different registers for the return value, this target hook
4546 should recognize only the caller's register numbers.
4548 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4551 @defmac APPLY_RESULT_SIZE
4552 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4553 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4554 saving and restoring an arbitrary return value.
4557 @deftypevr {Target Hook} bool TARGET_OMIT_STRUCT_RETURN_REG
4558 Normally, when a function returns a structure by memory, the address
4559 is passed as an invisible pointer argument, but the compiler also
4560 arranges to return the address from the function like it would a normal
4561 pointer return value. Define this to true if that behavior is
4562 undesirable on your target.
4565 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4566 This hook should return true if values of type @var{type} are returned
4567 at the most significant end of a register (in other words, if they are
4568 padded at the least significant end). You can assume that @var{type}
4569 is returned in a register; the caller is required to check this.
4571 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4572 be able to hold the complete return value. For example, if a 1-, 2-
4573 or 3-byte structure is returned at the most significant end of a
4574 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4578 @node Aggregate Return
4579 @subsection How Large Values Are Returned
4580 @cindex aggregates as return values
4581 @cindex large return values
4582 @cindex returning aggregate values
4583 @cindex structure value address
4585 When a function value's mode is @code{BLKmode} (and in some other
4586 cases), the value is not returned according to
4587 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4588 caller passes the address of a block of memory in which the value
4589 should be stored. This address is called the @dfn{structure value
4592 This section describes how to control returning structure values in
4595 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4596 This target hook should return a nonzero value to say to return the
4597 function value in memory, just as large structures are always returned.
4598 Here @var{type} will be the data type of the value, and @var{fntype}
4599 will be the type of the function doing the returning, or @code{NULL} for
4602 Note that values of mode @code{BLKmode} must be explicitly handled
4603 by this function. Also, the option @option{-fpcc-struct-return}
4604 takes effect regardless of this macro. On most systems, it is
4605 possible to leave the hook undefined; this causes a default
4606 definition to be used, whose value is the constant 1 for @code{BLKmode}
4607 values, and 0 otherwise.
4609 Do not use this hook to indicate that structures and unions should always
4610 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4614 @defmac DEFAULT_PCC_STRUCT_RETURN
4615 Define this macro to be 1 if all structure and union return values must be
4616 in memory. Since this results in slower code, this should be defined
4617 only if needed for compatibility with other compilers or with an ABI@.
4618 If you define this macro to be 0, then the conventions used for structure
4619 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4622 If not defined, this defaults to the value 1.
4625 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4626 This target hook should return the location of the structure value
4627 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4628 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4629 be @code{NULL}, for libcalls. You do not need to define this target
4630 hook if the address is always passed as an ``invisible'' first
4633 On some architectures the place where the structure value address
4634 is found by the called function is not the same place that the
4635 caller put it. This can be due to register windows, or it could
4636 be because the function prologue moves it to a different place.
4637 @var{incoming} is @code{1} or @code{2} when the location is needed in
4638 the context of the called function, and @code{0} in the context of
4641 If @var{incoming} is nonzero and the address is to be found on the
4642 stack, return a @code{mem} which refers to the frame pointer. If
4643 @var{incoming} is @code{2}, the result is being used to fetch the
4644 structure value address at the beginning of a function. If you need
4645 to emit adjusting code, you should do it at this point.
4648 @defmac PCC_STATIC_STRUCT_RETURN
4649 Define this macro if the usual system convention on the target machine
4650 for returning structures and unions is for the called function to return
4651 the address of a static variable containing the value.
4653 Do not define this if the usual system convention is for the caller to
4654 pass an address to the subroutine.
4656 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4657 nothing when you use @option{-freg-struct-return} mode.
4660 @deftypefn {Target Hook} fixed_size_mode TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4661 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.
4664 @deftypefn {Target Hook} fixed_size_mode TARGET_GET_RAW_ARG_MODE (int @var{regno})
4665 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.
4668 @deftypefn {Target Hook} bool TARGET_EMPTY_RECORD_P (const_tree @var{type})
4669 This target hook returns true if the type is an empty record. The default
4670 is to return @code{false}.
4673 @deftypefn {Target Hook} void TARGET_WARN_PARAMETER_PASSING_ABI (cumulative_args_t @var{ca}, tree @var{type})
4674 This target hook warns about the change in empty class parameter passing
4679 @subsection Caller-Saves Register Allocation
4681 If you enable it, GCC can save registers around function calls. This
4682 makes it possible to use call-clobbered registers to hold variables that
4683 must live across calls.
4685 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4686 A C expression specifying which mode is required for saving @var{nregs}
4687 of a pseudo-register in call-clobbered hard register @var{regno}. If
4688 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4689 returned. For most machines this macro need not be defined since GCC
4690 will select the smallest suitable mode.
4693 @node Function Entry
4694 @subsection Function Entry and Exit
4695 @cindex function entry and exit
4699 This section describes the macros that output function entry
4700 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4702 @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})
4703 Generate a patchable area at the function start, consisting of
4704 @var{patch_area_size} NOP instructions. If the target supports named
4705 sections and if @var{record_p} is true, insert a pointer to the current
4706 location in the table of patchable functions. The default implementation
4707 of the hook places the table of pointers in the special section named
4708 @code{__patchable_function_entries}.
4711 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file})
4712 If defined, a function that outputs the assembler code for entry to a
4713 function. The prologue is responsible for setting up the stack frame,
4714 initializing the frame pointer register, saving registers that must be
4715 saved, and allocating @var{size} additional bytes of storage for the
4716 local variables. @var{file} is a stdio stream to which the assembler
4717 code should be output.
4719 The label for the beginning of the function need not be output by this
4720 macro. That has already been done when the macro is run.
4722 @findex regs_ever_live
4723 To determine which registers to save, the macro can refer to the array
4724 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4725 @var{r} is used anywhere within the function. This implies the function
4726 prologue should save register @var{r}, provided it is not one of the
4727 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4728 @code{regs_ever_live}.)
4730 On machines that have ``register windows'', the function entry code does
4731 not save on the stack the registers that are in the windows, even if
4732 they are supposed to be preserved by function calls; instead it takes
4733 appropriate steps to ``push'' the register stack, if any non-call-used
4734 registers are used in the function.
4736 @findex frame_pointer_needed
4737 On machines where functions may or may not have frame-pointers, the
4738 function entry code must vary accordingly; it must set up the frame
4739 pointer if one is wanted, and not otherwise. To determine whether a
4740 frame pointer is in wanted, the macro can refer to the variable
4741 @code{frame_pointer_needed}. The variable's value will be 1 at run
4742 time in a function that needs a frame pointer. @xref{Elimination}.
4744 The function entry code is responsible for allocating any stack space
4745 required for the function. This stack space consists of the regions
4746 listed below. In most cases, these regions are allocated in the
4747 order listed, with the last listed region closest to the top of the
4748 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4749 the highest address if it is not defined). You can use a different order
4750 for a machine if doing so is more convenient or required for
4751 compatibility reasons. Except in cases where required by standard
4752 or by a debugger, there is no reason why the stack layout used by GCC
4753 need agree with that used by other compilers for a machine.
4756 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4757 If defined, a function that outputs assembler code at the end of a
4758 prologue. This should be used when the function prologue is being
4759 emitted as RTL, and you have some extra assembler that needs to be
4760 emitted. @xref{prologue instruction pattern}.
4763 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4764 If defined, a function that outputs assembler code at the start of an
4765 epilogue. This should be used when the function epilogue is being
4766 emitted as RTL, and you have some extra assembler that needs to be
4767 emitted. @xref{epilogue instruction pattern}.
4770 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file})
4771 If defined, a function that outputs the assembler code for exit from a
4772 function. The epilogue is responsible for restoring the saved
4773 registers and stack pointer to their values when the function was
4774 called, and returning control to the caller. This macro takes the
4775 same argument as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4776 registers to restore are determined from @code{regs_ever_live} and
4777 @code{CALL_USED_REGISTERS} in the same way.
4779 On some machines, there is a single instruction that does all the work
4780 of returning from the function. On these machines, give that
4781 instruction the name @samp{return} and do not define the macro
4782 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4784 Do not define a pattern named @samp{return} if you want the
4785 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4786 switches to control whether return instructions or epilogues are used,
4787 define a @samp{return} pattern with a validity condition that tests the
4788 target switches appropriately. If the @samp{return} pattern's validity
4789 condition is false, epilogues will be used.
4791 On machines where functions may or may not have frame-pointers, the
4792 function exit code must vary accordingly. Sometimes the code for these
4793 two cases is completely different. To determine whether a frame pointer
4794 is wanted, the macro can refer to the variable
4795 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4796 a function that needs a frame pointer.
4798 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4799 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4800 The C variable @code{current_function_is_leaf} is nonzero for such a
4801 function. @xref{Leaf Functions}.
4803 On some machines, some functions pop their arguments on exit while
4804 others leave that for the caller to do. For example, the 68020 when
4805 given @option{-mrtd} pops arguments in functions that take a fixed
4806 number of arguments.
4809 @findex crtl->args.pops_args
4810 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4811 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4812 needs to know what was decided. The number of bytes of the current
4813 function's arguments that this function should pop is available in
4814 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4819 @findex pretend_args_size
4820 @findex crtl->args.pretend_args_size
4821 A region of @code{crtl->args.pretend_args_size} bytes of
4822 uninitialized space just underneath the first argument arriving on the
4823 stack. (This may not be at the very start of the allocated stack region
4824 if the calling sequence has pushed anything else since pushing the stack
4825 arguments. But usually, on such machines, nothing else has been pushed
4826 yet, because the function prologue itself does all the pushing.) This
4827 region is used on machines where an argument may be passed partly in
4828 registers and partly in memory, and, in some cases to support the
4829 features in @code{<stdarg.h>}.
4832 An area of memory used to save certain registers used by the function.
4833 The size of this area, which may also include space for such things as
4834 the return address and pointers to previous stack frames, is
4835 machine-specific and usually depends on which registers have been used
4836 in the function. Machines with register windows often do not require
4840 A region of at least @var{size} bytes, possibly rounded up to an allocation
4841 boundary, to contain the local variables of the function. On some machines,
4842 this region and the save area may occur in the opposite order, with the
4843 save area closer to the top of the stack.
4846 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4847 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4848 @code{crtl->outgoing_args_size} bytes to be used for outgoing
4849 argument lists of the function. @xref{Stack Arguments}.
4852 @defmac EXIT_IGNORE_STACK
4853 Define this macro as a C expression that is nonzero if the return
4854 instruction or the function epilogue ignores the value of the stack
4855 pointer; in other words, if it is safe to delete an instruction to
4856 adjust the stack pointer before a return from the function. The
4859 Note that this macro's value is relevant only for functions for which
4860 frame pointers are maintained. It is never safe to delete a final
4861 stack adjustment in a function that has no frame pointer, and the
4862 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4865 @defmac EPILOGUE_USES (@var{regno})
4866 Define this macro as a C expression that is nonzero for registers that are
4867 used by the epilogue or the @samp{return} pattern. The stack and frame
4868 pointer registers are already assumed to be used as needed.
4871 @defmac EH_USES (@var{regno})
4872 Define this macro as a C expression that is nonzero for registers that are
4873 used by the exception handling mechanism, and so should be considered live
4874 on entry to an exception edge.
4877 @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})
4878 A function that outputs the assembler code for a thunk
4879 function, used to implement C++ virtual function calls with multiple
4880 inheritance. The thunk acts as a wrapper around a virtual function,
4881 adjusting the implicit object parameter before handing control off to
4884 First, emit code to add the integer @var{delta} to the location that
4885 contains the incoming first argument. Assume that this argument
4886 contains a pointer, and is the one used to pass the @code{this} pointer
4887 in C++. This is the incoming argument @emph{before} the function prologue,
4888 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4889 all other incoming arguments.
4891 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4892 made after adding @code{delta}. In particular, if @var{p} is the
4893 adjusted pointer, the following adjustment should be made:
4896 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4899 After the additions, emit code to jump to @var{function}, which is a
4900 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4901 not touch the return address. Hence returning from @var{FUNCTION} will
4902 return to whoever called the current @samp{thunk}.
4904 The effect must be as if @var{function} had been called directly with
4905 the adjusted first argument. This macro is responsible for emitting all
4906 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4907 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4909 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4910 have already been extracted from it.) It might possibly be useful on
4911 some targets, but probably not.
4913 If you do not define this macro, the target-independent code in the C++
4914 front end will generate a less efficient heavyweight thunk that calls
4915 @var{function} instead of jumping to it. The generic approach does
4916 not support varargs.
4919 @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})
4920 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4921 to output the assembler code for the thunk function specified by the
4922 arguments it is passed, and false otherwise. In the latter case, the
4923 generic approach will be used by the C++ front end, with the limitations
4928 @subsection Generating Code for Profiling
4929 @cindex profiling, code generation
4931 These macros will help you generate code for profiling.
4933 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4934 A C statement or compound statement to output to @var{file} some
4935 assembler code to call the profiling subroutine @code{mcount}.
4938 The details of how @code{mcount} expects to be called are determined by
4939 your operating system environment, not by GCC@. To figure them out,
4940 compile a small program for profiling using the system's installed C
4941 compiler and look at the assembler code that results.
4943 Older implementations of @code{mcount} expect the address of a counter
4944 variable to be loaded into some register. The name of this variable is
4945 @samp{LP} followed by the number @var{labelno}, so you would generate
4946 the name using @samp{LP%d} in a @code{fprintf}.
4949 @defmac PROFILE_HOOK
4950 A C statement or compound statement to output to @var{file} some assembly
4951 code to call the profiling subroutine @code{mcount} even the target does
4952 not support profiling.
4955 @defmac NO_PROFILE_COUNTERS
4956 Define this macro to be an expression with a nonzero value if the
4957 @code{mcount} subroutine on your system does not need a counter variable
4958 allocated for each function. This is true for almost all modern
4959 implementations. If you define this macro, you must not use the
4960 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4963 @defmac PROFILE_BEFORE_PROLOGUE
4964 Define this macro if the code for function profiling should come before
4965 the function prologue. Normally, the profiling code comes after.
4968 @deftypefn {Target Hook} bool TARGET_KEEP_LEAF_WHEN_PROFILED (void)
4969 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.
4973 @subsection Permitting tail calls
4976 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4977 True if it is OK to do sibling call optimization for the specified
4978 call expression @var{exp}. @var{decl} will be the called function,
4979 or @code{NULL} if this is an indirect call.
4981 It is not uncommon for limitations of calling conventions to prevent
4982 tail calls to functions outside the current unit of translation, or
4983 during PIC compilation. The hook is used to enforce these restrictions,
4984 as the @code{sibcall} md pattern cannot fail, or fall over to a
4985 ``normal'' call. The criteria for successful sibling call optimization
4986 may vary greatly between different architectures.
4989 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4990 Add any hard registers to @var{regs} that are live on entry to the
4991 function. This hook only needs to be defined to provide registers that
4992 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4993 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4994 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4995 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4998 @deftypefn {Target Hook} void TARGET_SET_UP_BY_PROLOGUE (struct hard_reg_set_container *@var{})
4999 This hook should add additional registers that are computed by the prologue to the hard regset for shrink-wrapping optimization purposes.
5002 @deftypefn {Target Hook} bool TARGET_WARN_FUNC_RETURN (tree)
5003 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.
5006 @node Shrink-wrapping separate components
5007 @subsection Shrink-wrapping separate components
5008 @cindex shrink-wrapping separate components
5010 The prologue may perform a variety of target dependent tasks such as
5011 saving callee-saved registers, saving the return address, aligning the
5012 stack, creating a stack frame, initializing the PIC register, setting
5013 up the static chain, etc.
5015 On some targets some of these tasks may be independent of others and
5016 thus may be shrink-wrapped separately. These independent tasks are
5017 referred to as components and are handled generically by the target
5018 independent parts of GCC.
5020 Using the following hooks those prologue or epilogue components can be
5021 shrink-wrapped separately, so that the initialization (and possibly
5022 teardown) those components do is not done as frequently on execution
5023 paths where this would unnecessary.
5025 What exactly those components are is up to the target code; the generic
5026 code treats them abstractly, as a bit in an @code{sbitmap}. These
5027 @code{sbitmap}s are allocated by the @code{shrink_wrap.get_separate_components}
5028 and @code{shrink_wrap.components_for_bb} hooks, and deallocated by the
5031 @deftypefn {Target Hook} sbitmap TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS (void)
5032 This hook should return an @code{sbitmap} with the bits set for those
5033 components that can be separately shrink-wrapped in the current function.
5034 Return @code{NULL} if the current function should not get any separate
5036 Don't define this hook if it would always return @code{NULL}.
5037 If it is defined, the other hooks in this group have to be defined as well.
5040 @deftypefn {Target Hook} sbitmap TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB (basic_block)
5041 This hook should return an @code{sbitmap} with the bits set for those
5042 components where either the prologue component has to be executed before
5043 the @code{basic_block}, or the epilogue component after it, or both.
5046 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS (sbitmap @var{components}, edge @var{e}, sbitmap @var{edge_components}, bool @var{is_prologue})
5047 This hook should clear the bits in the @var{components} bitmap for those
5048 components in @var{edge_components} that the target cannot handle on edge
5049 @var{e}, where @var{is_prologue} says if this is for a prologue or an
5053 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS (sbitmap)
5054 Emit prologue insns for the components indicated by the parameter.
5057 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS (sbitmap)
5058 Emit epilogue insns for the components indicated by the parameter.
5061 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS (sbitmap)
5062 Mark the components in the parameter as handled, so that the
5063 @code{prologue} and @code{epilogue} named patterns know to ignore those
5064 components. The target code should not hang on to the @code{sbitmap}, it
5065 will be deleted after this call.
5068 @node Stack Smashing Protection
5069 @subsection Stack smashing protection
5070 @cindex stack smashing protection
5072 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
5073 This hook returns a @code{DECL} node for the external variable to use
5074 for the stack protection guard. This variable is initialized by the
5075 runtime to some random value and is used to initialize the guard value
5076 that is placed at the top of the local stack frame. The type of this
5077 variable must be @code{ptr_type_node}.
5079 The default version of this hook creates a variable called
5080 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
5083 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
5084 This hook returns a @code{CALL_EXPR} that alerts the runtime that the
5085 stack protect guard variable has been modified. This expression should
5086 involve a call to a @code{noreturn} function.
5088 The default version of this hook invokes a function called
5089 @samp{__stack_chk_fail}, taking no arguments. This function is
5090 normally defined in @file{libgcc2.c}.
5093 @deftypefn {Target Hook} bool TARGET_STACK_PROTECT_RUNTIME_ENABLED_P (void)
5094 Returns true if the target wants GCC's default stack protect runtime support, otherwise return false. The default implementation always returns true.
5097 @deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
5098 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
5101 @deftypefn {Common Target Hook} {vec<const char *>} TARGET_GET_VALID_OPTION_VALUES (int @var{option_code}, const char *@var{prefix})
5102 The hook is used for options that have a non-trivial list of possible option values. OPTION_CODE is option code of opt_code enum type. PREFIX is used for bash completion and allows an implementation to return more specific completion based on the prefix. All string values should be allocated from heap memory and consumers should release them. The result will be pruned to cases with PREFIX if not NULL.
5105 @node Miscellaneous Register Hooks
5106 @subsection Miscellaneous register hooks
5107 @cindex miscellaneous register hooks
5109 @deftypevr {Target Hook} bool TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
5110 Set to true if each call that binds to a local definition explicitly
5111 clobbers or sets all non-fixed registers modified by performing the call.
5112 That is, by the call pattern itself, or by code that might be inserted by the
5113 linker (e.g.@: stubs, veneers, branch islands), but not including those
5114 modifiable by the callee. The affected registers may be mentioned explicitly
5115 in the call pattern, or included as clobbers in CALL_INSN_FUNCTION_USAGE.
5116 The default version of this hook is set to false. The purpose of this hook
5117 is to enable the fipa-ra optimization.
5121 @section Implementing the Varargs Macros
5122 @cindex varargs implementation
5124 GCC comes with an implementation of @code{<varargs.h>} and
5125 @code{<stdarg.h>} that work without change on machines that pass arguments
5126 on the stack. Other machines require their own implementations of
5127 varargs, and the two machine independent header files must have
5128 conditionals to include it.
5130 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
5131 the calling convention for @code{va_start}. The traditional
5132 implementation takes just one argument, which is the variable in which
5133 to store the argument pointer. The ISO implementation of
5134 @code{va_start} takes an additional second argument. The user is
5135 supposed to write the last named argument of the function here.
5137 However, @code{va_start} should not use this argument. The way to find
5138 the end of the named arguments is with the built-in functions described
5141 @defmac __builtin_saveregs ()
5142 Use this built-in function to save the argument registers in memory so
5143 that the varargs mechanism can access them. Both ISO and traditional
5144 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
5145 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
5147 On some machines, @code{__builtin_saveregs} is open-coded under the
5148 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
5149 other machines, it calls a routine written in assembler language,
5150 found in @file{libgcc2.c}.
5152 Code generated for the call to @code{__builtin_saveregs} appears at the
5153 beginning of the function, as opposed to where the call to
5154 @code{__builtin_saveregs} is written, regardless of what the code is.
5155 This is because the registers must be saved before the function starts
5156 to use them for its own purposes.
5157 @c i rewrote the first sentence above to fix an overfull hbox. --mew
5161 @defmac __builtin_next_arg (@var{lastarg})
5162 This builtin returns the address of the first anonymous stack
5163 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
5164 returns the address of the location above the first anonymous stack
5165 argument. Use it in @code{va_start} to initialize the pointer for
5166 fetching arguments from the stack. Also use it in @code{va_start} to
5167 verify that the second parameter @var{lastarg} is the last named argument
5168 of the current function.
5171 @defmac __builtin_classify_type (@var{object})
5172 Since each machine has its own conventions for which data types are
5173 passed in which kind of register, your implementation of @code{va_arg}
5174 has to embody these conventions. The easiest way to categorize the
5175 specified data type is to use @code{__builtin_classify_type} together
5176 with @code{sizeof} and @code{__alignof__}.
5178 @code{__builtin_classify_type} ignores the value of @var{object},
5179 considering only its data type. It returns an integer describing what
5180 kind of type that is---integer, floating, pointer, structure, and so on.
5182 The file @file{typeclass.h} defines an enumeration that you can use to
5183 interpret the values of @code{__builtin_classify_type}.
5186 These machine description macros help implement varargs:
5188 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5189 If defined, this hook produces the machine-specific code for a call to
5190 @code{__builtin_saveregs}. This code will be moved to the very
5191 beginning of the function, before any parameter access are made. The
5192 return value of this function should be an RTX that contains the value
5193 to use as the return of @code{__builtin_saveregs}.
5196 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (cumulative_args_t @var{args_so_far}, const function_arg_info @var{&arg}, int *@var{pretend_args_size}, int @var{second_time})
5197 This target hook offers an alternative to using
5198 @code{__builtin_saveregs} and defining the hook
5199 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5200 register arguments into the stack so that all the arguments appear to
5201 have been passed consecutively on the stack. Once this is done, you can
5202 use the standard implementation of varargs that works for machines that
5203 pass all their arguments on the stack.
5205 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5206 structure, containing the values that are obtained after processing the
5207 named arguments. The argument @var{arg} describes the last of these named
5210 The target hook should do two things: first, push onto the stack all the
5211 argument registers @emph{not} used for the named arguments, and second,
5212 store the size of the data thus pushed into the @code{int}-valued
5213 variable pointed to by @var{pretend_args_size}. The value that you
5214 store here will serve as additional offset for setting up the stack
5217 Because you must generate code to push the anonymous arguments at
5218 compile time without knowing their data types,
5219 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5220 have just a single category of argument register and use it uniformly
5223 If the argument @var{second_time} is nonzero, it means that the
5224 arguments of the function are being analyzed for the second time. This
5225 happens for an inline function, which is not actually compiled until the
5226 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5227 not generate any instructions in this case.
5230 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca})
5231 Define this hook to return @code{true} if the location where a function
5232 argument is passed depends on whether or not it is a named argument.
5234 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5235 is set for varargs and stdarg functions. If this hook returns
5236 @code{true}, the @var{named} argument is always true for named
5237 arguments, and false for unnamed arguments. If it returns @code{false},
5238 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5239 then all arguments are treated as named. Otherwise, all named arguments
5240 except the last are treated as named.
5242 You need not define this hook if it always returns @code{false}.
5245 @deftypefn {Target Hook} void TARGET_CALL_ARGS (rtx, @var{tree})
5246 While generating RTL for a function call, this target hook is invoked once
5247 for each argument passed to the function, either a register returned by
5248 @code{TARGET_FUNCTION_ARG} or a memory location. It is called just
5249 before the point where argument registers are stored. The type of the
5250 function to be called is also passed as the second argument; it is
5251 @code{NULL_TREE} for libcalls. The @code{TARGET_END_CALL_ARGS} hook is
5252 invoked just after the code to copy the return reg has been emitted.
5253 This functionality can be used to perform special setup of call argument
5254 registers if a target needs it.
5255 For functions without arguments, the hook is called once with @code{pc_rtx}
5256 passed instead of an argument register.
5257 Most ports do not need to implement anything for this hook.
5260 @deftypefn {Target Hook} void TARGET_END_CALL_ARGS (void)
5261 This target hook is invoked while generating RTL for a function call,
5262 just after the point where the return reg is copied into a pseudo. It
5263 signals that all the call argument and return registers for the just
5264 emitted call are now no longer in use.
5265 Most ports do not need to implement anything for this hook.
5268 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca})
5269 If you need to conditionally change ABIs so that one works with
5270 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5271 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5272 defined, then define this hook to return @code{true} if
5273 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5274 Otherwise, you should not define this hook.
5277 @deftypefn {Target Hook} rtx TARGET_LOAD_BOUNDS_FOR_ARG (rtx @var{slot}, rtx @var{arg}, rtx @var{slot_no})
5278 This hook is used by expand pass to emit insn to load bounds of
5279 @var{arg} passed in @var{slot}. Expand pass uses this hook in case
5280 bounds of @var{arg} are not passed in register. If @var{slot} is a
5281 memory, then bounds are loaded as for regular pointer loaded from
5282 memory. If @var{slot} is not a memory then @var{slot_no} is an integer
5283 constant holding number of the target dependent special slot which
5284 should be used to obtain bounds. Hook returns RTX holding loaded bounds.
5287 @deftypefn {Target Hook} void TARGET_STORE_BOUNDS_FOR_ARG (rtx @var{arg}, rtx @var{slot}, rtx @var{bounds}, rtx @var{slot_no})
5288 This hook is used by expand pass to emit insns to store @var{bounds} of
5289 @var{arg} passed in @var{slot}. Expand pass uses this hook in case
5290 @var{bounds} of @var{arg} are not passed in register. If @var{slot} is a
5291 memory, then @var{bounds} are stored as for regular pointer stored in
5292 memory. If @var{slot} is not a memory then @var{slot_no} is an integer
5293 constant holding number of the target dependent special slot which
5294 should be used to store @var{bounds}.
5297 @deftypefn {Target Hook} rtx TARGET_LOAD_RETURNED_BOUNDS (rtx @var{slot})
5298 This hook is used by expand pass to emit insn to load bounds
5299 returned by function call in @var{slot}. Hook returns RTX holding
5303 @deftypefn {Target Hook} void TARGET_STORE_RETURNED_BOUNDS (rtx @var{slot}, rtx @var{bounds})
5304 This hook is used by expand pass to emit insn to store @var{bounds}
5305 returned by function call into @var{slot}.
5309 @section Support for Nested Functions
5310 @cindex support for nested functions
5311 @cindex trampolines for nested functions
5312 @cindex descriptors for nested functions
5313 @cindex nested functions, support for
5315 Taking the address of a nested function requires special compiler
5316 handling to ensure that the static chain register is loaded when
5317 the function is invoked via an indirect call.
5319 GCC has traditionally supported nested functions by creating an
5320 executable @dfn{trampoline} at run time when the address of a nested
5321 function is taken. This is a small piece of code which normally
5322 resides on the stack, in the stack frame of the containing function.
5323 The trampoline loads the static chain register and then jumps to the
5324 real address of the nested function.
5326 The use of trampolines requires an executable stack, which is a
5327 security risk. To avoid this problem, GCC also supports another
5328 strategy: using descriptors for nested functions. Under this model,
5329 taking the address of a nested function results in a pointer to a
5330 non-executable function descriptor object. Initializing the static chain
5331 from the descriptor is handled at indirect call sites.
5333 On some targets, including HPPA and IA-64, function descriptors may be
5334 mandated by the ABI or be otherwise handled in a target-specific way
5335 by the back end in its code generation strategy for indirect calls.
5336 GCC also provides its own generic descriptor implementation to support the
5337 @option{-fno-trampolines} option. In this case runtime detection of
5338 function descriptors at indirect call sites relies on descriptor
5339 pointers being tagged with a bit that is never set in bare function
5340 addresses. Since GCC's generic function descriptors are
5341 not ABI-compliant, this option is typically used only on a
5342 per-language basis (notably by Ada) or when it can otherwise be
5343 applied to the whole program.
5345 Define the following hook if your backend either implements ABI-specified
5346 descriptor support, or can use GCC's generic descriptor implementation
5347 for nested functions.
5349 @deftypevr {Target Hook} int TARGET_CUSTOM_FUNCTION_DESCRIPTORS
5350 If the target can use GCC's generic descriptor mechanism for nested
5351 functions, define this hook to a power of 2 representing an unused bit
5352 in function pointers which can be used to differentiate descriptors at
5353 run time. This value gives the number of bytes by which descriptor
5354 pointers are misaligned compared to function pointers. For example, on
5355 targets that require functions to be aligned to a 4-byte boundary, a
5356 value of either 1 or 2 is appropriate unless the architecture already
5357 reserves the bit for another purpose, such as on ARM.
5359 Define this hook to 0 if the target implements ABI support for
5360 function descriptors in its standard calling sequence, like for example
5363 Using descriptors for nested functions
5364 eliminates the need for trampolines that reside on the stack and require
5365 it to be made executable.
5368 The following macros tell GCC how to generate code to allocate and
5369 initialize an executable trampoline. You can also use this interface
5370 if your back end needs to create ABI-specified non-executable descriptors; in
5371 this case the "trampoline" created is the descriptor containing data only.
5373 The instructions in an executable trampoline must do two things: load
5374 a constant address into the static chain register, and jump to the real
5375 address of the nested function. On CISC machines such as the m68k,
5376 this requires two instructions, a move immediate and a jump. Then the
5377 two addresses exist in the trampoline as word-long immediate operands.
5378 On RISC machines, it is often necessary to load each address into a
5379 register in two parts. Then pieces of each address form separate
5382 The code generated to initialize the trampoline must store the variable
5383 parts---the static chain value and the function address---into the
5384 immediate operands of the instructions. On a CISC machine, this is
5385 simply a matter of copying each address to a memory reference at the
5386 proper offset from the start of the trampoline. On a RISC machine, it
5387 may be necessary to take out pieces of the address and store them
5390 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5391 This hook is called by @code{assemble_trampoline_template} to output,
5392 on the stream @var{f}, assembler code for a block of data that contains
5393 the constant parts of a trampoline. This code should not include a
5394 label---the label is taken care of automatically.
5396 If you do not define this hook, it means no template is needed
5397 for the target. Do not define this hook on systems where the block move
5398 code to copy the trampoline into place would be larger than the code
5399 to generate it on the spot.
5402 @defmac TRAMPOLINE_SECTION
5403 Return the section into which the trampoline template is to be placed
5404 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5407 @defmac TRAMPOLINE_SIZE
5408 A C expression for the size in bytes of the trampoline, as an integer.
5411 @defmac TRAMPOLINE_ALIGNMENT
5412 Alignment required for trampolines, in bits.
5414 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5415 is used for aligning trampolines.
5418 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5419 This hook is called to initialize a trampoline.
5420 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5421 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5422 RTX for the static chain value that should be passed to the function
5425 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5426 first thing this hook should do is emit a block move into @var{m_tramp}
5427 from the memory block returned by @code{assemble_trampoline_template}.
5428 Note that the block move need only cover the constant parts of the
5429 trampoline. If the target isolates the variable parts of the trampoline
5430 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5432 If the target requires any other actions, such as flushing caches or
5433 enabling stack execution, these actions should be performed after
5434 initializing the trampoline proper.
5437 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5438 This hook should perform any machine-specific adjustment in
5439 the address of the trampoline. Its argument contains the address of the
5440 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5441 the address to be used for a function call should be different from the
5442 address at which the template was stored, the different address should
5443 be returned; otherwise @var{addr} should be returned unchanged.
5444 If this hook is not defined, @var{addr} will be used for function calls.
5447 Implementing trampolines is difficult on many machines because they have
5448 separate instruction and data caches. Writing into a stack location
5449 fails to clear the memory in the instruction cache, so when the program
5450 jumps to that location, it executes the old contents.
5452 Here are two possible solutions. One is to clear the relevant parts of
5453 the instruction cache whenever a trampoline is set up. The other is to
5454 make all trampolines identical, by having them jump to a standard
5455 subroutine. The former technique makes trampoline execution faster; the
5456 latter makes initialization faster.
5458 To clear the instruction cache when a trampoline is initialized, define
5459 the following macro.
5461 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5462 If defined, expands to a C expression clearing the @emph{instruction
5463 cache} in the specified interval. The definition of this macro would
5464 typically be a series of @code{asm} statements. Both @var{beg} and
5465 @var{end} are both pointer expressions.
5468 To use a standard subroutine, define the following macro. In addition,
5469 you must make sure that the instructions in a trampoline fill an entire
5470 cache line with identical instructions, or else ensure that the
5471 beginning of the trampoline code is always aligned at the same point in
5472 its cache line. Look in @file{m68k.h} as a guide.
5474 @defmac TRANSFER_FROM_TRAMPOLINE
5475 Define this macro if trampolines need a special subroutine to do their
5476 work. The macro should expand to a series of @code{asm} statements
5477 which will be compiled with GCC@. They go in a library function named
5478 @code{__transfer_from_trampoline}.
5480 If you need to avoid executing the ordinary prologue code of a compiled
5481 C function when you jump to the subroutine, you can do so by placing a
5482 special label of your own in the assembler code. Use one @code{asm}
5483 statement to generate an assembler label, and another to make the label
5484 global. Then trampolines can use that label to jump directly to your
5485 special assembler code.
5489 @section Implicit Calls to Library Routines
5490 @cindex library subroutine names
5491 @cindex @file{libgcc.a}
5493 @c prevent bad page break with this line
5494 Here is an explanation of implicit calls to library routines.
5496 @defmac DECLARE_LIBRARY_RENAMES
5497 This macro, if defined, should expand to a piece of C code that will get
5498 expanded when compiling functions for libgcc.a. It can be used to
5499 provide alternate names for GCC's internal library functions if there
5500 are ABI-mandated names that the compiler should provide.
5503 @findex set_optab_libfunc
5504 @findex init_one_libfunc
5505 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5506 This hook should declare additional library routines or rename
5507 existing ones, using the functions @code{set_optab_libfunc} and
5508 @code{init_one_libfunc} defined in @file{optabs.c}.
5509 @code{init_optabs} calls this macro after initializing all the normal
5512 The default is to do nothing. Most ports don't need to define this hook.
5515 @deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX
5516 If false (the default), internal library routines start with two
5517 underscores. If set to true, these routines start with @code{__gnu_}
5518 instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This
5519 currently only affects functions defined in @file{libgcc2.c}. If this
5520 is set to true, the @file{tm.h} file must also
5521 @code{#define LIBGCC2_GNU_PREFIX}.
5524 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5525 This macro should return @code{true} if the library routine that
5526 implements the floating point comparison operator @var{comparison} in
5527 mode @var{mode} will return a boolean, and @var{false} if it will
5530 GCC's own floating point libraries return tristates from the
5531 comparison operators, so the default returns false always. Most ports
5532 don't need to define this macro.
5535 @defmac TARGET_LIB_INT_CMP_BIASED
5536 This macro should evaluate to @code{true} if the integer comparison
5537 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5538 operand is smaller than the second, 1 to indicate that they are equal,
5539 and 2 to indicate that the first operand is greater than the second.
5540 If this macro evaluates to @code{false} the comparison functions return
5541 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5542 in @file{libgcc.a}, you do not need to define this macro.
5545 @defmac TARGET_HAS_NO_HW_DIVIDE
5546 This macro should be defined if the target has no hardware divide
5547 instructions. If this macro is defined, GCC will use an algorithm which
5548 make use of simple logical and arithmetic operations for 64-bit
5549 division. If the macro is not defined, GCC will use an algorithm which
5550 make use of a 64-bit by 32-bit divide primitive.
5553 @cindex @code{EDOM}, implicit usage
5556 The value of @code{EDOM} on the target machine, as a C integer constant
5557 expression. If you don't define this macro, GCC does not attempt to
5558 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5559 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5562 If you do not define @code{TARGET_EDOM}, then compiled code reports
5563 domain errors by calling the library function and letting it report the
5564 error. If mathematical functions on your system use @code{matherr} when
5565 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5566 that @code{matherr} is used normally.
5569 @cindex @code{errno}, implicit usage
5570 @defmac GEN_ERRNO_RTX
5571 Define this macro as a C expression to create an rtl expression that
5572 refers to the global ``variable'' @code{errno}. (On certain systems,
5573 @code{errno} may not actually be a variable.) If you don't define this
5574 macro, a reasonable default is used.
5577 @deftypefn {Target Hook} bool TARGET_LIBC_HAS_FUNCTION (enum function_class @var{fn_class})
5578 This hook determines whether a function from a class of functions
5579 @var{fn_class} is present in the target C library.
5582 @deftypefn {Target Hook} bool TARGET_LIBC_HAS_FAST_FUNCTION (int @var{fcode})
5583 This hook determines whether a function from a class of functions
5584 @code{(enum function_class)}@var{fcode} has a fast implementation.
5587 @defmac NEXT_OBJC_RUNTIME
5588 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
5589 by default. This calling convention involves passing the object, the selector
5590 and the method arguments all at once to the method-lookup library function.
5591 This is the usual setting when targeting Darwin/Mac OS X systems, which have
5592 the NeXT runtime installed.
5594 If the macro is set to 0, the "GNU" Objective-C message sending convention
5595 will be used by default. This convention passes just the object and the
5596 selector to the method-lookup function, which returns a pointer to the method.
5598 In either case, it remains possible to select code-generation for the alternate
5599 scheme, by means of compiler command line switches.
5602 @node Addressing Modes
5603 @section Addressing Modes
5604 @cindex addressing modes
5606 @c prevent bad page break with this line
5607 This is about addressing modes.
5609 @defmac HAVE_PRE_INCREMENT
5610 @defmacx HAVE_PRE_DECREMENT
5611 @defmacx HAVE_POST_INCREMENT
5612 @defmacx HAVE_POST_DECREMENT
5613 A C expression that is nonzero if the machine supports pre-increment,
5614 pre-decrement, post-increment, or post-decrement addressing respectively.
5617 @defmac HAVE_PRE_MODIFY_DISP
5618 @defmacx HAVE_POST_MODIFY_DISP
5619 A C expression that is nonzero if the machine supports pre- or
5620 post-address side-effect generation involving constants other than
5621 the size of the memory operand.
5624 @defmac HAVE_PRE_MODIFY_REG
5625 @defmacx HAVE_POST_MODIFY_REG
5626 A C expression that is nonzero if the machine supports pre- or
5627 post-address side-effect generation involving a register displacement.
5630 @defmac CONSTANT_ADDRESS_P (@var{x})
5631 A C expression that is 1 if the RTX @var{x} is a constant which
5632 is a valid address. On most machines the default definition of
5633 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5634 is acceptable, but a few machines are more restrictive as to which
5635 constant addresses are supported.
5638 @defmac CONSTANT_P (@var{x})
5639 @code{CONSTANT_P}, which is defined by target-independent code,
5640 accepts integer-values expressions whose values are not explicitly
5641 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5642 expressions and @code{const} arithmetic expressions, in addition to
5643 @code{const_int} and @code{const_double} expressions.
5646 @defmac MAX_REGS_PER_ADDRESS
5647 A number, the maximum number of registers that can appear in a valid
5648 memory address. Note that it is up to you to specify a value equal to
5649 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5653 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5654 A function that returns whether @var{x} (an RTX) is a legitimate memory
5655 address on the target machine for a memory operand of mode @var{mode}.
5657 Legitimate addresses are defined in two variants: a strict variant and a
5658 non-strict one. The @var{strict} parameter chooses which variant is
5659 desired by the caller.
5661 The strict variant is used in the reload pass. It must be defined so
5662 that any pseudo-register that has not been allocated a hard register is
5663 considered a memory reference. This is because in contexts where some
5664 kind of register is required, a pseudo-register with no hard register
5665 must be rejected. For non-hard registers, the strict variant should look
5666 up the @code{reg_renumber} array; it should then proceed using the hard
5667 register number in the array, or treat the pseudo as a memory reference
5668 if the array holds @code{-1}.
5670 The non-strict variant is used in other passes. It must be defined to
5671 accept all pseudo-registers in every context where some kind of
5672 register is required.
5674 Normally, constant addresses which are the sum of a @code{symbol_ref}
5675 and an integer are stored inside a @code{const} RTX to mark them as
5676 constant. Therefore, there is no need to recognize such sums
5677 specifically as legitimate addresses. Normally you would simply
5678 recognize any @code{const} as legitimate.
5680 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5681 sums that are not marked with @code{const}. It assumes that a naked
5682 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5683 naked constant sums as illegitimate addresses, so that none of them will
5684 be given to @code{PRINT_OPERAND_ADDRESS}.
5686 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5687 On some machines, whether a symbolic address is legitimate depends on
5688 the section that the address refers to. On these machines, define the
5689 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5690 into the @code{symbol_ref}, and then check for it here. When you see a
5691 @code{const}, you will have to look inside it to find the
5692 @code{symbol_ref} in order to determine the section. @xref{Assembler
5695 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5696 Some ports are still using a deprecated legacy substitute for
5697 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5701 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5705 and should @code{goto @var{label}} if the address @var{x} is a valid
5706 address on the target machine for a memory operand of mode @var{mode}.
5708 @findex REG_OK_STRICT
5709 Compiler source files that want to use the strict variant of this
5710 macro define the macro @code{REG_OK_STRICT}. You should use an
5711 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5712 that case and the non-strict variant otherwise.
5714 Using the hook is usually simpler because it limits the number of
5715 files that are recompiled when changes are made.
5718 @defmac TARGET_MEM_CONSTRAINT
5719 A single character to be used instead of the default @code{'m'}
5720 character for general memory addresses. This defines the constraint
5721 letter which matches the memory addresses accepted by
5722 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5723 support new address formats in your back end without changing the
5724 semantics of the @code{'m'} constraint. This is necessary in order to
5725 preserve functionality of inline assembly constructs using the
5726 @code{'m'} constraint.
5729 @defmac FIND_BASE_TERM (@var{x})
5730 A C expression to determine the base term of address @var{x},
5731 or to provide a simplified version of @var{x} from which @file{alias.c}
5732 can easily find the base term. This macro is used in only two places:
5733 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5735 It is always safe for this macro to not be defined. It exists so
5736 that alias analysis can understand machine-dependent addresses.
5738 The typical use of this macro is to handle addresses containing
5739 a label_ref or symbol_ref within an UNSPEC@.
5742 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode})
5743 This hook is given an invalid memory address @var{x} for an
5744 operand of mode @var{mode} and should try to return a valid memory
5747 @findex break_out_memory_refs
5748 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5749 and @var{oldx} will be the operand that was given to that function to produce
5752 The code of the hook should not alter the substructure of
5753 @var{x}. If it transforms @var{x} into a more legitimate form, it
5754 should return the new @var{x}.
5756 It is not necessary for this hook to come up with a legitimate address,
5757 with the exception of native TLS addresses (@pxref{Emulated TLS}).
5758 The compiler has standard ways of doing so in all cases. In fact, if
5759 the target supports only emulated TLS, it
5760 is safe to omit this hook or make it return @var{x} if it cannot find
5761 a valid way to legitimize the address. But often a machine-dependent
5762 strategy can generate better code.
5765 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5766 A C compound statement that attempts to replace @var{x}, which is an address
5767 that needs reloading, with a valid memory address for an operand of mode
5768 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5769 It is not necessary to define this macro, but it might be useful for
5770 performance reasons.
5772 For example, on the i386, it is sometimes possible to use a single
5773 reload register instead of two by reloading a sum of two pseudo
5774 registers into a register. On the other hand, for number of RISC
5775 processors offsets are limited so that often an intermediate address
5776 needs to be generated in order to address a stack slot. By defining
5777 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5778 generated for adjacent some stack slots can be made identical, and thus
5781 @emph{Note}: This macro should be used with caution. It is necessary
5782 to know something of how reload works in order to effectively use this,
5783 and it is quite easy to produce macros that build in too much knowledge
5784 of reload internals.
5786 @emph{Note}: This macro must be able to reload an address created by a
5787 previous invocation of this macro. If it fails to handle such addresses
5788 then the compiler may generate incorrect code or abort.
5791 The macro definition should use @code{push_reload} to indicate parts that
5792 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5793 suitable to be passed unaltered to @code{push_reload}.
5795 The code generated by this macro must not alter the substructure of
5796 @var{x}. If it transforms @var{x} into a more legitimate form, it
5797 should assign @var{x} (which will always be a C variable) a new value.
5798 This also applies to parts that you change indirectly by calling
5801 @findex strict_memory_address_p
5802 The macro definition may use @code{strict_memory_address_p} to test if
5803 the address has become legitimate.
5806 If you want to change only a part of @var{x}, one standard way of doing
5807 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5808 single level of rtl. Thus, if the part to be changed is not at the
5809 top level, you'll need to replace first the top level.
5810 It is not necessary for this macro to come up with a legitimate
5811 address; but often a machine-dependent strategy can generate better code.
5814 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr}, addr_space_t @var{addrspace})
5815 This hook returns @code{true} if memory address @var{addr} in address
5816 space @var{addrspace} can have
5817 different meanings depending on the machine mode of the memory
5818 reference it is used for or if the address is valid for some modes
5821 Autoincrement and autodecrement addresses typically have mode-dependent
5822 effects because the amount of the increment or decrement is the size
5823 of the operand being addressed. Some machines have other mode-dependent
5824 addresses. Many RISC machines have no mode-dependent addresses.
5826 You may assume that @var{addr} is a valid address for the machine.
5828 The default version of this hook returns @code{false}.
5831 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (machine_mode @var{mode}, rtx @var{x})
5832 This hook returns true if @var{x} is a legitimate constant for a
5833 @var{mode}-mode immediate operand on the target machine. You can assume that
5834 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5836 The default definition returns true.
5839 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5840 This hook is used to undo the possibly obfuscating effects of the
5841 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5842 macros. Some backend implementations of these macros wrap symbol
5843 references inside an @code{UNSPEC} rtx to represent PIC or similar
5844 addressing modes. This target hook allows GCC's optimizers to understand
5845 the semantics of these opaque @code{UNSPEC}s by converting them back
5846 into their original form.
5849 @deftypefn {Target Hook} bool TARGET_CONST_NOT_OK_FOR_DEBUG_P (rtx @var{x})
5850 This hook should return true if @var{x} should not be emitted into
5854 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (machine_mode @var{mode}, rtx @var{x})
5855 This hook should return true if @var{x} is of a form that cannot (or
5856 should not) be spilled to the constant pool. @var{mode} is the mode
5859 The default version of this hook returns false.
5861 The primary reason to define this hook is to prevent reload from
5862 deciding that a non-legitimate constant would be better reloaded
5863 from the constant pool instead of spilling and reloading a register
5864 holding the constant. This restriction is often true of addresses
5865 of TLS symbols for various targets.
5868 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (machine_mode @var{mode}, const_rtx @var{x})
5869 This hook should return true if pool entries for constant @var{x} can
5870 be placed in an @code{object_block} structure. @var{mode} is the mode
5873 The default version returns false for all constants.
5876 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_DECL_P (const_tree @var{decl})
5877 This hook should return true if pool entries for @var{decl} should
5878 be placed in an @code{object_block} structure.
5880 The default version returns true for all decls.
5883 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (tree @var{fndecl})
5884 This hook should return the DECL of a function that implements the
5885 reciprocal of the machine-specific builtin function @var{fndecl}, or
5886 @code{NULL_TREE} if such a function is not available.
5889 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5890 This hook should return the DECL of a function @var{f} that given an
5891 address @var{addr} as an argument returns a mask @var{m} that can be
5892 used to extract from two vectors the relevant data that resides in
5893 @var{addr} in case @var{addr} is not properly aligned.
5895 The autovectorizer, when vectorizing a load operation from an address
5896 @var{addr} that may be unaligned, will generate two vector loads from
5897 the two aligned addresses around @var{addr}. It then generates a
5898 @code{REALIGN_LOAD} operation to extract the relevant data from the
5899 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5900 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5901 the third argument, @var{OFF}, defines how the data will be extracted
5902 from these two vectors: if @var{OFF} is 0, then the returned vector is
5903 @var{v2}; otherwise, the returned vector is composed from the last
5904 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5905 @var{OFF} elements of @var{v2}.
5907 If this hook is defined, the autovectorizer will generate a call
5908 to @var{f} (using the DECL tree that this hook returns) and will
5909 use the return value of @var{f} as the argument @var{OFF} to
5910 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5911 should comply with the semantics expected by @code{REALIGN_LOAD}
5913 If this hook is not defined, then @var{addr} will be used as
5914 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5915 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5918 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5919 Returns cost of different scalar or vector statements for vectorization cost model.
5920 For vector memory operations the cost may depend on type (@var{vectype}) and
5921 misalignment value (@var{misalign}).
5924 @deftypefn {Target Hook} poly_uint64 TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT (const_tree @var{type})
5925 This hook returns the preferred alignment in bits for accesses to
5926 vectors of type @var{type} in vectorized code. This might be less than
5927 or greater than the ABI-defined value returned by
5928 @code{TARGET_VECTOR_ALIGNMENT}. It can be equal to the alignment of
5929 a single element, in which case the vectorizer will not try to optimize
5932 The default hook returns @code{TYPE_ALIGN (@var{type})}, which is
5933 correct for most targets.
5936 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5937 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.
5940 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VEC_PERM_CONST (machine_mode @var{mode}, rtx @var{output}, rtx @var{in0}, rtx @var{in1}, const vec_perm_indices @var{&sel})
5941 This hook is used to test whether the target can permute up to two
5942 vectors of mode @var{mode} using the permutation vector @code{sel}, and
5943 also to emit such a permutation. In the former case @var{in0}, @var{in1}
5944 and @var{out} are all null. In the latter case @var{in0} and @var{in1} are
5945 the source vectors and @var{out} is the destination vector; all three are
5946 registers of mode @var{mode}. @var{in1} is the same as @var{in0} if
5947 @var{sel} describes a permutation on one vector instead of two.
5949 Return true if the operation is possible, emitting instructions for it
5950 if rtxes are provided.
5952 @cindex @code{vec_perm@var{m}} instruction pattern
5953 If the hook returns false for a mode with multibyte elements, GCC will
5954 try the equivalent byte operation. If that also fails, it will try forcing
5955 the selector into a register and using the @var{vec_perm@var{mode}}
5956 instruction pattern. There is no need for the hook to handle these two
5957 implementation approaches itself.
5960 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5961 This hook should return the DECL of a function that implements conversion of the
5962 input vector of type @var{src_type} to type @var{dest_type}.
5963 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5964 specifies how the conversion is to be applied
5965 (truncation, rounding, etc.).
5967 If this hook is defined, the autovectorizer will use the
5968 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5969 conversion. Otherwise, it will return @code{NULL_TREE}.
5972 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (unsigned @var{code}, tree @var{vec_type_out}, tree @var{vec_type_in})
5973 This hook should return the decl of a function that implements the
5974 vectorized variant of the function with the @code{combined_fn} code
5975 @var{code} or @code{NULL_TREE} if such a function is not available.
5976 The return type of the vectorized function shall be of vector type
5977 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5980 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5981 This hook should return the decl of a function that implements the
5982 vectorized variant of target built-in function @code{fndecl}. The
5983 return type of the vectorized function shall be of vector type
5984 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5987 @deftypefn {Target Hook} bool TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT (machine_mode @var{mode}, const_tree @var{type}, int @var{misalignment}, bool @var{is_packed})
5988 This hook should return true if the target supports misaligned vector
5989 store/load of a specific factor denoted in the @var{misalignment}
5990 parameter. The vector store/load should be of machine mode @var{mode} and
5991 the elements in the vectors should be of type @var{type}. @var{is_packed}
5992 parameter is true if the memory access is defined in a packed struct.
5995 @deftypefn {Target Hook} machine_mode TARGET_VECTORIZE_PREFERRED_SIMD_MODE (scalar_mode @var{mode})
5996 This hook should return the preferred mode for vectorizing scalar
5997 mode @var{mode}. The default is
5998 equal to @code{word_mode}, because the vectorizer can do some
5999 transformations even in absence of specialized @acronym{SIMD} hardware.
6002 @deftypefn {Target Hook} machine_mode TARGET_VECTORIZE_SPLIT_REDUCTION (machine_mode)
6003 This hook should return the preferred mode to split the final reduction
6004 step on @var{mode} to. The reduction is then carried out reducing upper
6005 against lower halves of vectors recursively until the specified mode is
6006 reached. The default is @var{mode} which means no splitting.
6009 @deftypefn {Target Hook} void TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (vector_sizes *@var{sizes}, bool @var{all})
6010 If the mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE} is not
6011 the only one that is worth considering, this hook should add all suitable
6012 vector sizes to @var{sizes}, in order of decreasing preference. The first
6013 one should be the size of @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
6014 If @var{all} is true, add suitable vector sizes even when they are generally
6015 not expected to be worthwhile.
6017 The hook does not need to do anything if the vector returned by
6018 @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE} is the only one relevant
6019 for autovectorization. The default implementation does nothing.
6022 @deftypefn {Target Hook} opt_machine_mode TARGET_VECTORIZE_GET_MASK_MODE (poly_uint64 @var{nunits}, poly_uint64 @var{length})
6023 A vector mask is a value that holds one boolean result for every element
6024 in a vector. This hook returns the machine mode that should be used to
6025 represent such a mask when the vector in question is @var{length} bytes
6026 long and contains @var{nunits} elements. The hook returns an empty
6027 @code{opt_machine_mode} if no such mode exists.
6029 The default implementation returns the mode of an integer vector that
6030 is @var{length} bytes long and that contains @var{nunits} elements,
6031 if such a mode exists.
6034 @deftypefn {Target Hook} bool TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE (unsigned @var{ifn})
6035 This hook returns true if masked internal function @var{ifn} (really of
6036 type @code{internal_fn}) should be considered expensive when the mask is
6037 all zeros. GCC can then try to branch around the instruction instead.
6040 @deftypefn {Target Hook} {void *} TARGET_VECTORIZE_INIT_COST (class loop *@var{loop_info})
6041 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.
6044 @deftypefn {Target Hook} unsigned TARGET_VECTORIZE_ADD_STMT_COST (void *@var{data}, int @var{count}, enum vect_cost_for_stmt @var{kind}, class _stmt_vec_info *@var{stmt_info}, int @var{misalign}, enum vect_cost_model_location @var{where})
6045 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.
6048 @deftypefn {Target Hook} void TARGET_VECTORIZE_FINISH_COST (void *@var{data}, unsigned *@var{prologue_cost}, unsigned *@var{body_cost}, unsigned *@var{epilogue_cost})
6049 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.
6052 @deftypefn {Target Hook} void TARGET_VECTORIZE_DESTROY_COST_DATA (void *@var{data})
6053 This hook should release @var{data} and any related data structures allocated by TARGET_VECTORIZE_INIT_COST. The default releases the accumulator.
6056 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_GATHER (const_tree @var{mem_vectype}, const_tree @var{index_type}, int @var{scale})
6057 Target builtin that implements vector gather operation. @var{mem_vectype}
6058 is the vector type of the load and @var{index_type} is scalar type of
6059 the index, scaled by @var{scale}.
6060 The default is @code{NULL_TREE} which means to not vectorize gather
6064 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_SCATTER (const_tree @var{vectype}, const_tree @var{index_type}, int @var{scale})
6065 Target builtin that implements vector scatter operation. @var{vectype}
6066 is the vector type of the store and @var{index_type} is scalar type of
6067 the index, scaled by @var{scale}.
6068 The default is @code{NULL_TREE} which means to not vectorize scatter
6072 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN (struct cgraph_node *@var{}, struct cgraph_simd_clone *@var{}, @var{tree}, @var{int})
6073 This hook should set @var{vecsize_mangle}, @var{vecsize_int}, @var{vecsize_float}
6074 fields in @var{simd_clone} structure pointed by @var{clone_info} argument and also
6075 @var{simdlen} field if it was previously 0.
6076 The hook should return 0 if SIMD clones shouldn't be emitted,
6077 or number of @var{vecsize_mangle} variants that should be emitted.
6080 @deftypefn {Target Hook} void TARGET_SIMD_CLONE_ADJUST (struct cgraph_node *@var{})
6081 This hook should add implicit @code{attribute(target("..."))} attribute
6082 to SIMD clone @var{node} if needed.
6085 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_USABLE (struct cgraph_node *@var{})
6086 This hook should return -1 if SIMD clone @var{node} shouldn't be used
6087 in vectorized loops in current function, or non-negative number if it is
6088 usable. In that case, the smaller the number is, the more desirable it is
6092 @deftypefn {Target Hook} int TARGET_SIMT_VF (void)
6093 Return number of threads in SIMT thread group on the target.
6096 @deftypefn {Target Hook} bool TARGET_GOACC_VALIDATE_DIMS (tree @var{decl}, int *@var{dims}, int @var{fn_level}, unsigned @var{used})
6097 This hook should check the launch dimensions provided for an OpenACC
6098 compute region, or routine. Defaulted values are represented as -1
6099 and non-constant values as 0. The @var{fn_level} is negative for the
6100 function corresponding to the compute region. For a routine is is the
6101 outermost level at which partitioned execution may be spawned. The hook
6102 should verify non-default values. If DECL is NULL, global defaults
6103 are being validated and unspecified defaults should be filled in.
6104 Diagnostics should be issued as appropriate. Return
6105 true, if changes have been made. You must override this hook to
6106 provide dimensions larger than 1.
6109 @deftypefn {Target Hook} int TARGET_GOACC_DIM_LIMIT (int @var{axis})
6110 This hook should return the maximum size of a particular dimension,
6111 or zero if unbounded.
6114 @deftypefn {Target Hook} bool TARGET_GOACC_FORK_JOIN (gcall *@var{call}, const int *@var{dims}, bool @var{is_fork})
6115 This hook can be used to convert IFN_GOACC_FORK and IFN_GOACC_JOIN
6116 function calls to target-specific gimple, or indicate whether they
6117 should be retained. It is executed during the oacc_device_lower pass.
6118 It should return true, if the call should be retained. It should
6119 return false, if it is to be deleted (either because target-specific
6120 gimple has been inserted before it, or there is no need for it).
6121 The default hook returns false, if there are no RTL expanders for them.
6124 @deftypefn {Target Hook} void TARGET_GOACC_REDUCTION (gcall *@var{call})
6125 This hook is used by the oacc_transform pass to expand calls to the
6126 @var{GOACC_REDUCTION} internal function, into a sequence of gimple
6127 instructions. @var{call} is gimple statement containing the call to
6128 the function. This hook removes statement @var{call} after the
6129 expanded sequence has been inserted. This hook is also responsible
6130 for allocating any storage for reductions when necessary.
6133 @deftypefn {Target Hook} tree TARGET_PREFERRED_ELSE_VALUE (unsigned @var{ifn}, tree @var{type}, unsigned @var{nops}, tree *@var{ops})
6134 This hook returns the target's preferred final argument for a call
6135 to conditional internal function @var{ifn} (really of type
6136 @code{internal_fn}). @var{type} specifies the return type of the
6137 function and @var{ops} are the operands to the conditional operation,
6138 of which there are @var{nops}.
6140 For example, if @var{ifn} is @code{IFN_COND_ADD}, the hook returns
6141 a value of type @var{type} that should be used when @samp{@var{ops}[0]}
6142 and @samp{@var{ops}[1]} are conditionally added together.
6144 This hook is only relevant if the target supports conditional patterns
6145 like @code{cond_add@var{m}}. The default implementation returns a zero
6146 constant of type @var{type}.
6149 @node Anchored Addresses
6150 @section Anchored Addresses
6151 @cindex anchored addresses
6152 @cindex @option{-fsection-anchors}
6154 GCC usually addresses every static object as a separate entity.
6155 For example, if we have:
6159 int foo (void) @{ return a + b + c; @}
6162 the code for @code{foo} will usually calculate three separate symbolic
6163 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
6164 it would be better to calculate just one symbolic address and access
6165 the three variables relative to it. The equivalent pseudocode would
6171 register int *xr = &x;
6172 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6176 (which isn't valid C). We refer to shared addresses like @code{x} as
6177 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
6179 The hooks below describe the target properties that GCC needs to know
6180 in order to make effective use of section anchors. It won't use
6181 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
6182 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
6184 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
6185 The minimum offset that should be applied to a section anchor.
6186 On most targets, it should be the smallest offset that can be
6187 applied to a base register while still giving a legitimate address
6188 for every mode. The default value is 0.
6191 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
6192 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
6193 offset that should be applied to section anchors. The default
6197 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
6198 Write the assembly code to define section anchor @var{x}, which is a
6199 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
6200 The hook is called with the assembly output position set to the beginning
6201 of @code{SYMBOL_REF_BLOCK (@var{x})}.
6203 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
6204 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
6205 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
6206 is @code{NULL}, which disables the use of section anchors altogether.
6209 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
6210 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
6211 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
6212 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
6214 The default version is correct for most targets, but you might need to
6215 intercept this hook to handle things like target-specific attributes
6216 or target-specific sections.
6219 @node Condition Code
6220 @section Condition Code Status
6221 @cindex condition code status
6223 The macros in this section can be split in two families, according to the
6224 two ways of representing condition codes in GCC.
6226 The first representation is the so called @code{(cc0)} representation
6227 (@pxref{Jump Patterns}), where all instructions can have an implicit
6228 clobber of the condition codes. The second is the condition code
6229 register representation, which provides better schedulability for
6230 architectures that do have a condition code register, but on which
6231 most instructions do not affect it. The latter category includes
6234 The implicit clobbering poses a strong restriction on the placement of
6235 the definition and use of the condition code. In the past the definition
6236 and use were always adjacent. However, recent changes to support trapping
6237 arithmatic may result in the definition and user being in different blocks.
6238 Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally,
6239 the definition may be the source of exception handling edges.
6241 These restrictions can prevent important
6242 optimizations on some machines. For example, on the IBM RS/6000, there
6243 is a delay for taken branches unless the condition code register is set
6244 three instructions earlier than the conditional branch. The instruction
6245 scheduler cannot perform this optimization if it is not permitted to
6246 separate the definition and use of the condition code register.
6248 For this reason, it is possible and suggested to use a register to
6249 represent the condition code for new ports. If there is a specific
6250 condition code register in the machine, use a hard register. If the
6251 condition code or comparison result can be placed in any general register,
6252 or if there are multiple condition registers, use a pseudo register.
6253 Registers used to store the condition code value will usually have a mode
6254 that is in class @code{MODE_CC}.
6256 Alternatively, you can use @code{BImode} if the comparison operator is
6257 specified already in the compare instruction. In this case, you are not
6258 interested in most macros in this section.
6261 * CC0 Condition Codes:: Old style representation of condition codes.
6262 * MODE_CC Condition Codes:: Modern representation of condition codes.
6265 @node CC0 Condition Codes
6266 @subsection Representation of condition codes using @code{(cc0)}
6270 The file @file{conditions.h} defines a variable @code{cc_status} to
6271 describe how the condition code was computed (in case the interpretation of
6272 the condition code depends on the instruction that it was set by). This
6273 variable contains the RTL expressions on which the condition code is
6274 currently based, and several standard flags.
6276 Sometimes additional machine-specific flags must be defined in the machine
6277 description header file. It can also add additional machine-specific
6278 information by defining @code{CC_STATUS_MDEP}.
6280 @defmac CC_STATUS_MDEP
6281 C code for a data type which is used for declaring the @code{mdep}
6282 component of @code{cc_status}. It defaults to @code{int}.
6284 This macro is not used on machines that do not use @code{cc0}.
6287 @defmac CC_STATUS_MDEP_INIT
6288 A C expression to initialize the @code{mdep} field to ``empty''.
6289 The default definition does nothing, since most machines don't use
6290 the field anyway. If you want to use the field, you should probably
6291 define this macro to initialize it.
6293 This macro is not used on machines that do not use @code{cc0}.
6296 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
6297 A C compound statement to set the components of @code{cc_status}
6298 appropriately for an insn @var{insn} whose body is @var{exp}. It is
6299 this macro's responsibility to recognize insns that set the condition
6300 code as a byproduct of other activity as well as those that explicitly
6303 This macro is not used on machines that do not use @code{cc0}.
6305 If there are insns that do not set the condition code but do alter
6306 other machine registers, this macro must check to see whether they
6307 invalidate the expressions that the condition code is recorded as
6308 reflecting. For example, on the 68000, insns that store in address
6309 registers do not set the condition code, which means that usually
6310 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
6311 insns. But suppose that the previous insn set the condition code
6312 based on location @samp{a4@@(102)} and the current insn stores a new
6313 value in @samp{a4}. Although the condition code is not changed by
6314 this, it will no longer be true that it reflects the contents of
6315 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
6316 @code{cc_status} in this case to say that nothing is known about the
6317 condition code value.
6319 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
6320 with the results of peephole optimization: insns whose patterns are
6321 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
6322 constants which are just the operands. The RTL structure of these
6323 insns is not sufficient to indicate what the insns actually do. What
6324 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
6325 @code{CC_STATUS_INIT}.
6327 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
6328 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
6329 @samp{cc}. This avoids having detailed information about patterns in
6330 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
6333 @node MODE_CC Condition Codes
6334 @subsection Representation of condition codes using registers
6338 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
6339 On many machines, the condition code may be produced by other instructions
6340 than compares, for example the branch can use directly the condition
6341 code set by a subtract instruction. However, on some machines
6342 when the condition code is set this way some bits (such as the overflow
6343 bit) are not set in the same way as a test instruction, so that a different
6344 branch instruction must be used for some conditional branches. When
6345 this happens, use the machine mode of the condition code register to
6346 record different formats of the condition code register. Modes can
6347 also be used to record which compare instruction (e.g.@: a signed or an
6348 unsigned comparison) produced the condition codes.
6350 If other modes than @code{CCmode} are required, add them to
6351 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
6352 a mode given an operand of a compare. This is needed because the modes
6353 have to be chosen not only during RTL generation but also, for example,
6354 by instruction combination. The result of @code{SELECT_CC_MODE} should
6355 be consistent with the mode used in the patterns; for example to support
6356 the case of the add on the SPARC discussed above, we have the pattern
6362 (plus:SI (match_operand:SI 0 "register_operand" "%r")
6363 (match_operand:SI 1 "arith_operand" "rI"))
6370 together with a @code{SELECT_CC_MODE} that returns @code{CCNZmode}
6371 for comparisons whose argument is a @code{plus}:
6374 #define SELECT_CC_MODE(OP,X,Y) \
6375 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
6376 ? ((OP == LT || OP == LE || OP == GT || OP == GE) \
6377 ? CCFPEmode : CCFPmode) \
6378 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
6379 || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \
6380 ? CCNZmode : CCmode))
6383 Another reason to use modes is to retain information on which operands
6384 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
6387 You should define this macro if and only if you define extra CC modes
6388 in @file{@var{machine}-modes.def}.
6391 @deftypefn {Target Hook} void TARGET_CANONICALIZE_COMPARISON (int *@var{code}, rtx *@var{op0}, rtx *@var{op1}, bool @var{op0_preserve_value})
6392 On some machines not all possible comparisons are defined, but you can
6393 convert an invalid comparison into a valid one. For example, the Alpha
6394 does not have a @code{GT} comparison, but you can use an @code{LT}
6395 comparison instead and swap the order of the operands.
6397 On such machines, implement this hook to do any required conversions.
6398 @var{code} is the initial comparison code and @var{op0} and @var{op1}
6399 are the left and right operands of the comparison, respectively. If
6400 @var{op0_preserve_value} is @code{true} the implementation is not
6401 allowed to change the value of @var{op0} since the value might be used
6402 in RTXs which aren't comparisons. E.g. the implementation is not
6403 allowed to swap operands in that case.
6405 GCC will not assume that the comparison resulting from this macro is
6406 valid but will see if the resulting insn matches a pattern in the
6409 You need not to implement this hook if it would never change the
6410 comparison code or operands.
6413 @defmac REVERSIBLE_CC_MODE (@var{mode})
6414 A C expression whose value is one if it is always safe to reverse a
6415 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6416 can ever return @var{mode} for a floating-point inequality comparison,
6417 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6419 You need not define this macro if it would always returns zero or if the
6420 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6421 For example, here is the definition used on the SPARC, where floating-point
6422 inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}:
6425 #define REVERSIBLE_CC_MODE(MODE) \
6426 ((MODE) != CCFPEmode && (MODE) != CCFPmode)
6430 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6431 A C expression whose value is reversed condition code of the @var{code} for
6432 comparison done in CC_MODE @var{mode}. The macro is used only in case
6433 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6434 machine has some non-standard way how to reverse certain conditionals. For
6435 instance in case all floating point conditions are non-trapping, compiler may
6436 freely convert unordered compares to ordered ones. Then definition may look
6440 #define REVERSE_CONDITION(CODE, MODE) \
6441 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6442 : reverse_condition_maybe_unordered (CODE))
6446 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6447 On targets which do not use @code{(cc0)}, and which use a hard
6448 register rather than a pseudo-register to hold condition codes, the
6449 regular CSE passes are often not able to identify cases in which the
6450 hard register is set to a common value. Use this hook to enable a
6451 small pass which optimizes such cases. This hook should return true
6452 to enable this pass, and it should set the integers to which its
6453 arguments point to the hard register numbers used for condition codes.
6454 When there is only one such register, as is true on most systems, the
6455 integer pointed to by @var{p2} should be set to
6456 @code{INVALID_REGNUM}.
6458 The default version of this hook returns false.
6461 @deftypefn {Target Hook} machine_mode TARGET_CC_MODES_COMPATIBLE (machine_mode @var{m1}, machine_mode @var{m2})
6462 On targets which use multiple condition code modes in class
6463 @code{MODE_CC}, it is sometimes the case that a comparison can be
6464 validly done in more than one mode. On such a system, define this
6465 target hook to take two mode arguments and to return a mode in which
6466 both comparisons may be validly done. If there is no such mode,
6467 return @code{VOIDmode}.
6469 The default version of this hook checks whether the modes are the
6470 same. If they are, it returns that mode. If they are different, it
6471 returns @code{VOIDmode}.
6474 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
6475 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.
6479 @section Describing Relative Costs of Operations
6480 @cindex costs of instructions
6481 @cindex relative costs
6482 @cindex speed of instructions
6484 These macros let you describe the relative speed of various operations
6485 on the target machine.
6487 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6488 A C expression for the cost of moving data of mode @var{mode} from a
6489 register in class @var{from} to one in class @var{to}. The classes are
6490 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6491 value of 2 is the default; other values are interpreted relative to
6494 It is not required that the cost always equal 2 when @var{from} is the
6495 same as @var{to}; on some machines it is expensive to move between
6496 registers if they are not general registers.
6498 If reload sees an insn consisting of a single @code{set} between two
6499 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6500 classes returns a value of 2, reload does not check to ensure that the
6501 constraints of the insn are met. Setting a cost of other than 2 will
6502 allow reload to verify that the constraints are met. You should do this
6503 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6505 These macros are obsolete, new ports should use the target hook
6506 @code{TARGET_REGISTER_MOVE_COST} instead.
6509 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6510 This target hook should return the cost of moving data of mode @var{mode}
6511 from a register in class @var{from} to one in class @var{to}. The classes
6512 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6513 A value of 2 is the default; other values are interpreted relative to
6516 It is not required that the cost always equal 2 when @var{from} is the
6517 same as @var{to}; on some machines it is expensive to move between
6518 registers if they are not general registers.
6520 If reload sees an insn consisting of a single @code{set} between two
6521 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6522 classes returns a value of 2, reload does not check to ensure that the
6523 constraints of the insn are met. Setting a cost of other than 2 will
6524 allow reload to verify that the constraints are met. You should do this
6525 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6527 The default version of this function returns 2.
6530 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6531 A C expression for the cost of moving data of mode @var{mode} between a
6532 register of class @var{class} and memory; @var{in} is zero if the value
6533 is to be written to memory, nonzero if it is to be read in. This cost
6534 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6535 registers and memory is more expensive than between two registers, you
6536 should define this macro to express the relative cost.
6538 If you do not define this macro, GCC uses a default cost of 4 plus
6539 the cost of copying via a secondary reload register, if one is
6540 needed. If your machine requires a secondary reload register to copy
6541 between memory and a register of @var{class} but the reload mechanism is
6542 more complex than copying via an intermediate, define this macro to
6543 reflect the actual cost of the move.
6545 GCC defines the function @code{memory_move_secondary_cost} if
6546 secondary reloads are needed. It computes the costs due to copying via
6547 a secondary register. If your machine copies from memory using a
6548 secondary register in the conventional way but the default base value of
6549 4 is not correct for your machine, define this macro to add some other
6550 value to the result of that function. The arguments to that function
6551 are the same as to this macro.
6553 These macros are obsolete, new ports should use the target hook
6554 @code{TARGET_MEMORY_MOVE_COST} instead.
6557 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6558 This target hook should return the cost of moving data of mode @var{mode}
6559 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6560 if the value is to be written to memory, @code{true} if it is to be read in.
6561 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6562 If moving between registers and memory is more expensive than between two
6563 registers, you should add this target hook to express the relative cost.
6565 If you do not add this target hook, GCC uses a default cost of 4 plus
6566 the cost of copying via a secondary reload register, if one is
6567 needed. If your machine requires a secondary reload register to copy
6568 between memory and a register of @var{rclass} but the reload mechanism is
6569 more complex than copying via an intermediate, use this target hook to
6570 reflect the actual cost of the move.
6572 GCC defines the function @code{memory_move_secondary_cost} if
6573 secondary reloads are needed. It computes the costs due to copying via
6574 a secondary register. If your machine copies from memory using a
6575 secondary register in the conventional way but the default base value of
6576 4 is not correct for your machine, use this target hook to add some other
6577 value to the result of that function. The arguments to that function
6578 are the same as to this target hook.
6581 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6582 A C expression for the cost of a branch instruction. A value of 1 is
6583 the default; other values are interpreted relative to that. Parameter
6584 @var{speed_p} is true when the branch in question should be optimized
6585 for speed. When it is false, @code{BRANCH_COST} should return a value
6586 optimal for code size rather than performance. @var{predictable_p} is
6587 true for well-predicted branches. On many architectures the
6588 @code{BRANCH_COST} can be reduced then.
6591 Here are additional macros which do not specify precise relative costs,
6592 but only that certain actions are more expensive than GCC would
6595 @defmac SLOW_BYTE_ACCESS
6596 Define this macro as a C expression which is nonzero if accessing less
6597 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6598 faster than accessing a word of memory, i.e., if such access
6599 require more than one instruction or if there is no difference in cost
6600 between byte and (aligned) word loads.
6602 When this macro is not defined, the compiler will access a field by
6603 finding the smallest containing object; when it is defined, a fullword
6604 load will be used if alignment permits. Unless bytes accesses are
6605 faster than word accesses, using word accesses is preferable since it
6606 may eliminate subsequent memory access if subsequent accesses occur to
6607 other fields in the same word of the structure, but to different bytes.
6610 @deftypefn {Target Hook} bool TARGET_SLOW_UNALIGNED_ACCESS (machine_mode @var{mode}, unsigned int @var{align})
6611 This hook returns true if memory accesses described by the
6612 @var{mode} and @var{alignment} parameters have a cost many times greater
6613 than aligned accesses, for example if they are emulated in a trap handler.
6614 This hook is invoked only for unaligned accesses, i.e.@: when
6615 @code{@var{alignment} < GET_MODE_ALIGNMENT (@var{mode})}.
6617 When this hook returns true, the compiler will act as if
6618 @code{STRICT_ALIGNMENT} were true when generating code for block
6619 moves. This can cause significantly more instructions to be produced.
6620 Therefore, do not make this hook return true if unaligned accesses only
6621 add a cycle or two to the time for a memory access.
6623 The hook must return true whenever @code{STRICT_ALIGNMENT} is true.
6624 The default implementation returns @code{STRICT_ALIGNMENT}.
6627 @defmac MOVE_RATIO (@var{speed})
6628 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6629 which a sequence of insns should be generated instead of a
6630 string move insn or a library call. Increasing the value will always
6631 make code faster, but eventually incurs high cost in increased code size.
6633 Note that on machines where the corresponding move insn is a
6634 @code{define_expand} that emits a sequence of insns, this macro counts
6635 the number of such sequences.
6637 The parameter @var{speed} is true if the code is currently being
6638 optimized for speed rather than size.
6640 If you don't define this, a reasonable default is used.
6643 @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})
6644 GCC will attempt several strategies when asked to copy between
6645 two areas of memory, or to set, clear or store to memory, for example
6646 when copying a @code{struct}. The @code{by_pieces} infrastructure
6647 implements such memory operations as a sequence of load, store or move
6648 insns. Alternate strategies are to expand the
6649 @code{cpymem} or @code{setmem} optabs, to emit a library call, or to emit
6650 unit-by-unit, loop-based operations.
6652 This target hook should return true if, for a memory operation with a
6653 given @var{size} and @var{alignment}, using the @code{by_pieces}
6654 infrastructure is expected to result in better code generation.
6655 Both @var{size} and @var{alignment} are measured in terms of storage
6658 The parameter @var{op} is one of: @code{CLEAR_BY_PIECES},
6659 @code{MOVE_BY_PIECES}, @code{SET_BY_PIECES}, @code{STORE_BY_PIECES} or
6660 @code{COMPARE_BY_PIECES}. These describe the type of memory operation
6661 under consideration.
6663 The parameter @var{speed_p} is true if the code is currently being
6664 optimized for speed rather than size.
6666 Returning true for higher values of @var{size} can improve code generation
6667 for speed if the target does not provide an implementation of the
6668 @code{cpymem} or @code{setmem} standard names, if the @code{cpymem} or
6669 @code{setmem} implementation would be more expensive than a sequence of
6670 insns, or if the overhead of a library call would dominate that of
6671 the body of the memory operation.
6673 Returning true for higher values of @code{size} may also cause an increase
6674 in code size, for example where the number of insns emitted to perform a
6675 move would be greater than that of a library call.
6678 @deftypefn {Target Hook} int TARGET_COMPARE_BY_PIECES_BRANCH_RATIO (machine_mode @var{mode})
6679 When expanding a block comparison in MODE, gcc can try to reduce the
6680 number of branches at the expense of more memory operations. This hook
6681 allows the target to override the default choice. It should return the
6682 factor by which branches should be reduced over the plain expansion with
6683 one comparison per @var{mode}-sized piece. A port can also prevent a
6684 particular mode from being used for block comparisons by returning a
6685 negative number from this hook.
6688 @defmac MOVE_MAX_PIECES
6689 A C expression used by @code{move_by_pieces} to determine the largest unit
6690 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6693 @defmac STORE_MAX_PIECES
6694 A C expression used by @code{store_by_pieces} to determine the largest unit
6695 a store used to memory is. Defaults to @code{MOVE_MAX_PIECES}, or two times
6696 the size of @code{HOST_WIDE_INT}, whichever is smaller.
6699 @defmac COMPARE_MAX_PIECES
6700 A C expression used by @code{compare_by_pieces} to determine the largest unit
6701 a load or store used to compare memory is. Defaults to
6702 @code{MOVE_MAX_PIECES}.
6705 @defmac CLEAR_RATIO (@var{speed})
6706 The threshold of number of scalar move insns, @emph{below} which a sequence
6707 of insns should be generated to clear memory instead of a string clear insn
6708 or a library call. Increasing the value will always make code faster, but
6709 eventually incurs high cost in increased code size.
6711 The parameter @var{speed} is true if the code is currently being
6712 optimized for speed rather than size.
6714 If you don't define this, a reasonable default is used.
6717 @defmac SET_RATIO (@var{speed})
6718 The threshold of number of scalar move insns, @emph{below} which a sequence
6719 of insns should be generated to set memory to a constant value, instead of
6720 a block set insn or a library call.
6721 Increasing the value will always make code faster, but
6722 eventually incurs high cost in increased code size.
6724 The parameter @var{speed} is true if the code is currently being
6725 optimized for speed rather than size.
6727 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6730 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6731 A C expression used to determine whether a load postincrement is a good
6732 thing to use for a given mode. Defaults to the value of
6733 @code{HAVE_POST_INCREMENT}.
6736 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6737 A C expression used to determine whether a load postdecrement is a good
6738 thing to use for a given mode. Defaults to the value of
6739 @code{HAVE_POST_DECREMENT}.
6742 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6743 A C expression used to determine whether a load preincrement is a good
6744 thing to use for a given mode. Defaults to the value of
6745 @code{HAVE_PRE_INCREMENT}.
6748 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6749 A C expression used to determine whether a load predecrement is a good
6750 thing to use for a given mode. Defaults to the value of
6751 @code{HAVE_PRE_DECREMENT}.
6754 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6755 A C expression used to determine whether a store postincrement is a good
6756 thing to use for a given mode. Defaults to the value of
6757 @code{HAVE_POST_INCREMENT}.
6760 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6761 A C expression used to determine whether a store postdecrement is a good
6762 thing to use for a given mode. Defaults to the value of
6763 @code{HAVE_POST_DECREMENT}.
6766 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6767 This macro is used to determine whether a store preincrement is a good
6768 thing to use for a given mode. Defaults to the value of
6769 @code{HAVE_PRE_INCREMENT}.
6772 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6773 This macro is used to determine whether a store predecrement is a good
6774 thing to use for a given mode. Defaults to the value of
6775 @code{HAVE_PRE_DECREMENT}.
6778 @defmac NO_FUNCTION_CSE
6779 Define this macro to be true if it is as good or better to call a constant
6780 function address than to call an address kept in a register.
6783 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
6784 Define this macro if a non-short-circuit operation produced by
6785 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6786 @code{BRANCH_COST} is greater than or equal to the value 2.
6789 @deftypefn {Target Hook} bool TARGET_OPTAB_SUPPORTED_P (int @var{op}, machine_mode @var{mode1}, machine_mode @var{mode2}, optimization_type @var{opt_type})
6790 Return true if the optimizers should use optab @var{op} with
6791 modes @var{mode1} and @var{mode2} for optimization type @var{opt_type}.
6792 The optab is known to have an associated @file{.md} instruction
6793 whose C condition is true. @var{mode2} is only meaningful for conversion
6794 optabs; for direct optabs it is a copy of @var{mode1}.
6796 For example, when called with @var{op} equal to @code{rint_optab} and
6797 @var{mode1} equal to @code{DFmode}, the hook should say whether the
6798 optimizers should use optab @code{rintdf2}.
6800 The default hook returns true for all inputs.
6803 @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})
6804 This target hook describes the relative costs of RTL expressions.
6806 The cost may depend on the precise form of the expression, which is
6807 available for examination in @var{x}, and the fact that @var{x} appears
6808 as operand @var{opno} of an expression with rtx code @var{outer_code}.
6809 That is, the hook can assume that there is some rtx @var{y} such
6810 that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that
6811 either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or
6812 (b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}.
6814 @var{mode} is @var{x}'s machine mode, or for cases like @code{const_int} that
6815 do not have a mode, the mode in which @var{x} is used.
6817 In implementing this hook, you can use the construct
6818 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6821 On entry to the hook, @code{*@var{total}} contains a default estimate
6822 for the cost of the expression. The hook should modify this value as
6823 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6824 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6825 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6827 When optimizing for code size, i.e.@: when @code{speed} is
6828 false, this target hook should be used to estimate the relative
6829 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6831 The hook returns true when all subexpressions of @var{x} have been
6832 processed, and false when @code{rtx_cost} should recurse.
6835 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, machine_mode @var{mode}, addr_space_t @var{as}, bool @var{speed})
6836 This hook computes the cost of an addressing mode that contains
6837 @var{address}. If not defined, the cost is computed from
6838 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6840 For most CISC machines, the default cost is a good approximation of the
6841 true cost of the addressing mode. However, on RISC machines, all
6842 instructions normally have the same length and execution time. Hence
6843 all addresses will have equal costs.
6845 In cases where more than one form of an address is known, the form with
6846 the lowest cost will be used. If multiple forms have the same, lowest,
6847 cost, the one that is the most complex will be used.
6849 For example, suppose an address that is equal to the sum of a register
6850 and a constant is used twice in the same basic block. When this macro
6851 is not defined, the address will be computed in a register and memory
6852 references will be indirect through that register. On machines where
6853 the cost of the addressing mode containing the sum is no higher than
6854 that of a simple indirect reference, this will produce an additional
6855 instruction and possibly require an additional register. Proper
6856 specification of this macro eliminates this overhead for such machines.
6858 This hook is never called with an invalid address.
6860 On machines where an address involving more than one register is as
6861 cheap as an address computation involving only one register, defining
6862 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6863 be live over a region of code where only one would have been if
6864 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6865 should be considered in the definition of this macro. Equivalent costs
6866 should probably only be given to addresses with different numbers of
6867 registers on machines with lots of registers.
6870 @deftypefn {Target Hook} int TARGET_INSN_COST (rtx_insn *@var{insn}, bool @var{speed})
6871 This target hook describes the relative costs of RTL instructions.
6873 In implementing this hook, you can use the construct
6874 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6877 When optimizing for code size, i.e.@: when @code{speed} is
6878 false, this target hook should be used to estimate the relative
6879 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6882 @deftypefn {Target Hook} {unsigned int} TARGET_MAX_NOCE_IFCVT_SEQ_COST (edge @var{e})
6883 This hook returns a value in the same units as @code{TARGET_RTX_COSTS},
6884 giving the maximum acceptable cost for a sequence generated by the RTL
6885 if-conversion pass when conditional execution is not available.
6886 The RTL if-conversion pass attempts to convert conditional operations
6887 that would require a branch to a series of unconditional operations and
6888 @code{mov@var{mode}cc} insns. This hook returns the maximum cost of the
6889 unconditional instructions and the @code{mov@var{mode}cc} insns.
6890 RTL if-conversion is cancelled if the cost of the converted sequence
6891 is greater than the value returned by this hook.
6893 @code{e} is the edge between the basic block containing the conditional
6894 branch to the basic block which would be executed if the condition
6897 The default implementation of this hook uses the
6898 @code{max-rtl-if-conversion-[un]predictable} parameters if they are set,
6899 and uses a multiple of @code{BRANCH_COST} otherwise.
6902 @deftypefn {Target Hook} bool TARGET_NOCE_CONVERSION_PROFITABLE_P (rtx_insn *@var{seq}, struct noce_if_info *@var{if_info})
6903 This hook returns true if the instruction sequence @code{seq} is a good
6904 candidate as a replacement for the if-convertible sequence described in
6908 @deftypefn {Target Hook} bool TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P (void)
6909 This predicate controls the use of the eager delay slot filler to disallow
6910 speculatively executed instructions being placed in delay slots. Targets
6911 such as certain MIPS architectures possess both branches with and without
6912 delay slots. As the eager delay slot filler can decrease performance,
6913 disabling it is beneficial when ordinary branches are available. Use of
6914 delay slot branches filled using the basic filler is often still desirable
6915 as the delay slot can hide a pipeline bubble.
6918 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_ESTIMATED_POLY_VALUE (poly_int64 @var{val})
6919 Return an estimate of the runtime value of @var{val}, for use in
6920 things like cost calculations or profiling frequencies. The default
6921 implementation returns the lowest possible value of @var{val}.
6925 @section Adjusting the Instruction Scheduler
6927 The instruction scheduler may need a fair amount of machine-specific
6928 adjustment in order to produce good code. GCC provides several target
6929 hooks for this purpose. It is usually enough to define just a few of
6930 them: try the first ones in this list first.
6932 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6933 This hook returns the maximum number of instructions that can ever
6934 issue at the same time on the target machine. The default is one.
6935 Although the insn scheduler can define itself the possibility of issue
6936 an insn on the same cycle, the value can serve as an additional
6937 constraint to issue insns on the same simulated processor cycle (see
6938 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6939 This value must be constant over the entire compilation. If you need
6940 it to vary depending on what the instructions are, you must use
6941 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6944 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx_insn *@var{insn}, int @var{more})
6945 This hook is executed by the scheduler after it has scheduled an insn
6946 from the ready list. It should return the number of insns which can
6947 still be issued in the current cycle. The default is
6948 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6949 @code{USE}, which normally are not counted against the issue rate.
6950 You should define this hook if some insns take more machine resources
6951 than others, so that fewer insns can follow them in the same cycle.
6952 @var{file} is either a null pointer, or a stdio stream to write any
6953 debug output to. @var{verbose} is the verbose level provided by
6954 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6958 @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})
6959 This function corrects the value of @var{cost} based on the
6960 relationship between @var{insn} and @var{dep_insn} through a
6961 dependence of type dep_type, and strength @var{dw}. It should return the new
6962 value. The default is to make no adjustment to @var{cost}. This can be
6963 used for example to specify to the scheduler using the traditional pipeline
6964 description that an output- or anti-dependence does not incur the same cost
6965 as a data-dependence. If the scheduler using the automaton based pipeline
6966 description, the cost of anti-dependence is zero and the cost of
6967 output-dependence is maximum of one and the difference of latency
6968 times of the first and the second insns. If these values are not
6969 acceptable, you could use the hook to modify them too. See also
6970 @pxref{Processor pipeline description}.
6973 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx_insn *@var{insn}, int @var{priority})
6974 This hook adjusts the integer scheduling priority @var{priority} of
6975 @var{insn}. It should return the new priority. Increase the priority to
6976 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6977 later. Do not define this hook if you do not need to adjust the
6978 scheduling priorities of insns.
6981 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock})
6982 This hook is executed by the scheduler after it has scheduled the ready
6983 list, to allow the machine description to reorder it (for example to
6984 combine two small instructions together on @samp{VLIW} machines).
6985 @var{file} is either a null pointer, or a stdio stream to write any
6986 debug output to. @var{verbose} is the verbose level provided by
6987 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6988 list of instructions that are ready to be scheduled. @var{n_readyp} is
6989 a pointer to the number of elements in the ready list. The scheduler
6990 reads the ready list in reverse order, starting with
6991 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6992 is the timer tick of the scheduler. You may modify the ready list and
6993 the number of ready insns. The return value is the number of insns that
6994 can issue this cycle; normally this is just @code{issue_rate}. See also
6995 @samp{TARGET_SCHED_REORDER2}.
6998 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock})
6999 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
7000 function is called whenever the scheduler starts a new cycle. This one
7001 is called once per iteration over a cycle, immediately after
7002 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
7003 return the number of insns to be scheduled in the same cycle. Defining
7004 this hook can be useful if there are frequent situations where
7005 scheduling one insn causes other insns to become ready in the same
7006 cycle. These other insns can then be taken into account properly.
7009 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_P (void)
7010 This hook is used to check whether target platform supports macro fusion.
7013 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_PAIR_P (rtx_insn *@var{prev}, rtx_insn *@var{curr})
7014 This hook is used to check whether two insns should be macro fused for
7015 a target microarchitecture. If this hook returns true for the given insn pair
7016 (@var{prev} and @var{curr}), the scheduler will put them into a sched
7017 group, and they will not be scheduled apart. The two insns will be either
7018 two SET insns or a compare and a conditional jump and this hook should
7019 validate any dependencies needed to fuse the two insns together.
7022 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx_insn *@var{head}, rtx_insn *@var{tail})
7023 This hook is called after evaluation forward dependencies of insns in
7024 chain given by two parameter values (@var{head} and @var{tail}
7025 correspondingly) but before insns scheduling of the insn chain. For
7026 example, it can be used for better insn classification if it requires
7027 analysis of dependencies. This hook can use backward and forward
7028 dependencies of the insn scheduler because they are already
7032 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
7033 This hook is executed by the scheduler at the beginning of each block of
7034 instructions that are to be scheduled. @var{file} is either a null
7035 pointer, or a stdio stream to write any debug output to. @var{verbose}
7036 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
7037 @var{max_ready} is the maximum number of insns in the current scheduling
7038 region that can be live at the same time. This can be used to allocate
7039 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
7042 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
7043 This hook is executed by the scheduler at the end of each block of
7044 instructions that are to be scheduled. It can be used to perform
7045 cleanup of any actions done by the other scheduling hooks. @var{file}
7046 is either a null pointer, or a stdio stream to write any debug output
7047 to. @var{verbose} is the verbose level provided by
7048 @option{-fsched-verbose-@var{n}}.
7051 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
7052 This hook is executed by the scheduler after function level initializations.
7053 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
7054 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
7055 @var{old_max_uid} is the maximum insn uid when scheduling begins.
7058 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
7059 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
7060 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
7061 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
7064 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
7065 The hook returns an RTL insn. The automaton state used in the
7066 pipeline hazard recognizer is changed as if the insn were scheduled
7067 when the new simulated processor cycle starts. Usage of the hook may
7068 simplify the automaton pipeline description for some @acronym{VLIW}
7069 processors. If the hook is defined, it is used only for the automaton
7070 based pipeline description. The default is not to change the state
7071 when the new simulated processor cycle starts.
7074 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
7075 The hook can be used to initialize data used by the previous hook.
7078 @deftypefn {Target Hook} {rtx_insn *} TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
7079 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
7080 to changed the state as if the insn were scheduled when the new
7081 simulated processor cycle finishes.
7084 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
7085 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
7086 used to initialize data used by the previous hook.
7089 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
7090 The hook to notify target that the current simulated cycle is about to finish.
7091 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
7092 to change the state in more complicated situations - e.g., when advancing
7093 state on a single insn is not enough.
7096 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
7097 The hook to notify target that new simulated cycle has just started.
7098 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
7099 to change the state in more complicated situations - e.g., when advancing
7100 state on a single insn is not enough.
7103 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
7104 This hook controls better choosing an insn from the ready insn queue
7105 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
7106 chooses the first insn from the queue. If the hook returns a positive
7107 value, an additional scheduler code tries all permutations of
7108 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
7109 subsequent ready insns to choose an insn whose issue will result in
7110 maximal number of issued insns on the same cycle. For the
7111 @acronym{VLIW} processor, the code could actually solve the problem of
7112 packing simple insns into the @acronym{VLIW} insn. Of course, if the
7113 rules of @acronym{VLIW} packing are described in the automaton.
7115 This code also could be used for superscalar @acronym{RISC}
7116 processors. Let us consider a superscalar @acronym{RISC} processor
7117 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
7118 @var{B}, some insns can be executed only in pipelines @var{B} or
7119 @var{C}, and one insn can be executed in pipeline @var{B}. The
7120 processor may issue the 1st insn into @var{A} and the 2nd one into
7121 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
7122 until the next cycle. If the scheduler issues the 3rd insn the first,
7123 the processor could issue all 3 insns per cycle.
7125 Actually this code demonstrates advantages of the automaton based
7126 pipeline hazard recognizer. We try quickly and easy many insn
7127 schedules to choose the best one.
7129 The default is no multipass scheduling.
7132 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx_insn *@var{insn}, int @var{ready_index})
7134 This hook controls what insns from the ready insn queue will be
7135 considered for the multipass insn scheduling. If the hook returns
7136 zero for @var{insn}, the insn will be considered in multipass scheduling.
7137 Positive return values will remove @var{insn} from consideration on
7138 the current round of multipass scheduling.
7139 Negative return values will remove @var{insn} from consideration for given
7141 Backends should be careful about returning non-zero for highest priority
7142 instruction at position 0 in the ready list. @var{ready_index} is passed
7143 to allow backends make correct judgements.
7145 The default is that any ready insns can be chosen to be issued.
7148 @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})
7149 This hook prepares the target backend for a new round of multipass
7153 @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})
7154 This hook is called when multipass scheduling evaluates instruction INSN.
7157 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, signed char *@var{ready_try}, int @var{n_ready})
7158 This is called when multipass scheduling backtracks from evaluation of
7162 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
7163 This hook notifies the target about the result of the concluded current
7164 round of multipass scheduling.
7167 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
7168 This hook initializes target-specific data used in multipass scheduling.
7171 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
7172 This hook finalizes target-specific data used in multipass scheduling.
7175 @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})
7176 This hook is called by the insn scheduler before issuing @var{insn}
7177 on cycle @var{clock}. If the hook returns nonzero,
7178 @var{insn} is not issued on this processor cycle. Instead,
7179 the processor cycle is advanced. If *@var{sort_p}
7180 is zero, the insn ready queue is not sorted on the new cycle
7181 start as usually. @var{dump} and @var{verbose} specify the file and
7182 verbosity level to use for debugging output.
7183 @var{last_clock} and @var{clock} are, respectively, the
7184 processor cycle on which the previous insn has been issued,
7185 and the current processor cycle.
7188 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
7189 This hook is used to define which dependences are considered costly by
7190 the target, so costly that it is not advisable to schedule the insns that
7191 are involved in the dependence too close to one another. The parameters
7192 to this hook are as follows: The first parameter @var{_dep} is the dependence
7193 being evaluated. The second parameter @var{cost} is the cost of the
7194 dependence as estimated by the scheduler, and the third
7195 parameter @var{distance} is the distance in cycles between the two insns.
7196 The hook returns @code{true} if considering the distance between the two
7197 insns the dependence between them is considered costly by the target,
7198 and @code{false} otherwise.
7200 Defining this hook can be useful in multiple-issue out-of-order machines,
7201 where (a) it's practically hopeless to predict the actual data/resource
7202 delays, however: (b) there's a better chance to predict the actual grouping
7203 that will be formed, and (c) correctly emulating the grouping can be very
7204 important. In such targets one may want to allow issuing dependent insns
7205 closer to one another---i.e., closer than the dependence distance; however,
7206 not in cases of ``costly dependences'', which this hooks allows to define.
7209 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
7210 This hook is called by the insn scheduler after emitting a new instruction to
7211 the instruction stream. The hook notifies a target backend to extend its
7212 per instruction data structures.
7215 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
7216 Return a pointer to a store large enough to hold target scheduling context.
7219 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
7220 Initialize store pointed to by @var{tc} to hold target scheduling context.
7221 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
7222 beginning of the block. Otherwise, copy the current context into @var{tc}.
7225 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
7226 Copy target scheduling context pointed to by @var{tc} to the current context.
7229 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
7230 Deallocate internal data in target scheduling context pointed to by @var{tc}.
7233 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
7234 Deallocate a store for target scheduling context pointed to by @var{tc}.
7237 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx_insn *@var{insn}, unsigned int @var{dep_status}, rtx *@var{new_pat})
7238 This hook is called by the insn scheduler when @var{insn} has only
7239 speculative dependencies and therefore can be scheduled speculatively.
7240 The hook is used to check if the pattern of @var{insn} has a speculative
7241 version and, in case of successful check, to generate that speculative
7242 pattern. The hook should return 1, if the instruction has a speculative form,
7243 or @minus{}1, if it doesn't. @var{request} describes the type of requested
7244 speculation. If the return value equals 1 then @var{new_pat} is assigned
7245 the generated speculative pattern.
7248 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (unsigned int @var{dep_status})
7249 This hook is called by the insn scheduler during generation of recovery code
7250 for @var{insn}. It should return @code{true}, if the corresponding check
7251 instruction should branch to recovery code, or @code{false} otherwise.
7254 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx_insn *@var{insn}, rtx_insn *@var{label}, unsigned int @var{ds})
7255 This hook is called by the insn scheduler to generate a pattern for recovery
7256 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
7257 speculative instruction for which the check should be generated.
7258 @var{label} is either a label of a basic block, where recovery code should
7259 be emitted, or a null pointer, when requested check doesn't branch to
7260 recovery code (a simple check). If @var{mutate_p} is nonzero, then
7261 a pattern for a branchy check corresponding to a simple check denoted by
7262 @var{insn} should be generated. In this case @var{label} can't be null.
7265 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
7266 This hook is used by the insn scheduler to find out what features should be
7268 The structure *@var{spec_info} should be filled in by the target.
7269 The structure describes speculation types that can be used in the scheduler.
7272 @deftypefn {Target Hook} bool TARGET_SCHED_CAN_SPECULATE_INSN (rtx_insn *@var{insn})
7273 Some instructions should never be speculated by the schedulers, usually
7274 because the instruction is too expensive to get this wrong. Often such
7275 instructions have long latency, and often they are not fully modeled in the
7276 pipeline descriptions. This hook should return @code{false} if @var{insn}
7277 should not be speculated.
7280 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
7281 This hook is called by the swing modulo scheduler to calculate a
7282 resource-based lower bound which is based on the resources available in
7283 the machine and the resources required by each instruction. The target
7284 backend can use @var{g} to calculate such bound. A very simple lower
7285 bound will be used in case this hook is not implemented: the total number
7286 of instructions divided by the issue rate.
7289 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx_insn *@var{insn}, int @var{x})
7290 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
7291 is supported in hardware and the condition specified in the parameter is true.
7294 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx_insn *@var{insn}, int @var{x})
7295 This hook is called by Haifa Scheduler. It performs the operation specified
7296 in its second parameter.
7299 @deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE
7300 True if the processor has an exposed pipeline, which means that not just
7301 the order of instructions is important for correctness when scheduling, but
7302 also the latencies of operations.
7305 @deftypefn {Target Hook} int TARGET_SCHED_REASSOCIATION_WIDTH (unsigned int @var{opc}, machine_mode @var{mode})
7306 This hook is called by tree reassociator to determine a level of
7307 parallelism required in output calculations chain.
7310 @deftypefn {Target Hook} void TARGET_SCHED_FUSION_PRIORITY (rtx_insn *@var{insn}, int @var{max_pri}, int *@var{fusion_pri}, int *@var{pri})
7311 This hook is called by scheduling fusion pass. It calculates fusion
7312 priorities for each instruction passed in by parameter. The priorities
7313 are returned via pointer parameters.
7315 @var{insn} is the instruction whose priorities need to be calculated.
7316 @var{max_pri} is the maximum priority can be returned in any cases.
7317 @var{fusion_pri} is the pointer parameter through which @var{insn}'s
7318 fusion priority should be calculated and returned.
7319 @var{pri} is the pointer parameter through which @var{insn}'s priority
7320 should be calculated and returned.
7322 Same @var{fusion_pri} should be returned for instructions which should
7323 be scheduled together. Different @var{pri} should be returned for
7324 instructions with same @var{fusion_pri}. @var{fusion_pri} is the major
7325 sort key, @var{pri} is the minor sort key. All instructions will be
7326 scheduled according to the two priorities. All priorities calculated
7327 should be between 0 (exclusive) and @var{max_pri} (inclusive). To avoid
7328 false dependencies, @var{fusion_pri} of instructions which need to be
7329 scheduled together should be smaller than @var{fusion_pri} of irrelevant
7332 Given below example:
7345 On targets like ARM/AArch64, the two pairs of consecutive loads should be
7346 merged. Since peephole2 pass can't help in this case unless consecutive
7347 loads are actually next to each other in instruction flow. That's where
7348 this scheduling fusion pass works. This hook calculates priority for each
7349 instruction based on its fustion type, like:
7352 ldr r10, [r1, 4] ; fusion_pri=99, pri=96
7353 add r4, r4, r10 ; fusion_pri=100, pri=100
7354 ldr r15, [r2, 8] ; fusion_pri=98, pri=92
7355 sub r5, r5, r15 ; fusion_pri=100, pri=100
7356 ldr r11, [r1, 0] ; fusion_pri=99, pri=100
7357 add r4, r4, r11 ; fusion_pri=100, pri=100
7358 ldr r16, [r2, 12] ; fusion_pri=98, pri=88
7359 sub r5, r5, r16 ; fusion_pri=100, pri=100
7362 Scheduling fusion pass then sorts all ready to issue instructions according
7363 to the priorities. As a result, instructions of same fusion type will be
7364 pushed together in instruction flow, like:
7377 Now peephole2 pass can simply merge the two pairs of loads.
7379 Since scheduling fusion pass relies on peephole2 to do real fusion
7380 work, it is only enabled by default when peephole2 is in effect.
7382 This is firstly introduced on ARM/AArch64 targets, please refer to
7383 the hook implementation for how different fusion types are supported.
7386 @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})
7387 Define this hook for enabling divmod transform if the port does not have
7388 hardware divmod insn but defines target-specific divmod libfuncs.
7392 @section Dividing the Output into Sections (Texts, Data, @dots{})
7393 @c the above section title is WAY too long. maybe cut the part between
7394 @c the (...)? --mew 10feb93
7396 An object file is divided into sections containing different types of
7397 data. In the most common case, there are three sections: the @dfn{text
7398 section}, which holds instructions and read-only data; the @dfn{data
7399 section}, which holds initialized writable data; and the @dfn{bss
7400 section}, which holds uninitialized data. Some systems have other kinds
7403 @file{varasm.c} provides several well-known sections, such as
7404 @code{text_section}, @code{data_section} and @code{bss_section}.
7405 The normal way of controlling a @code{@var{foo}_section} variable
7406 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
7407 as described below. The macros are only read once, when @file{varasm.c}
7408 initializes itself, so their values must be run-time constants.
7409 They may however depend on command-line flags.
7411 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
7412 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
7413 to be string literals.
7415 Some assemblers require a different string to be written every time a
7416 section is selected. If your assembler falls into this category, you
7417 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
7418 @code{get_unnamed_section} to set up the sections.
7420 You must always create a @code{text_section}, either by defining
7421 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
7422 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
7423 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
7424 create a distinct @code{readonly_data_section}, the default is to
7425 reuse @code{text_section}.
7427 All the other @file{varasm.c} sections are optional, and are null
7428 if the target does not provide them.
7430 @defmac TEXT_SECTION_ASM_OP
7431 A C expression whose value is a string, including spacing, containing the
7432 assembler operation that should precede instructions and read-only data.
7433 Normally @code{"\t.text"} is right.
7436 @defmac HOT_TEXT_SECTION_NAME
7437 If defined, a C string constant for the name of the section containing most
7438 frequently executed functions of the program. If not defined, GCC will provide
7439 a default definition if the target supports named sections.
7442 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
7443 If defined, a C string constant for the name of the section containing unlikely
7444 executed functions in the program.
7447 @defmac DATA_SECTION_ASM_OP
7448 A C expression whose value is a string, including spacing, containing the
7449 assembler operation to identify the following data as writable initialized
7450 data. Normally @code{"\t.data"} is right.
7453 @defmac SDATA_SECTION_ASM_OP
7454 If defined, a C expression whose value is a string, including spacing,
7455 containing the assembler operation to identify the following data as
7456 initialized, writable small data.
7459 @defmac READONLY_DATA_SECTION_ASM_OP
7460 A C expression whose value is a string, including spacing, containing the
7461 assembler operation to identify the following data as read-only initialized
7465 @defmac BSS_SECTION_ASM_OP
7466 If defined, a C expression whose value is a string, including spacing,
7467 containing the assembler operation to identify the following data as
7468 uninitialized global data. If not defined, and
7469 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
7470 uninitialized global data will be output in the data section if
7471 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
7475 @defmac SBSS_SECTION_ASM_OP
7476 If defined, a C expression whose value is a string, including spacing,
7477 containing the assembler operation to identify the following data as
7478 uninitialized, writable small data.
7481 @defmac TLS_COMMON_ASM_OP
7482 If defined, a C expression whose value is a string containing the
7483 assembler operation to identify the following data as thread-local
7484 common data. The default is @code{".tls_common"}.
7487 @defmac TLS_SECTION_ASM_FLAG
7488 If defined, a C expression whose value is a character constant
7489 containing the flag used to mark a section as a TLS section. The
7490 default is @code{'T'}.
7493 @defmac INIT_SECTION_ASM_OP
7494 If defined, a C expression whose value is a string, including spacing,
7495 containing the assembler operation to identify the following data as
7496 initialization code. If not defined, GCC will assume such a section does
7497 not exist. This section has no corresponding @code{init_section}
7498 variable; it is used entirely in runtime code.
7501 @defmac FINI_SECTION_ASM_OP
7502 If defined, a C expression whose value is a string, including spacing,
7503 containing the assembler operation to identify the following data as
7504 finalization code. If not defined, GCC will assume such a section does
7505 not exist. This section has no corresponding @code{fini_section}
7506 variable; it is used entirely in runtime code.
7509 @defmac INIT_ARRAY_SECTION_ASM_OP
7510 If defined, a C expression whose value is a string, including spacing,
7511 containing the assembler operation to identify the following data as
7512 part of the @code{.init_array} (or equivalent) section. If not
7513 defined, GCC will assume such a section does not exist. Do not define
7514 both this macro and @code{INIT_SECTION_ASM_OP}.
7517 @defmac FINI_ARRAY_SECTION_ASM_OP
7518 If defined, a C expression whose value is a string, including spacing,
7519 containing the assembler operation to identify the following data as
7520 part of the @code{.fini_array} (or equivalent) section. If not
7521 defined, GCC will assume such a section does not exist. Do not define
7522 both this macro and @code{FINI_SECTION_ASM_OP}.
7525 @defmac MACH_DEP_SECTION_ASM_FLAG
7526 If defined, a C expression whose value is a character constant
7527 containing the flag used to mark a machine-dependent section. This
7528 corresponds to the @code{SECTION_MACH_DEP} section flag.
7531 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
7532 If defined, an ASM statement that switches to a different section
7533 via @var{section_op}, calls @var{function}, and switches back to
7534 the text section. This is used in @file{crtstuff.c} if
7535 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
7536 to initialization and finalization functions from the init and fini
7537 sections. By default, this macro uses a simple function call. Some
7538 ports need hand-crafted assembly code to avoid dependencies on
7539 registers initialized in the function prologue or to ensure that
7540 constant pools don't end up too far way in the text section.
7543 @defmac TARGET_LIBGCC_SDATA_SECTION
7544 If defined, a string which names the section into which small
7545 variables defined in crtstuff and libgcc should go. This is useful
7546 when the target has options for optimizing access to small data, and
7547 you want the crtstuff and libgcc routines to be conservative in what
7548 they expect of your application yet liberal in what your application
7549 expects. For example, for targets with a @code{.sdata} section (like
7550 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
7551 require small data support from your application, but use this macro
7552 to put small data into @code{.sdata} so that your application can
7553 access these variables whether it uses small data or not.
7556 @defmac FORCE_CODE_SECTION_ALIGN
7557 If defined, an ASM statement that aligns a code section to some
7558 arbitrary boundary. This is used to force all fragments of the
7559 @code{.init} and @code{.fini} sections to have to same alignment
7560 and thus prevent the linker from having to add any padding.
7563 @defmac JUMP_TABLES_IN_TEXT_SECTION
7564 Define this macro to be an expression with a nonzero value if jump
7565 tables (for @code{tablejump} insns) should be output in the text
7566 section, along with the assembler instructions. Otherwise, the
7567 readonly data section is used.
7569 This macro is irrelevant if there is no separate readonly data section.
7572 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
7573 Define this hook if you need to do something special to set up the
7574 @file{varasm.c} sections, or if your target has some special sections
7575 of its own that you need to create.
7577 GCC calls this hook after processing the command line, but before writing
7578 any assembly code, and before calling any of the section-returning hooks
7582 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
7583 Return a mask describing how relocations should be treated when
7584 selecting sections. Bit 1 should be set if global relocations
7585 should be placed in a read-write section; bit 0 should be set if
7586 local relocations should be placed in a read-write section.
7588 The default version of this function returns 3 when @option{-fpic}
7589 is in effect, and 0 otherwise. The hook is typically redefined
7590 when the target cannot support (some kinds of) dynamic relocations
7591 in read-only sections even in executables.
7594 @deftypefn {Target Hook} bool TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC (void)
7595 Return true to generate ADDR_DIF_VEC table
7596 or false to generate ADDR_VEC table for jumps in case of -fPIC.
7598 The default version of this function returns true if flag_pic
7599 equals true and false otherwise
7602 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
7603 Return the section into which @var{exp} should be placed. You can
7604 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
7605 some sort. @var{reloc} indicates whether the initial value of @var{exp}
7606 requires link-time relocations. Bit 0 is set when variable contains
7607 local relocations only, while bit 1 is set for global relocations.
7608 @var{align} is the constant alignment in bits.
7610 The default version of this function takes care of putting read-only
7611 variables in @code{readonly_data_section}.
7613 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7616 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7617 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7618 for @code{FUNCTION_DECL}s as well as for variables and constants.
7620 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7621 function has been determined to be likely to be called, and nonzero if
7622 it is unlikely to be called.
7625 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7626 Build up a unique section name, expressed as a @code{STRING_CST} node,
7627 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7628 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7629 the initial value of @var{exp} requires link-time relocations.
7631 The default version of this function appends the symbol name to the
7632 ELF section name that would normally be used for the symbol. For
7633 example, the function @code{foo} would be placed in @code{.text.foo}.
7634 Whatever the actual target object format, this is often good enough.
7637 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7638 Return the readonly data section associated with
7639 @samp{DECL_SECTION_NAME (@var{decl})}.
7640 The default version of this function selects @code{.gnu.linkonce.r.name} if
7641 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7642 if function is in @code{.text.name}, and the normal readonly-data section
7646 @deftypevr {Target Hook} {const char *} TARGET_ASM_MERGEABLE_RODATA_PREFIX
7647 Usually, the compiler uses the prefix @code{".rodata"} to construct
7648 section names for mergeable constant data. Define this macro to override
7649 the string if a different section name should be used.
7652 @deftypefn {Target Hook} {section *} TARGET_ASM_TM_CLONE_TABLE_SECTION (void)
7653 Return the section that should be used for transactional memory clone tables.
7656 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7657 Return the section into which a constant @var{x}, of mode @var{mode},
7658 should be placed. You can assume that @var{x} is some kind of
7659 constant in RTL@. The argument @var{mode} is redundant except in the
7660 case of a @code{const_int} rtx. @var{align} is the constant alignment
7663 The default version of this function takes care of putting symbolic
7664 constants in @code{flag_pic} mode in @code{data_section} and everything
7665 else in @code{readonly_data_section}.
7668 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7669 Define this hook if you need to postprocess the assembler name generated
7670 by target-independent code. The @var{id} provided to this hook will be
7671 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7672 or the mangled name of the @var{decl} in C++). The return value of the
7673 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7674 your target system. The default implementation of this hook just
7675 returns the @var{id} provided.
7678 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7679 Define this hook if references to a symbol or a constant must be
7680 treated differently depending on something about the variable or
7681 function named by the symbol (such as what section it is in).
7683 The hook is executed immediately after rtl has been created for
7684 @var{decl}, which may be a variable or function declaration or
7685 an entry in the constant pool. In either case, @var{rtl} is the
7686 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7687 in this hook; that field may not have been initialized yet.
7689 In the case of a constant, it is safe to assume that the rtl is
7690 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7691 will also have this form, but that is not guaranteed. Global
7692 register variables, for instance, will have a @code{reg} for their
7693 rtl. (Normally the right thing to do with such unusual rtl is
7696 The @var{new_decl_p} argument will be true if this is the first time
7697 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7698 be false for subsequent invocations, which will happen for duplicate
7699 declarations. Whether or not anything must be done for the duplicate
7700 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7701 @var{new_decl_p} is always true when the hook is called for a constant.
7703 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7704 The usual thing for this hook to do is to record flags in the
7705 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7706 Historically, the name string was modified if it was necessary to
7707 encode more than one bit of information, but this practice is now
7708 discouraged; use @code{SYMBOL_REF_FLAGS}.
7710 The default definition of this hook, @code{default_encode_section_info}
7711 in @file{varasm.c}, sets a number of commonly-useful bits in
7712 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7713 before overriding it.
7716 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7717 Decode @var{name} and return the real name part, sans
7718 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7722 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7723 Returns true if @var{exp} should be placed into a ``small data'' section.
7724 The default version of this hook always returns false.
7727 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7728 Contains the value true if the target places read-only
7729 ``small data'' into a separate section. The default value is false.
7732 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7733 It returns true if target wants profile code emitted before prologue.
7735 The default version of this hook use the target macro
7736 @code{PROFILE_BEFORE_PROLOGUE}.
7739 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7740 Returns true if @var{exp} names an object for which name resolution
7741 rules must resolve to the current ``module'' (dynamic shared library
7742 or executable image).
7744 The default version of this hook implements the name resolution rules
7745 for ELF, which has a looser model of global name binding than other
7746 currently supported object file formats.
7749 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7750 Contains the value true if the target supports thread-local storage.
7751 The default value is false.
7756 @section Position Independent Code
7757 @cindex position independent code
7760 This section describes macros that help implement generation of position
7761 independent code. Simply defining these macros is not enough to
7762 generate valid PIC; you must also add support to the hook
7763 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7764 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7765 must modify the definition of @samp{movsi} to do something appropriate
7766 when the source operand contains a symbolic address. You may also
7767 need to alter the handling of switch statements so that they use
7769 @c i rearranged the order of the macros above to try to force one of
7770 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7772 @defmac PIC_OFFSET_TABLE_REGNUM
7773 The register number of the register used to address a table of static
7774 data addresses in memory. In some cases this register is defined by a
7775 processor's ``application binary interface'' (ABI)@. When this macro
7776 is defined, RTL is generated for this register once, as with the stack
7777 pointer and frame pointer registers. If this macro is not defined, it
7778 is up to the machine-dependent files to allocate such a register (if
7779 necessary). Note that this register must be fixed when in use (e.g.@:
7780 when @code{flag_pic} is true).
7783 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7784 A C expression that is nonzero if the register defined by
7785 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7786 the default is zero. Do not define
7787 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7790 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7791 A C expression that is nonzero if @var{x} is a legitimate immediate
7792 operand on the target machine when generating position independent code.
7793 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7794 check this. You can also assume @var{flag_pic} is true, so you need not
7795 check it either. You need not define this macro if all constants
7796 (including @code{SYMBOL_REF}) can be immediate operands when generating
7797 position independent code.
7800 @node Assembler Format
7801 @section Defining the Output Assembler Language
7803 This section describes macros whose principal purpose is to describe how
7804 to write instructions in assembler language---rather than what the
7808 * File Framework:: Structural information for the assembler file.
7809 * Data Output:: Output of constants (numbers, strings, addresses).
7810 * Uninitialized Data:: Output of uninitialized variables.
7811 * Label Output:: Output and generation of labels.
7812 * Initialization:: General principles of initialization
7813 and termination routines.
7814 * Macros for Initialization::
7815 Specific macros that control the handling of
7816 initialization and termination routines.
7817 * Instruction Output:: Output of actual instructions.
7818 * Dispatch Tables:: Output of jump tables.
7819 * Exception Region Output:: Output of exception region code.
7820 * Alignment Output:: Pseudo ops for alignment and skipping data.
7823 @node File Framework
7824 @subsection The Overall Framework of an Assembler File
7825 @cindex assembler format
7826 @cindex output of assembler code
7828 @c prevent bad page break with this line
7829 This describes the overall framework of an assembly file.
7831 @findex default_file_start
7832 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7833 Output to @code{asm_out_file} any text which the assembler expects to
7834 find at the beginning of a file. The default behavior is controlled
7835 by two flags, documented below. Unless your target's assembler is
7836 quite unusual, if you override the default, you should call
7837 @code{default_file_start} at some point in your target hook. This
7838 lets other target files rely on these variables.
7841 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7842 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7843 printed as the very first line in the assembly file, unless
7844 @option{-fverbose-asm} is in effect. (If that macro has been defined
7845 to the empty string, this variable has no effect.) With the normal
7846 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7847 assembler that it need not bother stripping comments or extra
7848 whitespace from its input. This allows it to work a bit faster.
7850 The default is false. You should not set it to true unless you have
7851 verified that your port does not generate any extra whitespace or
7852 comments that will cause GAS to issue errors in NO_APP mode.
7855 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7856 If this flag is true, @code{output_file_directive} will be called
7857 for the primary source file, immediately after printing
7858 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7859 this to be done. The default is false.
7862 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7863 Output to @code{asm_out_file} any text which the assembler expects
7864 to find at the end of a file. The default is to output nothing.
7867 @deftypefun void file_end_indicate_exec_stack ()
7868 Some systems use a common convention, the @samp{.note.GNU-stack}
7869 special section, to indicate whether or not an object file relies on
7870 the stack being executable. If your system uses this convention, you
7871 should define @code{TARGET_ASM_FILE_END} to this function. If you
7872 need to do other things in that hook, have your hook function call
7876 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7877 Output to @code{asm_out_file} any text which the assembler expects
7878 to find at the start of an LTO section. The default is to output
7882 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7883 Output to @code{asm_out_file} any text which the assembler expects
7884 to find at the end of an LTO section. The default is to output
7888 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7889 Output to @code{asm_out_file} any text which is needed before emitting
7890 unwind info and debug info at the end of a file. Some targets emit
7891 here PIC setup thunks that cannot be emitted at the end of file,
7892 because they couldn't have unwind info then. The default is to output
7896 @defmac ASM_COMMENT_START
7897 A C string constant describing how to begin a comment in the target
7898 assembler language. The compiler assumes that the comment will end at
7899 the end of the line.
7903 A C string constant for text to be output before each @code{asm}
7904 statement or group of consecutive ones. Normally this is
7905 @code{"#APP"}, which is a comment that has no effect on most
7906 assemblers but tells the GNU assembler that it must check the lines
7907 that follow for all valid assembler constructs.
7911 A C string constant for text to be output after each @code{asm}
7912 statement or group of consecutive ones. Normally this is
7913 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7914 time-saving assumptions that are valid for ordinary compiler output.
7917 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7918 A C statement to output COFF information or DWARF debugging information
7919 which indicates that filename @var{name} is the current source file to
7920 the stdio stream @var{stream}.
7922 This macro need not be defined if the standard form of output
7923 for the file format in use is appropriate.
7926 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7927 Output DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7929 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7932 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_IDENT (const char *@var{name})
7933 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.
7936 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7937 A C statement to output the string @var{string} to the stdio stream
7938 @var{stream}. If you do not call the function @code{output_quoted_string}
7939 in your config files, GCC will only call it to output filenames to
7940 the assembler source. So you can use it to canonicalize the format
7941 of the filename using this macro.
7944 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7945 Output assembly directives to switch to section @var{name}. The section
7946 should have attributes as specified by @var{flags}, which is a bit mask
7947 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7948 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7949 this section is associated.
7952 @deftypefn {Target Hook} bool TARGET_ASM_ELF_FLAGS_NUMERIC (unsigned int @var{flags}, unsigned int *@var{num})
7953 This hook can be used to encode ELF section flags for which no letter
7954 code has been defined in the assembler. It is called by
7955 @code{default_asm_named_section} whenever the section flags need to be
7956 emitted in the assembler output. If the hook returns true, then the
7957 numerical value for ELF section flags should be calculated from
7958 @var{flags} and saved in @var{*num}; the value is printed out instead of the
7959 normal sequence of letter codes. If the hook is not defined, or if it
7960 returns false, then @var{num} is ignored and the traditional letter sequence
7964 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
7965 Return preferred text (sub)section for function @var{decl}.
7966 Main purpose of this function is to separate cold, normal and hot
7967 functions. @var{startup} is true when function is known to be used only
7968 at startup (from static constructors or it is @code{main()}).
7969 @var{exit} is true when function is known to be used only at exit
7970 (from static destructors).
7971 Return NULL if function should go to default text section.
7974 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
7975 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}.
7978 @deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7979 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7980 It must not be modified by command-line option processing.
7983 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7984 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7985 This flag is true if we can create zeroed data by switching to a BSS
7986 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7987 This is true on most ELF targets.
7990 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7991 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7992 based on a variable or function decl, a section name, and whether or not the
7993 declaration's initializer may contain runtime relocations. @var{decl} may be
7994 null, in which case read-write data should be assumed.
7996 The default version of this function handles choosing code vs data,
7997 read-only vs read-write data, and @code{flag_pic}. You should only
7998 need to override this if your target has special flags that might be
7999 set via @code{__attribute__}.
8002 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
8003 Provides the target with the ability to record the gcc command line
8004 switches that have been passed to the compiler, and options that are
8005 enabled. The @var{type} argument specifies what is being recorded.
8006 It can take the following values:
8009 @item SWITCH_TYPE_PASSED
8010 @var{text} is a command line switch that has been set by the user.
8012 @item SWITCH_TYPE_ENABLED
8013 @var{text} is an option which has been enabled. This might be as a
8014 direct result of a command line switch, or because it is enabled by
8015 default or because it has been enabled as a side effect of a different
8016 command line switch. For example, the @option{-O2} switch enables
8017 various different individual optimization passes.
8019 @item SWITCH_TYPE_DESCRIPTIVE
8020 @var{text} is either NULL or some descriptive text which should be
8021 ignored. If @var{text} is NULL then it is being used to warn the
8022 target hook that either recording is starting or ending. The first
8023 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
8024 warning is for start up and the second time the warning is for
8025 wind down. This feature is to allow the target hook to make any
8026 necessary preparations before it starts to record switches and to
8027 perform any necessary tidying up after it has finished recording
8030 @item SWITCH_TYPE_LINE_START
8031 This option can be ignored by this target hook.
8033 @item SWITCH_TYPE_LINE_END
8034 This option can be ignored by this target hook.
8037 The hook's return value must be zero. Other return values may be
8038 supported in the future.
8040 By default this hook is set to NULL, but an example implementation is
8041 provided for ELF based targets. Called @var{elf_record_gcc_switches},
8042 it records the switches as ASCII text inside a new, string mergeable
8043 section in the assembler output file. The name of the new section is
8044 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
8048 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
8049 This is the name of the section that will be created by the example
8050 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
8056 @subsection Output of Data
8059 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
8060 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
8061 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_PSI_OP
8062 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
8063 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_PDI_OP
8064 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
8065 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_PTI_OP
8066 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
8067 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
8068 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_PSI_OP
8069 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
8070 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_PDI_OP
8071 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
8072 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_PTI_OP
8073 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
8074 These hooks specify assembly directives for creating certain kinds
8075 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
8076 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
8077 aligned two-byte object, and so on. Any of the hooks may be
8078 @code{NULL}, indicating that no suitable directive is available.
8080 The compiler will print these strings at the start of a new line,
8081 followed immediately by the object's initial value. In most cases,
8082 the string should contain a tab, a pseudo-op, and then another tab.
8085 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
8086 The @code{assemble_integer} function uses this hook to output an
8087 integer object. @var{x} is the object's value, @var{size} is its size
8088 in bytes and @var{aligned_p} indicates whether it is aligned. The
8089 function should return @code{true} if it was able to output the
8090 object. If it returns false, @code{assemble_integer} will try to
8091 split the object into smaller parts.
8093 The default implementation of this hook will use the
8094 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
8095 when the relevant string is @code{NULL}.
8098 @deftypefn {Target Hook} void TARGET_ASM_DECL_END (void)
8099 Define this hook if the target assembler requires a special marker to
8100 terminate an initialized variable declaration.
8103 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
8104 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
8105 can't deal with, and output assembly code to @var{file} corresponding to
8106 the pattern @var{x}. This may be used to allow machine-dependent
8107 @code{UNSPEC}s to appear within constants.
8109 If target hook fails to recognize a pattern, it must return @code{false},
8110 so that a standard error message is printed. If it prints an error message
8111 itself, by calling, for example, @code{output_operand_lossage}, it may just
8115 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
8116 A C statement to output to the stdio stream @var{stream} an assembler
8117 instruction to assemble a string constant containing the @var{len}
8118 bytes at @var{ptr}. @var{ptr} will be a C expression of type
8119 @code{char *} and @var{len} a C expression of type @code{int}.
8121 If the assembler has a @code{.ascii} pseudo-op as found in the
8122 Berkeley Unix assembler, do not define the macro
8123 @code{ASM_OUTPUT_ASCII}.
8126 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
8127 A C statement to output word @var{n} of a function descriptor for
8128 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
8129 is defined, and is otherwise unused.
8132 @defmac CONSTANT_POOL_BEFORE_FUNCTION
8133 You may define this macro as a C expression. You should define the
8134 expression to have a nonzero value if GCC should output the constant
8135 pool for a function before the code for the function, or a zero value if
8136 GCC should output the constant pool after the function. If you do
8137 not define this macro, the usual case, GCC will output the constant
8138 pool before the function.
8141 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
8142 A C statement to output assembler commands to define the start of the
8143 constant pool for a function. @var{funname} is a string giving
8144 the name of the function. Should the return type of the function
8145 be required, it can be obtained via @var{fundecl}. @var{size}
8146 is the size, in bytes, of the constant pool that will be written
8147 immediately after this call.
8149 If no constant-pool prefix is required, the usual case, this macro need
8153 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
8154 A C statement (with or without semicolon) to output a constant in the
8155 constant pool, if it needs special treatment. (This macro need not do
8156 anything for RTL expressions that can be output normally.)
8158 The argument @var{file} is the standard I/O stream to output the
8159 assembler code on. @var{x} is the RTL expression for the constant to
8160 output, and @var{mode} is the machine mode (in case @var{x} is a
8161 @samp{const_int}). @var{align} is the required alignment for the value
8162 @var{x}; you should output an assembler directive to force this much
8165 The argument @var{labelno} is a number to use in an internal label for
8166 the address of this pool entry. The definition of this macro is
8167 responsible for outputting the label definition at the proper place.
8168 Here is how to do this:
8171 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
8174 When you output a pool entry specially, you should end with a
8175 @code{goto} to the label @var{jumpto}. This will prevent the same pool
8176 entry from being output a second time in the usual manner.
8178 You need not define this macro if it would do nothing.
8181 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
8182 A C statement to output assembler commands to at the end of the constant
8183 pool for a function. @var{funname} is a string giving the name of the
8184 function. Should the return type of the function be required, you can
8185 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
8186 constant pool that GCC wrote immediately before this call.
8188 If no constant-pool epilogue is required, the usual case, you need not
8192 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
8193 Define this macro as a C expression which is nonzero if @var{C} is
8194 used as a logical line separator by the assembler. @var{STR} points
8195 to the position in the string where @var{C} was found; this can be used if
8196 a line separator uses multiple characters.
8198 If you do not define this macro, the default is that only
8199 the character @samp{;} is treated as a logical line separator.
8202 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
8203 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
8204 These target hooks are C string constants, describing the syntax in the
8205 assembler for grouping arithmetic expressions. If not overridden, they
8206 default to normal parentheses, which is correct for most assemblers.
8209 These macros are provided by @file{real.h} for writing the definitions
8210 of @code{ASM_OUTPUT_DOUBLE} and the like:
8212 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
8213 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
8214 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
8215 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
8216 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
8217 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
8218 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
8219 target's floating point representation, and store its bit pattern in
8220 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
8221 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
8222 simple @code{long int}. For the others, it should be an array of
8223 @code{long int}. The number of elements in this array is determined
8224 by the size of the desired target floating point data type: 32 bits of
8225 it go in each @code{long int} array element. Each array element holds
8226 32 bits of the result, even if @code{long int} is wider than 32 bits
8227 on the host machine.
8229 The array element values are designed so that you can print them out
8230 using @code{fprintf} in the order they should appear in the target
8234 @node Uninitialized Data
8235 @subsection Output of Uninitialized Variables
8237 Each of the macros in this section is used to do the whole job of
8238 outputting a single uninitialized variable.
8240 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
8241 A C statement (sans semicolon) to output to the stdio stream
8242 @var{stream} the assembler definition of a common-label named
8243 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
8244 is the size rounded up to whatever alignment the caller wants. It is
8245 possible that @var{size} may be zero, for instance if a struct with no
8246 other member than a zero-length array is defined. In this case, the
8247 backend must output a symbol definition that allocates at least one
8248 byte, both so that the address of the resulting object does not compare
8249 equal to any other, and because some object formats cannot even express
8250 the concept of a zero-sized common symbol, as that is how they represent
8251 an ordinary undefined external.
8253 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8254 output the name itself; before and after that, output the additional
8255 assembler syntax for defining the name, and a newline.
8257 This macro controls how the assembler definitions of uninitialized
8258 common global variables are output.
8261 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
8262 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
8263 separate, explicit argument. If you define this macro, it is used in
8264 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
8265 handling the required alignment of the variable. The alignment is specified
8266 as the number of bits.
8269 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
8270 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
8271 variable to be output, if there is one, or @code{NULL_TREE} if there
8272 is no corresponding variable. If you define this macro, GCC will use it
8273 in place of both @code{ASM_OUTPUT_COMMON} and
8274 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
8275 the variable's decl in order to chose what to output.
8278 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
8279 A C statement (sans semicolon) to output to the stdio stream
8280 @var{stream} the assembler definition of uninitialized global @var{decl} named
8281 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
8282 is the alignment specified as the number of bits.
8284 Try to use function @code{asm_output_aligned_bss} defined in file
8285 @file{varasm.c} when defining this macro. If unable, use the expression
8286 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
8287 before and after that, output the additional assembler syntax for defining
8288 the name, and a newline.
8290 There are two ways of handling global BSS@. One is to define this macro.
8291 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
8292 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
8293 You do not need to do both.
8295 Some languages do not have @code{common} data, and require a
8296 non-common form of global BSS in order to handle uninitialized globals
8297 efficiently. C++ is one example of this. However, if the target does
8298 not support global BSS, the front end may choose to make globals
8299 common in order to save space in the object file.
8302 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
8303 A C statement (sans semicolon) to output to the stdio stream
8304 @var{stream} the assembler definition of a local-common-label named
8305 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
8306 is the size rounded up to whatever alignment the caller wants.
8308 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8309 output the name itself; before and after that, output the additional
8310 assembler syntax for defining the name, and a newline.
8312 This macro controls how the assembler definitions of uninitialized
8313 static variables are output.
8316 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
8317 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
8318 separate, explicit argument. If you define this macro, it is used in
8319 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
8320 handling the required alignment of the variable. The alignment is specified
8321 as the number of bits.
8324 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
8325 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
8326 variable to be output, if there is one, or @code{NULL_TREE} if there
8327 is no corresponding variable. If you define this macro, GCC will use it
8328 in place of both @code{ASM_OUTPUT_DECL} and
8329 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
8330 the variable's decl in order to chose what to output.
8334 @subsection Output and Generation of Labels
8336 @c prevent bad page break with this line
8337 This is about outputting labels.
8339 @findex assemble_name
8340 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
8341 A C statement (sans semicolon) to output to the stdio stream
8342 @var{stream} the assembler definition of a label named @var{name}.
8343 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8344 output the name itself; before and after that, output the additional
8345 assembler syntax for defining the name, and a newline. A default
8346 definition of this macro is provided which is correct for most systems.
8349 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
8350 A C statement (sans semicolon) to output to the stdio stream
8351 @var{stream} the assembler definition of a label named @var{name} of
8353 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8354 output the name itself; before and after that, output the additional
8355 assembler syntax for defining the name, and a newline. A default
8356 definition of this macro is provided which is correct for most systems.
8358 If this macro is not defined, then the function name is defined in the
8359 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8362 @findex assemble_name_raw
8363 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
8364 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
8365 to refer to a compiler-generated label. The default definition uses
8366 @code{assemble_name_raw}, which is like @code{assemble_name} except
8367 that it is more efficient.
8371 A C string containing the appropriate assembler directive to specify the
8372 size of a symbol, without any arguments. On systems that use ELF, the
8373 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
8374 systems, the default is not to define this macro.
8376 Define this macro only if it is correct to use the default definitions
8377 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
8378 for your system. If you need your own custom definitions of those
8379 macros, or if you do not need explicit symbol sizes at all, do not
8383 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
8384 A C statement (sans semicolon) to output to the stdio stream
8385 @var{stream} a directive telling the assembler that the size of the
8386 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
8387 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
8391 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
8392 A C statement (sans semicolon) to output to the stdio stream
8393 @var{stream} a directive telling the assembler to calculate the size of
8394 the symbol @var{name} by subtracting its address from the current
8397 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
8398 provided. The default assumes that the assembler recognizes a special
8399 @samp{.} symbol as referring to the current address, and can calculate
8400 the difference between this and another symbol. If your assembler does
8401 not recognize @samp{.} or cannot do calculations with it, you will need
8402 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
8405 @defmac NO_DOLLAR_IN_LABEL
8406 Define this macro if the assembler does not accept the character
8407 @samp{$} in label names. By default constructors and destructors in
8408 G++ have @samp{$} in the identifiers. If this macro is defined,
8409 @samp{.} is used instead.
8412 @defmac NO_DOT_IN_LABEL
8413 Define this macro if the assembler does not accept the character
8414 @samp{.} in label names. By default constructors and destructors in G++
8415 have names that use @samp{.}. If this macro is defined, these names
8416 are rewritten to avoid @samp{.}.
8420 A C string containing the appropriate assembler directive to specify the
8421 type of a symbol, without any arguments. On systems that use ELF, the
8422 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
8423 systems, the default is not to define this macro.
8425 Define this macro only if it is correct to use the default definition of
8426 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
8427 custom definition of this macro, or if you do not need explicit symbol
8428 types at all, do not define this macro.
8431 @defmac TYPE_OPERAND_FMT
8432 A C string which specifies (using @code{printf} syntax) the format of
8433 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
8434 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
8435 the default is not to define this macro.
8437 Define this macro only if it is correct to use the default definition of
8438 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
8439 custom definition of this macro, or if you do not need explicit symbol
8440 types at all, do not define this macro.
8443 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
8444 A C statement (sans semicolon) to output to the stdio stream
8445 @var{stream} a directive telling the assembler that the type of the
8446 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
8447 that string is always either @samp{"function"} or @samp{"object"}, but
8448 you should not count on this.
8450 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
8451 definition of this macro is provided.
8454 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
8455 A C statement (sans semicolon) to output to the stdio stream
8456 @var{stream} any text necessary for declaring the name @var{name} of a
8457 function which is being defined. This macro is responsible for
8458 outputting the label definition (perhaps using
8459 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
8460 @code{FUNCTION_DECL} tree node representing the function.
8462 If this macro is not defined, then the function name is defined in the
8463 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
8465 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
8469 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
8470 A C statement (sans semicolon) to output to the stdio stream
8471 @var{stream} any text necessary for declaring the size of a function
8472 which is being defined. The argument @var{name} is the name of the
8473 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
8474 representing the function.
8476 If this macro is not defined, then the function size is not defined.
8478 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
8482 @defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
8483 A C statement (sans semicolon) to output to the stdio stream
8484 @var{stream} any text necessary for declaring the name @var{name} of a
8485 cold function partition which is being defined. This macro is responsible
8486 for outputting the label definition (perhaps using
8487 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
8488 @code{FUNCTION_DECL} tree node representing the function.
8490 If this macro is not defined, then the cold partition name is defined in the
8491 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8493 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
8497 @defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
8498 A C statement (sans semicolon) to output to the stdio stream
8499 @var{stream} any text necessary for declaring the size of a cold function
8500 partition which is being defined. The argument @var{name} is the name of the
8501 cold partition of the function. The argument @var{decl} is the
8502 @code{FUNCTION_DECL} tree node representing the function.
8504 If this macro is not defined, then the partition size is not defined.
8506 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
8510 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
8511 A C statement (sans semicolon) to output to the stdio stream
8512 @var{stream} any text necessary for declaring the name @var{name} of an
8513 initialized variable which is being defined. This macro must output the
8514 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
8515 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
8517 If this macro is not defined, then the variable name is defined in the
8518 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8520 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
8521 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
8524 @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})
8525 A target hook to output to the stdio stream @var{file} any text necessary
8526 for declaring the name @var{name} of a constant which is being defined. This
8527 target hook is responsible for outputting the label definition (perhaps using
8528 @code{assemble_label}). The argument @var{exp} is the value of the constant,
8529 and @var{size} is the size of the constant in bytes. The @var{name}
8530 will be an internal label.
8532 The default version of this target hook, define the @var{name} in the
8533 usual manner as a label (by means of @code{assemble_label}).
8535 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
8538 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
8539 A C statement (sans semicolon) to output to the stdio stream
8540 @var{stream} any text necessary for claiming a register @var{regno}
8541 for a global variable @var{decl} with name @var{name}.
8543 If you don't define this macro, that is equivalent to defining it to do
8547 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
8548 A C statement (sans semicolon) to finish up declaring a variable name
8549 once the compiler has processed its initializer fully and thus has had a
8550 chance to determine the size of an array when controlled by an
8551 initializer. This is used on systems where it's necessary to declare
8552 something about the size of the object.
8554 If you don't define this macro, that is equivalent to defining it to do
8557 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
8558 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
8561 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
8562 This target hook is a function to output to the stdio stream
8563 @var{stream} some commands that will make the label @var{name} global;
8564 that is, available for reference from other files.
8566 The default implementation relies on a proper definition of
8567 @code{GLOBAL_ASM_OP}.
8570 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
8571 This target hook is a function to output to the stdio stream
8572 @var{stream} some commands that will make the name associated with @var{decl}
8573 global; that is, available for reference from other files.
8575 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
8578 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_UNDEFINED_DECL (FILE *@var{stream}, const char *@var{name}, const_tree @var{decl})
8579 This target hook is a function to output to the stdio stream
8580 @var{stream} some commands that will declare the name associated with
8581 @var{decl} which is not defined in the current translation unit. Most
8582 assemblers do not require anything to be output in this case.
8585 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
8586 A C statement (sans semicolon) to output to the stdio stream
8587 @var{stream} some commands that will make the label @var{name} weak;
8588 that is, available for reference from other files but only used if
8589 no other definition is available. Use the expression
8590 @code{assemble_name (@var{stream}, @var{name})} to output the name
8591 itself; before and after that, output the additional assembler syntax
8592 for making that name weak, and a newline.
8594 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
8595 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
8599 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
8600 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
8601 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
8602 or variable decl. If @var{value} is not @code{NULL}, this C statement
8603 should output to the stdio stream @var{stream} assembler code which
8604 defines (equates) the weak symbol @var{name} to have the value
8605 @var{value}. If @var{value} is @code{NULL}, it should output commands
8606 to make @var{name} weak.
8609 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
8610 Outputs a directive that enables @var{name} to be used to refer to
8611 symbol @var{value} with weak-symbol semantics. @code{decl} is the
8612 declaration of @code{name}.
8615 @defmac SUPPORTS_WEAK
8616 A preprocessor constant expression which evaluates to true if the target
8617 supports weak symbols.
8619 If you don't define this macro, @file{defaults.h} provides a default
8620 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
8621 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
8624 @defmac TARGET_SUPPORTS_WEAK
8625 A C expression which evaluates to true if the target supports weak symbols.
8627 If you don't define this macro, @file{defaults.h} provides a default
8628 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
8629 this macro if you want to control weak symbol support with a compiler
8630 flag such as @option{-melf}.
8633 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
8634 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
8635 public symbol such that extra copies in multiple translation units will
8636 be discarded by the linker. Define this macro if your object file
8637 format provides support for this concept, such as the @samp{COMDAT}
8638 section flags in the Microsoft Windows PE/COFF format, and this support
8639 requires changes to @var{decl}, such as putting it in a separate section.
8642 @defmac SUPPORTS_ONE_ONLY
8643 A C expression which evaluates to true if the target supports one-only
8646 If you don't define this macro, @file{varasm.c} provides a default
8647 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
8648 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
8649 you want to control one-only symbol support with a compiler flag, or if
8650 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
8651 be emitted as one-only.
8654 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
8655 This target hook is a function to output to @var{asm_out_file} some
8656 commands that will make the symbol(s) associated with @var{decl} have
8657 hidden, protected or internal visibility as specified by @var{visibility}.
8660 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
8661 A C expression that evaluates to true if the target's linker expects
8662 that weak symbols do not appear in a static archive's table of contents.
8663 The default is @code{0}.
8665 Leaving weak symbols out of an archive's table of contents means that,
8666 if a symbol will only have a definition in one translation unit and
8667 will have undefined references from other translation units, that
8668 symbol should not be weak. Defining this macro to be nonzero will
8669 thus have the effect that certain symbols that would normally be weak
8670 (explicit template instantiations, and vtables for polymorphic classes
8671 with noninline key methods) will instead be nonweak.
8673 The C++ ABI requires this macro to be zero. Define this macro for
8674 targets where full C++ ABI compliance is impossible and where linker
8675 restrictions require weak symbols to be left out of a static archive's
8679 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8680 A C statement (sans semicolon) to output to the stdio stream
8681 @var{stream} any text necessary for declaring the name of an external
8682 symbol named @var{name} which is referenced in this compilation but
8683 not defined. The value of @var{decl} is the tree node for the
8686 This macro need not be defined if it does not need to output anything.
8687 The GNU assembler and most Unix assemblers don't require anything.
8690 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8691 This target hook is a function to output to @var{asm_out_file} an assembler
8692 pseudo-op to declare a library function name external. The name of the
8693 library function is given by @var{symref}, which is a @code{symbol_ref}.
8696 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8697 This target hook is a function to output to @var{asm_out_file} an assembler
8698 directive to annotate @var{symbol} as used. The Darwin target uses the
8699 .no_dead_code_strip directive.
8702 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8703 A C statement (sans semicolon) to output to the stdio stream
8704 @var{stream} a reference in assembler syntax to a label named
8705 @var{name}. This should add @samp{_} to the front of the name, if that
8706 is customary on your operating system, as it is in most Berkeley Unix
8707 systems. This macro is used in @code{assemble_name}.
8710 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8711 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.
8714 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8715 A C statement (sans semicolon) to output a reference to
8716 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8717 will be used to output the name of the symbol. This macro may be used
8718 to modify the way a symbol is referenced depending on information
8719 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8722 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8723 A C statement (sans semicolon) to output a reference to @var{buf}, the
8724 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8725 @code{assemble_name} will be used to output the name of the symbol.
8726 This macro is not used by @code{output_asm_label}, or the @code{%l}
8727 specifier that calls it; the intention is that this macro should be set
8728 when it is necessary to output a label differently when its address is
8732 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8733 A function to output to the stdio stream @var{stream} a label whose
8734 name is made from the string @var{prefix} and the number @var{labelno}.
8736 It is absolutely essential that these labels be distinct from the labels
8737 used for user-level functions and variables. Otherwise, certain programs
8738 will have name conflicts with internal labels.
8740 It is desirable to exclude internal labels from the symbol table of the
8741 object file. Most assemblers have a naming convention for labels that
8742 should be excluded; on many systems, the letter @samp{L} at the
8743 beginning of a label has this effect. You should find out what
8744 convention your system uses, and follow it.
8746 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8749 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8750 A C statement to output to the stdio stream @var{stream} a debug info
8751 label whose name is made from the string @var{prefix} and the number
8752 @var{num}. This is useful for VLIW targets, where debug info labels
8753 may need to be treated differently than branch target labels. On some
8754 systems, branch target labels must be at the beginning of instruction
8755 bundles, but debug info labels can occur in the middle of instruction
8758 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8762 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8763 A C statement to store into the string @var{string} a label whose name
8764 is made from the string @var{prefix} and the number @var{num}.
8766 This string, when output subsequently by @code{assemble_name}, should
8767 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8768 with the same @var{prefix} and @var{num}.
8770 If the string begins with @samp{*}, then @code{assemble_name} will
8771 output the rest of the string unchanged. It is often convenient for
8772 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8773 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8774 to output the string, and may change it. (Of course,
8775 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8776 you should know what it does on your machine.)
8779 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8780 A C expression to assign to @var{outvar} (which is a variable of type
8781 @code{char *}) a newly allocated string made from the string
8782 @var{name} and the number @var{number}, with some suitable punctuation
8783 added. Use @code{alloca} to get space for the string.
8785 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8786 produce an assembler label for an internal static variable whose name is
8787 @var{name}. Therefore, the string must be such as to result in valid
8788 assembler code. The argument @var{number} is different each time this
8789 macro is executed; it prevents conflicts between similarly-named
8790 internal static variables in different scopes.
8792 Ideally this string should not be a valid C identifier, to prevent any
8793 conflict with the user's own symbols. Most assemblers allow periods
8794 or percent signs in assembler symbols; putting at least one of these
8795 between the name and the number will suffice.
8797 If this macro is not defined, a default definition will be provided
8798 which is correct for most systems.
8801 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8802 A C statement to output to the stdio stream @var{stream} assembler code
8803 which defines (equates) the symbol @var{name} to have the value @var{value}.
8806 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8807 correct for most systems.
8810 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8811 A C statement to output to the stdio stream @var{stream} assembler code
8812 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8813 to have the value of the tree node @var{decl_of_value}. This macro will
8814 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8815 the tree nodes are available.
8818 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8819 correct for most systems.
8822 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8823 A C statement that evaluates to true if the assembler code which defines
8824 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8825 of the tree node @var{decl_of_value} should be emitted near the end of the
8826 current compilation unit. The default is to not defer output of defines.
8827 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8828 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8831 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8832 A C statement to output to the stdio stream @var{stream} assembler code
8833 which defines (equates) the weak symbol @var{name} to have the value
8834 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8835 an undefined weak symbol.
8837 Define this macro if the target only supports weak aliases; define
8838 @code{ASM_OUTPUT_DEF} instead if possible.
8841 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8842 Define this macro to override the default assembler names used for
8843 Objective-C methods.
8845 The default name is a unique method number followed by the name of the
8846 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8847 the category is also included in the assembler name (e.g.@:
8850 These names are safe on most systems, but make debugging difficult since
8851 the method's selector is not present in the name. Therefore, particular
8852 systems define other ways of computing names.
8854 @var{buf} is an expression of type @code{char *} which gives you a
8855 buffer in which to store the name; its length is as long as
8856 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8857 50 characters extra.
8859 The argument @var{is_inst} specifies whether the method is an instance
8860 method or a class method; @var{class_name} is the name of the class;
8861 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8862 in a category); and @var{sel_name} is the name of the selector.
8864 On systems where the assembler can handle quoted names, you can use this
8865 macro to provide more human-readable names.
8868 @node Initialization
8869 @subsection How Initialization Functions Are Handled
8870 @cindex initialization routines
8871 @cindex termination routines
8872 @cindex constructors, output of
8873 @cindex destructors, output of
8875 The compiled code for certain languages includes @dfn{constructors}
8876 (also called @dfn{initialization routines})---functions to initialize
8877 data in the program when the program is started. These functions need
8878 to be called before the program is ``started''---that is to say, before
8879 @code{main} is called.
8881 Compiling some languages generates @dfn{destructors} (also called
8882 @dfn{termination routines}) that should be called when the program
8885 To make the initialization and termination functions work, the compiler
8886 must output something in the assembler code to cause those functions to
8887 be called at the appropriate time. When you port the compiler to a new
8888 system, you need to specify how to do this.
8890 There are two major ways that GCC currently supports the execution of
8891 initialization and termination functions. Each way has two variants.
8892 Much of the structure is common to all four variations.
8894 @findex __CTOR_LIST__
8895 @findex __DTOR_LIST__
8896 The linker must build two lists of these functions---a list of
8897 initialization functions, called @code{__CTOR_LIST__}, and a list of
8898 termination functions, called @code{__DTOR_LIST__}.
8900 Each list always begins with an ignored function pointer (which may hold
8901 0, @minus{}1, or a count of the function pointers after it, depending on
8902 the environment). This is followed by a series of zero or more function
8903 pointers to constructors (or destructors), followed by a function
8904 pointer containing zero.
8906 Depending on the operating system and its executable file format, either
8907 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8908 time and exit time. Constructors are called in reverse order of the
8909 list; destructors in forward order.
8911 The best way to handle static constructors works only for object file
8912 formats which provide arbitrarily-named sections. A section is set
8913 aside for a list of constructors, and another for a list of destructors.
8914 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8915 object file that defines an initialization function also puts a word in
8916 the constructor section to point to that function. The linker
8917 accumulates all these words into one contiguous @samp{.ctors} section.
8918 Termination functions are handled similarly.
8920 This method will be chosen as the default by @file{target-def.h} if
8921 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8922 support arbitrary sections, but does support special designated
8923 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8924 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8926 When arbitrary sections are available, there are two variants, depending
8927 upon how the code in @file{crtstuff.c} is called. On systems that
8928 support a @dfn{.init} section which is executed at program startup,
8929 parts of @file{crtstuff.c} are compiled into that section. The
8930 program is linked by the @command{gcc} driver like this:
8933 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8936 The prologue of a function (@code{__init}) appears in the @code{.init}
8937 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8938 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8939 files are provided by the operating system or by the GNU C library, but
8940 are provided by GCC for a few targets.
8942 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8943 compiled from @file{crtstuff.c}. They contain, among other things, code
8944 fragments within the @code{.init} and @code{.fini} sections that branch
8945 to routines in the @code{.text} section. The linker will pull all parts
8946 of a section together, which results in a complete @code{__init} function
8947 that invokes the routines we need at startup.
8949 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8952 If no init section is available, when GCC compiles any function called
8953 @code{main} (or more accurately, any function designated as a program
8954 entry point by the language front end calling @code{expand_main_function}),
8955 it inserts a procedure call to @code{__main} as the first executable code
8956 after the function prologue. The @code{__main} function is defined
8957 in @file{libgcc2.c} and runs the global constructors.
8959 In file formats that don't support arbitrary sections, there are again
8960 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8961 and an `a.out' format must be used. In this case,
8962 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8963 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8964 and with the address of the void function containing the initialization
8965 code as its value. The GNU linker recognizes this as a request to add
8966 the value to a @dfn{set}; the values are accumulated, and are eventually
8967 placed in the executable as a vector in the format described above, with
8968 a leading (ignored) count and a trailing zero element.
8969 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8970 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8971 the compilation of @code{main} to call @code{__main} as above, starting
8972 the initialization process.
8974 The last variant uses neither arbitrary sections nor the GNU linker.
8975 This is preferable when you want to do dynamic linking and when using
8976 file formats which the GNU linker does not support, such as `ECOFF'@. In
8977 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8978 termination functions are recognized simply by their names. This requires
8979 an extra program in the linkage step, called @command{collect2}. This program
8980 pretends to be the linker, for use with GCC; it does its job by running
8981 the ordinary linker, but also arranges to include the vectors of
8982 initialization and termination functions. These functions are called
8983 via @code{__main} as described above. In order to use this method,
8984 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8987 The following section describes the specific macros that control and
8988 customize the handling of initialization and termination functions.
8991 @node Macros for Initialization
8992 @subsection Macros Controlling Initialization Routines
8994 Here are the macros that control how the compiler handles initialization
8995 and termination functions:
8997 @defmac INIT_SECTION_ASM_OP
8998 If defined, a C string constant, including spacing, for the assembler
8999 operation to identify the following data as initialization code. If not
9000 defined, GCC will assume such a section does not exist. When you are
9001 using special sections for initialization and termination functions, this
9002 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
9003 run the initialization functions.
9006 @defmac HAS_INIT_SECTION
9007 If defined, @code{main} will not call @code{__main} as described above.
9008 This macro should be defined for systems that control start-up code
9009 on a symbol-by-symbol basis, such as OSF/1, and should not
9010 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
9013 @defmac LD_INIT_SWITCH
9014 If defined, a C string constant for a switch that tells the linker that
9015 the following symbol is an initialization routine.
9018 @defmac LD_FINI_SWITCH
9019 If defined, a C string constant for a switch that tells the linker that
9020 the following symbol is a finalization routine.
9023 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
9024 If defined, a C statement that will write a function that can be
9025 automatically called when a shared library is loaded. The function
9026 should call @var{func}, which takes no arguments. If not defined, and
9027 the object format requires an explicit initialization function, then a
9028 function called @code{_GLOBAL__DI} will be generated.
9030 This function and the following one are used by collect2 when linking a
9031 shared library that needs constructors or destructors, or has DWARF2
9032 exception tables embedded in the code.
9035 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
9036 If defined, a C statement that will write a function that can be
9037 automatically called when a shared library is unloaded. The function
9038 should call @var{func}, which takes no arguments. If not defined, and
9039 the object format requires an explicit finalization function, then a
9040 function called @code{_GLOBAL__DD} will be generated.
9043 @defmac INVOKE__main
9044 If defined, @code{main} will call @code{__main} despite the presence of
9045 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
9046 where the init section is not actually run automatically, but is still
9047 useful for collecting the lists of constructors and destructors.
9050 @defmac SUPPORTS_INIT_PRIORITY
9051 If nonzero, the C++ @code{init_priority} attribute is supported and the
9052 compiler should emit instructions to control the order of initialization
9053 of objects. If zero, the compiler will issue an error message upon
9054 encountering an @code{init_priority} attribute.
9057 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
9058 This value is true if the target supports some ``native'' method of
9059 collecting constructors and destructors to be run at startup and exit.
9060 It is false if we must use @command{collect2}.
9063 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
9064 If defined, a function that outputs assembler code to arrange to call
9065 the function referenced by @var{symbol} at initialization time.
9067 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
9068 no arguments and with no return value. If the target supports initialization
9069 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
9070 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
9072 If this macro is not defined by the target, a suitable default will
9073 be chosen if (1) the target supports arbitrary section names, (2) the
9074 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
9078 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
9079 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
9080 functions rather than initialization functions.
9083 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
9084 generated for the generated object file will have static linkage.
9086 If your system uses @command{collect2} as the means of processing
9087 constructors, then that program normally uses @command{nm} to scan
9088 an object file for constructor functions to be called.
9090 On certain kinds of systems, you can define this macro to make
9091 @command{collect2} work faster (and, in some cases, make it work at all):
9093 @defmac OBJECT_FORMAT_COFF
9094 Define this macro if the system uses COFF (Common Object File Format)
9095 object files, so that @command{collect2} can assume this format and scan
9096 object files directly for dynamic constructor/destructor functions.
9098 This macro is effective only in a native compiler; @command{collect2} as
9099 part of a cross compiler always uses @command{nm} for the target machine.
9102 @defmac REAL_NM_FILE_NAME
9103 Define this macro as a C string constant containing the file name to use
9104 to execute @command{nm}. The default is to search the path normally for
9109 @command{collect2} calls @command{nm} to scan object files for static
9110 constructors and destructors and LTO info. By default, @option{-n} is
9111 passed. Define @code{NM_FLAGS} to a C string constant if other options
9112 are needed to get the same output format as GNU @command{nm -n}
9116 If your system supports shared libraries and has a program to list the
9117 dynamic dependencies of a given library or executable, you can define
9118 these macros to enable support for running initialization and
9119 termination functions in shared libraries:
9122 Define this macro to a C string constant containing the name of the program
9123 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
9126 @defmac PARSE_LDD_OUTPUT (@var{ptr})
9127 Define this macro to be C code that extracts filenames from the output
9128 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
9129 of type @code{char *} that points to the beginning of a line of output
9130 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
9131 code must advance @var{ptr} to the beginning of the filename on that
9132 line. Otherwise, it must set @var{ptr} to @code{NULL}.
9135 @defmac SHLIB_SUFFIX
9136 Define this macro to a C string constant containing the default shared
9137 library extension of the target (e.g., @samp{".so"}). @command{collect2}
9138 strips version information after this suffix when generating global
9139 constructor and destructor names. This define is only needed on targets
9140 that use @command{collect2} to process constructors and destructors.
9143 @node Instruction Output
9144 @subsection Output of Assembler Instructions
9146 @c prevent bad page break with this line
9147 This describes assembler instruction output.
9149 @defmac REGISTER_NAMES
9150 A C initializer containing the assembler's names for the machine
9151 registers, each one as a C string constant. This is what translates
9152 register numbers in the compiler into assembler language.
9155 @defmac ADDITIONAL_REGISTER_NAMES
9156 If defined, a C initializer for an array of structures containing a name
9157 and a register number. This macro defines additional names for hard
9158 registers, thus allowing the @code{asm} option in declarations to refer
9159 to registers using alternate names.
9162 @defmac OVERLAPPING_REGISTER_NAMES
9163 If defined, a C initializer for an array of structures containing a
9164 name, a register number and a count of the number of consecutive
9165 machine registers the name overlaps. This macro defines additional
9166 names for hard registers, thus allowing the @code{asm} option in
9167 declarations to refer to registers using alternate names. Unlike
9168 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
9169 register name implies multiple underlying registers.
9171 This macro should be used when it is important that a clobber in an
9172 @code{asm} statement clobbers all the underlying values implied by the
9173 register name. For example, on ARM, clobbering the double-precision
9174 VFP register ``d0'' implies clobbering both single-precision registers
9178 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
9179 Define this macro if you are using an unusual assembler that
9180 requires different names for the machine instructions.
9182 The definition is a C statement or statements which output an
9183 assembler instruction opcode to the stdio stream @var{stream}. The
9184 macro-operand @var{ptr} is a variable of type @code{char *} which
9185 points to the opcode name in its ``internal'' form---the form that is
9186 written in the machine description. The definition should output the
9187 opcode name to @var{stream}, performing any translation you desire, and
9188 increment the variable @var{ptr} to point at the end of the opcode
9189 so that it will not be output twice.
9191 In fact, your macro definition may process less than the entire opcode
9192 name, or more than the opcode name; but if you want to process text
9193 that includes @samp{%}-sequences to substitute operands, you must take
9194 care of the substitution yourself. Just be sure to increment
9195 @var{ptr} over whatever text should not be output normally.
9197 @findex recog_data.operand
9198 If you need to look at the operand values, they can be found as the
9199 elements of @code{recog_data.operand}.
9201 If the macro definition does nothing, the instruction is output
9205 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
9206 If defined, a C statement to be executed just prior to the output of
9207 assembler code for @var{insn}, to modify the extracted operands so
9208 they will be output differently.
9210 Here the argument @var{opvec} is the vector containing the operands
9211 extracted from @var{insn}, and @var{noperands} is the number of
9212 elements of the vector which contain meaningful data for this insn.
9213 The contents of this vector are what will be used to convert the insn
9214 template into assembler code, so you can change the assembler output
9215 by changing the contents of the vector.
9217 This macro is useful when various assembler syntaxes share a single
9218 file of instruction patterns; by defining this macro differently, you
9219 can cause a large class of instructions to be output differently (such
9220 as with rearranged operands). Naturally, variations in assembler
9221 syntax affecting individual insn patterns ought to be handled by
9222 writing conditional output routines in those patterns.
9224 If this macro is not defined, it is equivalent to a null statement.
9227 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx_insn *@var{insn}, rtx *@var{opvec}, int @var{noperands})
9228 If defined, this target hook is a function which is executed just after the
9229 output of assembler code for @var{insn}, to change the mode of the assembler
9232 Here the argument @var{opvec} is the vector containing the operands
9233 extracted from @var{insn}, and @var{noperands} is the number of
9234 elements of the vector which contain meaningful data for this insn.
9235 The contents of this vector are what was used to convert the insn
9236 template into assembler code, so you can change the assembler mode
9237 by checking the contents of the vector.
9240 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
9241 A C compound statement to output to stdio stream @var{stream} the
9242 assembler syntax for an instruction operand @var{x}. @var{x} is an
9245 @var{code} is a value that can be used to specify one of several ways
9246 of printing the operand. It is used when identical operands must be
9247 printed differently depending on the context. @var{code} comes from
9248 the @samp{%} specification that was used to request printing of the
9249 operand. If the specification was just @samp{%@var{digit}} then
9250 @var{code} is 0; if the specification was @samp{%@var{ltr}
9251 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
9254 If @var{x} is a register, this macro should print the register's name.
9255 The names can be found in an array @code{reg_names} whose type is
9256 @code{char *[]}. @code{reg_names} is initialized from
9257 @code{REGISTER_NAMES}.
9259 When the machine description has a specification @samp{%@var{punct}}
9260 (a @samp{%} followed by a punctuation character), this macro is called
9261 with a null pointer for @var{x} and the punctuation character for
9265 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
9266 A C expression which evaluates to true if @var{code} is a valid
9267 punctuation character for use in the @code{PRINT_OPERAND} macro. If
9268 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
9269 punctuation characters (except for the standard one, @samp{%}) are used
9273 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
9274 A C compound statement to output to stdio stream @var{stream} the
9275 assembler syntax for an instruction operand that is a memory reference
9276 whose address is @var{x}. @var{x} is an RTL expression.
9278 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
9279 On some machines, the syntax for a symbolic address depends on the
9280 section that the address refers to. On these machines, define the hook
9281 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
9282 @code{symbol_ref}, and then check for it here. @xref{Assembler
9286 @findex dbr_sequence_length
9287 @defmac DBR_OUTPUT_SEQEND (@var{file})
9288 A C statement, to be executed after all slot-filler instructions have
9289 been output. If necessary, call @code{dbr_sequence_length} to
9290 determine the number of slots filled in a sequence (zero if not
9291 currently outputting a sequence), to decide how many no-ops to output,
9294 Don't define this macro if it has nothing to do, but it is helpful in
9295 reading assembly output if the extent of the delay sequence is made
9296 explicit (e.g.@: with white space).
9299 @findex final_sequence
9300 Note that output routines for instructions with delay slots must be
9301 prepared to deal with not being output as part of a sequence
9302 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
9303 found.) The variable @code{final_sequence} is null when not
9304 processing a sequence, otherwise it contains the @code{sequence} rtx
9308 @defmac REGISTER_PREFIX
9309 @defmacx LOCAL_LABEL_PREFIX
9310 @defmacx USER_LABEL_PREFIX
9311 @defmacx IMMEDIATE_PREFIX
9312 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
9313 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
9314 @file{final.c}). These are useful when a single @file{md} file must
9315 support multiple assembler formats. In that case, the various @file{tm.h}
9316 files can define these macros differently.
9319 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
9320 If defined this macro should expand to a series of @code{case}
9321 statements which will be parsed inside the @code{switch} statement of
9322 the @code{asm_fprintf} function. This allows targets to define extra
9323 printf formats which may useful when generating their assembler
9324 statements. Note that uppercase letters are reserved for future
9325 generic extensions to asm_fprintf, and so are not available to target
9326 specific code. The output file is given by the parameter @var{file}.
9327 The varargs input pointer is @var{argptr} and the rest of the format
9328 string, starting the character after the one that is being switched
9329 upon, is pointed to by @var{format}.
9332 @defmac ASSEMBLER_DIALECT
9333 If your target supports multiple dialects of assembler language (such as
9334 different opcodes), define this macro as a C expression that gives the
9335 numeric index of the assembler language dialect to use, with zero as the
9338 If this macro is defined, you may use constructs of the form
9340 @samp{@{option0|option1|option2@dots{}@}}
9343 in the output templates of patterns (@pxref{Output Template}) or in the
9344 first argument of @code{asm_fprintf}. This construct outputs
9345 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
9346 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
9347 within these strings retain their usual meaning. If there are fewer
9348 alternatives within the braces than the value of
9349 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
9350 to print curly braces or @samp{|} character in assembler output directly,
9351 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
9353 If you do not define this macro, the characters @samp{@{}, @samp{|} and
9354 @samp{@}} do not have any special meaning when used in templates or
9355 operands to @code{asm_fprintf}.
9357 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
9358 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
9359 the variations in assembler language syntax with that mechanism. Define
9360 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
9361 if the syntax variant are larger and involve such things as different
9362 opcodes or operand order.
9365 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
9366 A C expression to output to @var{stream} some assembler code
9367 which will push hard register number @var{regno} onto the stack.
9368 The code need not be optimal, since this macro is used only when
9372 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
9373 A C expression to output to @var{stream} some assembler code
9374 which will pop hard register number @var{regno} off of the stack.
9375 The code need not be optimal, since this macro is used only when
9379 @node Dispatch Tables
9380 @subsection Output of Dispatch Tables
9382 @c prevent bad page break with this line
9383 This concerns dispatch tables.
9385 @cindex dispatch table
9386 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
9387 A C statement to output to the stdio stream @var{stream} an assembler
9388 pseudo-instruction to generate a difference between two labels.
9389 @var{value} and @var{rel} are the numbers of two internal labels. The
9390 definitions of these labels are output using
9391 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
9392 way here. For example,
9395 fprintf (@var{stream}, "\t.word L%d-L%d\n",
9396 @var{value}, @var{rel})
9399 You must provide this macro on machines where the addresses in a
9400 dispatch table are relative to the table's own address. If defined, GCC
9401 will also use this macro on all machines when producing PIC@.
9402 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
9403 mode and flags can be read.
9406 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
9407 This macro should be provided on machines where the addresses
9408 in a dispatch table are absolute.
9410 The definition should be a C statement to output to the stdio stream
9411 @var{stream} an assembler pseudo-instruction to generate a reference to
9412 a label. @var{value} is the number of an internal label whose
9413 definition is output using @code{(*targetm.asm_out.internal_label)}.
9417 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
9421 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
9422 Define this if the label before a jump-table needs to be output
9423 specially. The first three arguments are the same as for
9424 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
9425 jump-table which follows (a @code{jump_table_data} containing an
9426 @code{addr_vec} or @code{addr_diff_vec}).
9428 This feature is used on system V to output a @code{swbeg} statement
9431 If this macro is not defined, these labels are output with
9432 @code{(*targetm.asm_out.internal_label)}.
9435 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
9436 Define this if something special must be output at the end of a
9437 jump-table. The definition should be a C statement to be executed
9438 after the assembler code for the table is written. It should write
9439 the appropriate code to stdio stream @var{stream}. The argument
9440 @var{table} is the jump-table insn, and @var{num} is the label-number
9441 of the preceding label.
9443 If this macro is not defined, nothing special is output at the end of
9447 @deftypefn {Target Hook} void TARGET_ASM_POST_CFI_STARTPROC (FILE *@var{}, @var{tree})
9448 This target hook is used to emit assembly strings required by the target
9449 after the .cfi_startproc directive. The first argument is the file stream to
9450 write the strings to and the second argument is the function's declaration. The
9451 expected use is to add more .cfi_* directives.
9453 The default is to not output any assembly strings.
9456 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
9457 This target hook emits a label at the beginning of each FDE@. It
9458 should be defined on targets where FDEs need special labels, and it
9459 should write the appropriate label, for the FDE associated with the
9460 function declaration @var{decl}, to the stdio stream @var{stream}.
9461 The third argument, @var{for_eh}, is a boolean: true if this is for an
9462 exception table. The fourth argument, @var{empty}, is a boolean:
9463 true if this is a placeholder label for an omitted FDE@.
9465 The default is that FDEs are not given nonlocal labels.
9468 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
9469 This target hook emits a label at the beginning of the exception table.
9470 It should be defined on targets where it is desirable for the table
9471 to be broken up according to function.
9473 The default is that no label is emitted.
9476 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
9477 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.
9480 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx_insn *@var{insn})
9481 This target hook emits assembly directives required to unwind the
9482 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
9483 returns @code{UI_TARGET}.
9486 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
9487 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.
9490 @node Exception Region Output
9491 @subsection Assembler Commands for Exception Regions
9493 @c prevent bad page break with this line
9495 This describes commands marking the start and the end of an exception
9498 @defmac EH_FRAME_SECTION_NAME
9499 If defined, a C string constant for the name of the section containing
9500 exception handling frame unwind information. If not defined, GCC will
9501 provide a default definition if the target supports named sections.
9502 @file{crtstuff.c} uses this macro to switch to the appropriate section.
9504 You should define this symbol if your target supports DWARF 2 frame
9505 unwind information and the default definition does not work.
9508 @defmac EH_FRAME_THROUGH_COLLECT2
9509 If defined, DWARF 2 frame unwind information will identified by
9510 specially named labels. The collect2 process will locate these
9511 labels and generate code to register the frames.
9513 This might be necessary, for instance, if the system linker will not
9514 place the eh_frames in-between the sentinals from @file{crtstuff.c},
9515 or if the system linker does garbage collection and sections cannot
9516 be marked as not to be collected.
9519 @defmac EH_TABLES_CAN_BE_READ_ONLY
9520 Define this macro to 1 if your target is such that no frame unwind
9521 information encoding used with non-PIC code will ever require a
9522 runtime relocation, but the linker may not support merging read-only
9523 and read-write sections into a single read-write section.
9526 @defmac MASK_RETURN_ADDR
9527 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
9528 that it does not contain any extraneous set bits in it.
9531 @defmac DWARF2_UNWIND_INFO
9532 Define this macro to 0 if your target supports DWARF 2 frame unwind
9533 information, but it does not yet work with exception handling.
9534 Otherwise, if your target supports this information (if it defines
9535 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
9536 GCC will provide a default definition of 1.
9539 @deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
9540 This hook defines the mechanism that will be used for exception handling
9541 by the target. If the target has ABI specified unwind tables, the hook
9542 should return @code{UI_TARGET}. If the target is to use the
9543 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
9544 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
9545 information, the hook should return @code{UI_DWARF2}.
9547 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
9548 This may end up simplifying other parts of target-specific code. The
9549 default implementation of this hook never returns @code{UI_NONE}.
9551 Note that the value returned by this hook should be constant. It should
9552 not depend on anything except the command-line switches described by
9553 @var{opts}. In particular, the
9554 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
9555 macros and builtin functions related to exception handling are set up
9556 depending on this setting.
9558 The default implementation of the hook first honors the
9559 @option{--enable-sjlj-exceptions} configure option, then
9560 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
9561 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
9562 must define this hook so that @var{opts} is used correctly.
9565 @deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
9566 This variable should be set to @code{true} if the target ABI requires unwinding
9567 tables even when exceptions are not used. It must not be modified by
9568 command-line option processing.
9571 @defmac DONT_USE_BUILTIN_SETJMP
9572 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
9573 should use the @code{setjmp}/@code{longjmp} functions from the C library
9574 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
9577 @defmac JMP_BUF_SIZE
9578 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
9579 defined. Define this macro if the default size of @code{jmp_buf} buffer
9580 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
9581 is not large enough, or if it is much too large.
9582 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
9585 @defmac DWARF_CIE_DATA_ALIGNMENT
9586 This macro need only be defined if the target might save registers in the
9587 function prologue at an offset to the stack pointer that is not aligned to
9588 @code{UNITS_PER_WORD}. The definition should be the negative minimum
9589 alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive
9590 minimum alignment otherwise. @xref{DWARF}. Only applicable if
9591 the target supports DWARF 2 frame unwind information.
9594 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
9595 Contains the value true if the target should add a zero word onto the
9596 end of a Dwarf-2 frame info section when used for exception handling.
9597 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
9601 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
9602 Given a register, this hook should return a parallel of registers to
9603 represent where to find the register pieces. Define this hook if the
9604 register and its mode are represented in Dwarf in non-contiguous
9605 locations, or if the register should be represented in more than one
9606 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
9607 If not defined, the default is to return @code{NULL_RTX}.
9610 @deftypefn {Target Hook} machine_mode TARGET_DWARF_FRAME_REG_MODE (int @var{regno})
9611 Given a register, this hook should return the mode which the
9612 corresponding Dwarf frame register should have. This is normally
9613 used to return a smaller mode than the raw mode to prevent call
9614 clobbered parts of a register altering the frame register size
9617 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
9618 If some registers are represented in Dwarf-2 unwind information in
9619 multiple pieces, define this hook to fill in information about the
9620 sizes of those pieces in the table used by the unwinder at runtime.
9621 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
9622 filling in a single size corresponding to each hard register;
9623 @var{address} is the address of the table.
9626 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
9627 This hook is used to output a reference from a frame unwinding table to
9628 the type_info object identified by @var{sym}. It should return @code{true}
9629 if the reference was output. Returning @code{false} will cause the
9630 reference to be output using the normal Dwarf2 routines.
9633 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
9634 This flag should be set to @code{true} on targets that use an ARM EABI
9635 based unwinding library, and @code{false} on other targets. This effects
9636 the format of unwinding tables, and how the unwinder in entered after
9637 running a cleanup. The default is @code{false}.
9640 @node Alignment Output
9641 @subsection Assembler Commands for Alignment
9643 @c prevent bad page break with this line
9644 This describes commands for alignment.
9646 @defmac JUMP_ALIGN (@var{label})
9647 The alignment (log base 2) to put in front of @var{label}, which is
9648 a common destination of jumps and has no fallthru incoming edge.
9650 This macro need not be defined if you don't want any special alignment
9651 to be done at such a time. Most machine descriptions do not currently
9654 Unless it's necessary to inspect the @var{label} parameter, it is better
9655 to set the variable @var{align_jumps} in the target's
9656 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9657 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
9660 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
9661 The alignment (log base 2) to put in front of @var{label}, which follows
9664 This macro need not be defined if you don't want any special alignment
9665 to be done at such a time. Most machine descriptions do not currently
9669 @defmac LOOP_ALIGN (@var{label})
9670 The alignment (log base 2) to put in front of @var{label} that heads
9671 a frequently executed basic block (usually the header of a loop).
9673 This macro need not be defined if you don't want any special alignment
9674 to be done at such a time. Most machine descriptions do not currently
9677 Unless it's necessary to inspect the @var{label} parameter, it is better
9678 to set the variable @code{align_loops} in the target's
9679 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9680 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
9683 @defmac LABEL_ALIGN (@var{label})
9684 The alignment (log base 2) to put in front of @var{label}.
9685 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
9686 the maximum of the specified values is used.
9688 Unless it's necessary to inspect the @var{label} parameter, it is better
9689 to set the variable @code{align_labels} in the target's
9690 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9691 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9694 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9695 A C statement to output to the stdio stream @var{stream} an assembler
9696 instruction to advance the location counter by @var{nbytes} bytes.
9697 Those bytes should be zero when loaded. @var{nbytes} will be a C
9698 expression of type @code{unsigned HOST_WIDE_INT}.
9701 @defmac ASM_NO_SKIP_IN_TEXT
9702 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9703 text section because it fails to put zeros in the bytes that are skipped.
9704 This is true on many Unix systems, where the pseudo--op to skip bytes
9705 produces no-op instructions rather than zeros when used in the text
9709 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9710 A C statement to output to the stdio stream @var{stream} an assembler
9711 command to advance the location counter to a multiple of 2 to the
9712 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9715 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9716 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9717 for padding, if necessary.
9720 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9721 A C statement to output to the stdio stream @var{stream} an assembler
9722 command to advance the location counter to a multiple of 2 to the
9723 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9724 satisfy the alignment request. @var{power} and @var{max_skip} will be
9725 a C expression of type @code{int}.
9729 @node Debugging Info
9730 @section Controlling Debugging Information Format
9732 @c prevent bad page break with this line
9733 This describes how to specify debugging information.
9736 * All Debuggers:: Macros that affect all debugging formats uniformly.
9737 * DBX Options:: Macros enabling specific options in DBX format.
9738 * DBX Hooks:: Hook macros for varying DBX format.
9739 * File Names and DBX:: Macros controlling output of file names in DBX format.
9740 * DWARF:: Macros for DWARF format.
9741 * VMS Debug:: Macros for VMS debug format.
9745 @subsection Macros Affecting All Debugging Formats
9747 @c prevent bad page break with this line
9748 These macros affect all debugging formats.
9750 @defmac DBX_REGISTER_NUMBER (@var{regno})
9751 A C expression that returns the DBX register number for the compiler
9752 register number @var{regno}. In the default macro provided, the value
9753 of this expression will be @var{regno} itself. But sometimes there are
9754 some registers that the compiler knows about and DBX does not, or vice
9755 versa. In such cases, some register may need to have one number in the
9756 compiler and another for DBX@.
9758 If two registers have consecutive numbers inside GCC, and they can be
9759 used as a pair to hold a multiword value, then they @emph{must} have
9760 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9761 Otherwise, debuggers will be unable to access such a pair, because they
9762 expect register pairs to be consecutive in their own numbering scheme.
9764 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9765 does not preserve register pairs, then what you must do instead is
9766 redefine the actual register numbering scheme.
9769 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9770 A C expression that returns the integer offset value for an automatic
9771 variable having address @var{x} (an RTL expression). The default
9772 computation assumes that @var{x} is based on the frame-pointer and
9773 gives the offset from the frame-pointer. This is required for targets
9774 that produce debugging output for DBX and allow the frame-pointer to be
9775 eliminated when the @option{-g} option is used.
9778 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9779 A C expression that returns the integer offset value for an argument
9780 having address @var{x} (an RTL expression). The nominal offset is
9784 @defmac PREFERRED_DEBUGGING_TYPE
9785 A C expression that returns the type of debugging output GCC should
9786 produce when the user specifies just @option{-g}. Define
9787 this if you have arranged for GCC to support more than one format of
9788 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9789 @code{DWARF2_DEBUG}, @code{XCOFF_DEBUG}, @code{VMS_DEBUG},
9790 and @code{VMS_AND_DWARF2_DEBUG}.
9792 When the user specifies @option{-ggdb}, GCC normally also uses the
9793 value of this macro to select the debugging output format, but with two
9794 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9795 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9796 defined, GCC uses @code{DBX_DEBUG}.
9798 The value of this macro only affects the default debugging output; the
9799 user can always get a specific type of output by using @option{-gstabs},
9800 @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9804 @subsection Specific Options for DBX Output
9806 @c prevent bad page break with this line
9807 These are specific options for DBX output.
9809 @defmac DBX_DEBUGGING_INFO
9810 Define this macro if GCC should produce debugging output for DBX
9811 in response to the @option{-g} option.
9814 @defmac XCOFF_DEBUGGING_INFO
9815 Define this macro if GCC should produce XCOFF format debugging output
9816 in response to the @option{-g} option. This is a variant of DBX format.
9819 @defmac DEFAULT_GDB_EXTENSIONS
9820 Define this macro to control whether GCC should by default generate
9821 GDB's extended version of DBX debugging information (assuming DBX-format
9822 debugging information is enabled at all). If you don't define the
9823 macro, the default is 1: always generate the extended information
9824 if there is any occasion to.
9827 @defmac DEBUG_SYMS_TEXT
9828 Define this macro if all @code{.stabs} commands should be output while
9829 in the text section.
9832 @defmac ASM_STABS_OP
9833 A C string constant, including spacing, naming the assembler pseudo op to
9834 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9835 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9836 applies only to DBX debugging information format.
9839 @defmac ASM_STABD_OP
9840 A C string constant, including spacing, naming the assembler pseudo op to
9841 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9842 value is the current location. If you don't define this macro,
9843 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9847 @defmac ASM_STABN_OP
9848 A C string constant, including spacing, naming the assembler pseudo op to
9849 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9850 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9851 macro applies only to DBX debugging information format.
9854 @defmac DBX_NO_XREFS
9855 Define this macro if DBX on your system does not support the construct
9856 @samp{xs@var{tagname}}. On some systems, this construct is used to
9857 describe a forward reference to a structure named @var{tagname}.
9858 On other systems, this construct is not supported at all.
9861 @defmac DBX_CONTIN_LENGTH
9862 A symbol name in DBX-format debugging information is normally
9863 continued (split into two separate @code{.stabs} directives) when it
9864 exceeds a certain length (by default, 80 characters). On some
9865 operating systems, DBX requires this splitting; on others, splitting
9866 must not be done. You can inhibit splitting by defining this macro
9867 with the value zero. You can override the default splitting-length by
9868 defining this macro as an expression for the length you desire.
9871 @defmac DBX_CONTIN_CHAR
9872 Normally continuation is indicated by adding a @samp{\} character to
9873 the end of a @code{.stabs} string when a continuation follows. To use
9874 a different character instead, define this macro as a character
9875 constant for the character you want to use. Do not define this macro
9876 if backslash is correct for your system.
9879 @defmac DBX_STATIC_STAB_DATA_SECTION
9880 Define this macro if it is necessary to go to the data section before
9881 outputting the @samp{.stabs} pseudo-op for a non-global static
9885 @defmac DBX_TYPE_DECL_STABS_CODE
9886 The value to use in the ``code'' field of the @code{.stabs} directive
9887 for a typedef. The default is @code{N_LSYM}.
9890 @defmac DBX_STATIC_CONST_VAR_CODE
9891 The value to use in the ``code'' field of the @code{.stabs} directive
9892 for a static variable located in the text section. DBX format does not
9893 provide any ``right'' way to do this. The default is @code{N_FUN}.
9896 @defmac DBX_REGPARM_STABS_CODE
9897 The value to use in the ``code'' field of the @code{.stabs} directive
9898 for a parameter passed in registers. DBX format does not provide any
9899 ``right'' way to do this. The default is @code{N_RSYM}.
9902 @defmac DBX_REGPARM_STABS_LETTER
9903 The letter to use in DBX symbol data to identify a symbol as a parameter
9904 passed in registers. DBX format does not customarily provide any way to
9905 do this. The default is @code{'P'}.
9908 @defmac DBX_FUNCTION_FIRST
9909 Define this macro if the DBX information for a function and its
9910 arguments should precede the assembler code for the function. Normally,
9911 in DBX format, the debugging information entirely follows the assembler
9915 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9916 Define this macro, with value 1, if the value of a symbol describing
9917 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9918 relative to the start of the enclosing function. Normally, GCC uses
9919 an absolute address.
9922 @defmac DBX_LINES_FUNCTION_RELATIVE
9923 Define this macro, with value 1, if the value of a symbol indicating
9924 the current line number (@code{N_SLINE}) should be relative to the
9925 start of the enclosing function. Normally, GCC uses an absolute address.
9928 @defmac DBX_USE_BINCL
9929 Define this macro if GCC should generate @code{N_BINCL} and
9930 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9931 macro also directs GCC to output a type number as a pair of a file
9932 number and a type number within the file. Normally, GCC does not
9933 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9934 number for a type number.
9938 @subsection Open-Ended Hooks for DBX Format
9940 @c prevent bad page break with this line
9941 These are hooks for DBX format.
9943 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9944 A C statement to output DBX debugging information before code for line
9945 number @var{line} of the current source file to the stdio stream
9946 @var{stream}. @var{counter} is the number of time the macro was
9947 invoked, including the current invocation; it is intended to generate
9948 unique labels in the assembly output.
9950 This macro should not be defined if the default output is correct, or
9951 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9954 @defmac NO_DBX_FUNCTION_END
9955 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9956 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9957 On those machines, define this macro to turn this feature off without
9958 disturbing the rest of the gdb extensions.
9961 @defmac NO_DBX_BNSYM_ENSYM
9962 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9963 extension construct. On those machines, define this macro to turn this
9964 feature off without disturbing the rest of the gdb extensions.
9967 @node File Names and DBX
9968 @subsection File Names in DBX Format
9970 @c prevent bad page break with this line
9971 This describes file names in DBX format.
9973 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9974 A C statement to output DBX debugging information to the stdio stream
9975 @var{stream}, which indicates that file @var{name} is the main source
9976 file---the file specified as the input file for compilation.
9977 This macro is called only once, at the beginning of compilation.
9979 This macro need not be defined if the standard form of output
9980 for DBX debugging information is appropriate.
9982 It may be necessary to refer to a label equal to the beginning of the
9983 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9984 to do so. If you do this, you must also set the variable
9985 @var{used_ltext_label_name} to @code{true}.
9988 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9989 Define this macro, with value 1, if GCC should not emit an indication
9990 of the current directory for compilation and current source language at
9991 the beginning of the file.
9994 @defmac NO_DBX_GCC_MARKER
9995 Define this macro, with value 1, if GCC should not emit an indication
9996 that this object file was compiled by GCC@. The default is to emit
9997 an @code{N_OPT} stab at the beginning of every source file, with
9998 @samp{gcc2_compiled.} for the string and value 0.
10001 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
10002 A C statement to output DBX debugging information at the end of
10003 compilation of the main source file @var{name}. Output should be
10004 written to the stdio stream @var{stream}.
10006 If you don't define this macro, nothing special is output at the end
10007 of compilation, which is correct for most machines.
10010 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
10011 Define this macro @emph{instead of} defining
10012 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
10013 the end of compilation is an @code{N_SO} stab with an empty string,
10014 whose value is the highest absolute text address in the file.
10019 @subsection Macros for DWARF Output
10021 @c prevent bad page break with this line
10022 Here are macros for DWARF output.
10024 @defmac DWARF2_DEBUGGING_INFO
10025 Define this macro if GCC should produce dwarf version 2 format
10026 debugging output in response to the @option{-g} option.
10028 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
10029 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
10030 be emitted for each function. Instead of an integer return the enum
10031 value for the @code{DW_CC_} tag.
10034 To support optional call frame debugging information, you must also
10035 define @code{INCOMING_RETURN_ADDR_RTX} and either set
10036 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
10037 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
10038 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
10041 @defmac DWARF2_FRAME_INFO
10042 Define this macro to a nonzero value if GCC should always output
10043 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
10044 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
10045 exceptions are enabled, GCC will output this information not matter
10046 how you define @code{DWARF2_FRAME_INFO}.
10049 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
10050 This hook defines the mechanism that will be used for describing frame
10051 unwind information to the debugger. Normally the hook will return
10052 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
10053 return @code{UI_NONE} otherwise.
10055 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
10056 is disabled in order to always output DWARF 2 frame information.
10058 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
10059 This will suppress generation of the normal debug frame unwind information.
10062 @defmac DWARF2_ASM_LINE_DEBUG_INFO
10063 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
10064 line debug info sections. This will result in much more compact line number
10065 tables, and hence is desirable if it works.
10068 @defmac DWARF2_ASM_VIEW_DEBUG_INFO
10069 Define this macro to be a nonzero value if the assembler supports view
10070 assignment and verification in @code{.loc}. If it does not, but the
10071 user enables location views, the compiler may have to fallback to
10072 internal line number tables.
10075 @deftypefn {Target Hook} int TARGET_RESET_LOCATION_VIEW (rtx_insn *@var{})
10076 This hook, if defined, enables -ginternal-reset-location-views, and
10077 uses its result to override cases in which the estimated min insn
10078 length might be nonzero even when a PC advance (i.e., a view reset)
10079 cannot be taken for granted.
10081 If the hook is defined, it must return a positive value to indicate
10082 the insn definitely advances the PC, and so the view number can be
10083 safely assumed to be reset; a negative value to mean the insn
10084 definitely does not advance the PC, and os the view number must not
10085 be reset; or zero to decide based on the estimated insn length.
10087 If insn length is to be regarded as reliable, set the hook to
10088 @code{hook_int_rtx_insn_0}.
10091 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
10092 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.
10095 @deftypevr {Target Hook} bool TARGET_DELAY_SCHED2
10096 True if sched2 is not to be run at its normal place.
10097 This usually means it will be run as part of machine-specific reorg.
10100 @deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK
10101 True if vartrack is not to be run at its normal place.
10102 This usually means it will be run as part of machine-specific reorg.
10105 @deftypevr {Target Hook} bool TARGET_NO_REGISTER_ALLOCATION
10106 True if register allocation and the passes
10107 following it should not be run. Usually true only for virtual assembler
10111 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
10112 A C statement to issue assembly directives that create a difference
10113 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
10116 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
10117 A C statement to issue assembly directives that create a difference
10118 between the two given labels in system defined units, e.g.@: instruction
10119 slots on IA64 VMS, using an integer of the given size.
10122 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{offset}, @var{section})
10123 A C statement to issue assembly directives that create a
10124 section-relative reference to the given @var{label} plus @var{offset}, using
10125 an integer of the given @var{size}. The label is known to be defined in the
10126 given @var{section}.
10129 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
10130 A C statement to issue assembly directives that create a self-relative
10131 reference to the given @var{label}, using an integer of the given @var{size}.
10134 @defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label})
10135 A C statement to issue assembly directives that create a reference to the
10136 given @var{label} relative to the dbase, using an integer of the given @var{size}.
10139 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
10140 A C statement to issue assembly directives that create a reference to
10141 the DWARF table identifier @var{label} from the current section. This
10142 is used on some systems to avoid garbage collecting a DWARF table which
10143 is referenced by a function.
10146 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
10147 If defined, this target hook is a function which outputs a DTP-relative
10148 reference to the given TLS symbol of the specified size.
10153 @subsection Macros for VMS Debug Format
10155 @c prevent bad page break with this line
10156 Here are macros for VMS debug format.
10158 @defmac VMS_DEBUGGING_INFO
10159 Define this macro if GCC should produce debugging output for VMS
10160 in response to the @option{-g} option. The default behavior for VMS
10161 is to generate minimal debug info for a traceback in the absence of
10162 @option{-g} unless explicitly overridden with @option{-g0}. This
10163 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
10164 @code{TARGET_OPTION_OVERRIDE}.
10167 @node Floating Point
10168 @section Cross Compilation and Floating Point
10169 @cindex cross compilation and floating point
10170 @cindex floating point and cross compilation
10172 While all modern machines use twos-complement representation for integers,
10173 there are a variety of representations for floating point numbers. This
10174 means that in a cross-compiler the representation of floating point numbers
10175 in the compiled program may be different from that used in the machine
10176 doing the compilation.
10178 Because different representation systems may offer different amounts of
10179 range and precision, all floating point constants must be represented in
10180 the target machine's format. Therefore, the cross compiler cannot
10181 safely use the host machine's floating point arithmetic; it must emulate
10182 the target's arithmetic. To ensure consistency, GCC always uses
10183 emulation to work with floating point values, even when the host and
10184 target floating point formats are identical.
10186 The following macros are provided by @file{real.h} for the compiler to
10187 use. All parts of the compiler which generate or optimize
10188 floating-point calculations must use these macros. They may evaluate
10189 their operands more than once, so operands must not have side effects.
10191 @defmac REAL_VALUE_TYPE
10192 The C data type to be used to hold a floating point value in the target
10193 machine's format. Typically this is a @code{struct} containing an
10194 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
10198 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
10199 Truncates @var{x} to a signed integer, rounding toward zero.
10202 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
10203 Truncates @var{x} to an unsigned integer, rounding toward zero. If
10204 @var{x} is negative, returns zero.
10207 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode})
10208 Converts @var{string} into a floating point number in the target machine's
10209 representation for mode @var{mode}. This routine can handle both
10210 decimal and hexadecimal floating point constants, using the syntax
10211 defined by the C language for both.
10214 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
10215 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
10218 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
10219 Determines whether @var{x} represents infinity (positive or negative).
10222 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
10223 Determines whether @var{x} represents a ``NaN'' (not-a-number).
10226 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
10227 Returns the negative of the floating point value @var{x}.
10230 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
10231 Returns the absolute value of @var{x}.
10234 @node Mode Switching
10235 @section Mode Switching Instructions
10236 @cindex mode switching
10237 The following macros control mode switching optimizations:
10239 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
10240 Define this macro if the port needs extra instructions inserted for mode
10241 switching in an optimizing compilation.
10243 For an example, the SH4 can perform both single and double precision
10244 floating point operations, but to perform a single precision operation,
10245 the FPSCR PR bit has to be cleared, while for a double precision
10246 operation, this bit has to be set. Changing the PR bit requires a general
10247 purpose register as a scratch register, hence these FPSCR sets have to
10248 be inserted before reload, i.e.@: you cannot put this into instruction emitting
10249 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
10251 You can have multiple entities that are mode-switched, and select at run time
10252 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
10253 return nonzero for any @var{entity} that needs mode-switching.
10254 If you define this macro, you also have to define
10255 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
10256 @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
10257 @code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
10261 @defmac NUM_MODES_FOR_MODE_SWITCHING
10262 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
10263 initializer for an array of integers. Each initializer element
10264 N refers to an entity that needs mode switching, and specifies the number
10265 of different modes that might need to be set for this entity.
10266 The position of the initializer in the initializer---starting counting at
10267 zero---determines the integer that is used to refer to the mode-switched
10268 entity in question.
10269 In macros that take mode arguments / yield a mode result, modes are
10270 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
10271 switch is needed / supplied.
10274 @deftypefn {Target Hook} void TARGET_MODE_EMIT (int @var{entity}, int @var{mode}, int @var{prev_mode}, HARD_REG_SET @var{regs_live})
10275 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.
10278 @deftypefn {Target Hook} int TARGET_MODE_NEEDED (int @var{entity}, rtx_insn *@var{insn})
10279 @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}.
10282 @deftypefn {Target Hook} int TARGET_MODE_AFTER (int @var{entity}, int @var{mode}, rtx_insn *@var{insn})
10283 @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).
10286 @deftypefn {Target Hook} int TARGET_MODE_ENTRY (int @var{entity})
10287 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.
10290 @deftypefn {Target Hook} int TARGET_MODE_EXIT (int @var{entity})
10291 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.
10294 @deftypefn {Target Hook} int TARGET_MODE_PRIORITY (int @var{entity}, int @var{n})
10295 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}.
10298 @node Target Attributes
10299 @section Defining target-specific uses of @code{__attribute__}
10300 @cindex target attributes
10301 @cindex machine attributes
10302 @cindex attributes, target-specific
10304 Target-specific attributes may be defined for functions, data and types.
10305 These are described using the following target hooks; they also need to
10306 be documented in @file{extend.texi}.
10308 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
10309 If defined, this target hook points to an array of @samp{struct
10310 attribute_spec} (defined in @file{tree-core.h}) specifying the machine
10311 specific attributes for this target and some of the restrictions on the
10312 entities to which these attributes are applied and the arguments they
10316 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
10317 If defined, this target hook is a function which returns true if the
10318 machine-specific attribute named @var{name} expects an identifier
10319 given as its first argument to be passed on as a plain identifier, not
10320 subjected to name lookup. If this is not defined, the default is
10321 false for all machine-specific attributes.
10324 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
10325 If defined, this target hook is a function which returns zero if the attributes on
10326 @var{type1} and @var{type2} are incompatible, one if they are compatible,
10327 and two if they are nearly compatible (which causes a warning to be
10328 generated). If this is not defined, machine-specific attributes are
10329 supposed always to be compatible.
10332 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
10333 If defined, this target hook is a function which assigns default attributes to
10334 the newly defined @var{type}.
10337 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
10338 Define this target hook if the merging of type attributes needs special
10339 handling. If defined, the result is a list of the combined
10340 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
10341 that @code{comptypes} has already been called and returned 1. This
10342 function may call @code{merge_attributes} to handle machine-independent
10346 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
10347 Define this target hook if the merging of decl attributes needs special
10348 handling. If defined, the result is a list of the combined
10349 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
10350 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
10351 when this is needed are when one attribute overrides another, or when an
10352 attribute is nullified by a subsequent definition. This function may
10353 call @code{merge_attributes} to handle machine-independent merging.
10355 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
10356 If the only target-specific handling you require is @samp{dllimport}
10357 for Microsoft Windows targets, you should define the macro
10358 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
10359 will then define a function called
10360 @code{merge_dllimport_decl_attributes} which can then be defined as
10361 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
10362 add @code{handle_dll_attribute} in the attribute table for your port
10363 to perform initial processing of the @samp{dllimport} and
10364 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
10365 @file{i386/i386.c}, for example.
10368 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
10369 @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}.
10372 @defmac TARGET_DECLSPEC
10373 Define this macro to a nonzero value if you want to treat
10374 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
10375 default, this behavior is enabled only for targets that define
10376 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
10377 of @code{__declspec} is via a built-in macro, but you should not rely
10378 on this implementation detail.
10381 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
10382 Define this target hook if you want to be able to add attributes to a decl
10383 when it is being created. This is normally useful for back ends which
10384 wish to implement a pragma by using the attributes which correspond to
10385 the pragma's effect. The @var{node} argument is the decl which is being
10386 created. The @var{attr_ptr} argument is a pointer to the attribute list
10387 for this decl. The list itself should not be modified, since it may be
10388 shared with other decls, but attributes may be chained on the head of
10389 the list and @code{*@var{attr_ptr}} modified to point to the new
10390 attributes, or a copy of the list may be made if further changes are
10394 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
10396 This target hook returns @code{true} if it is OK to inline @var{fndecl}
10397 into the current function, despite its having target-specific
10398 attributes, @code{false} otherwise. By default, if a function has a
10399 target specific attribute attached to it, it will not be inlined.
10402 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
10403 This hook is called to parse @code{attribute(target("..."))}, which
10404 allows setting target-specific options on individual functions.
10405 These function-specific options may differ
10406 from the options specified on the command line. The hook should return
10407 @code{true} if the options are valid.
10409 The hook should set the @code{DECL_FUNCTION_SPECIFIC_TARGET} field in
10410 the function declaration to hold a pointer to a target-specific
10411 @code{struct cl_target_option} structure.
10414 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr}, struct gcc_options *@var{opts})
10415 This hook is called to save any additional target-specific information
10416 in the @code{struct cl_target_option} structure for function-specific
10417 options from the @code{struct gcc_options} structure.
10418 @xref{Option file format}.
10421 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct gcc_options *@var{opts}, struct cl_target_option *@var{ptr})
10422 This hook is called to restore any additional target-specific
10423 information in the @code{struct cl_target_option} structure for
10424 function-specific options to the @code{struct gcc_options} structure.
10427 @deftypefn {Target Hook} void TARGET_OPTION_POST_STREAM_IN (struct cl_target_option *@var{ptr})
10428 This hook is called to update target-specific information in the
10429 @code{struct cl_target_option} structure after it is streamed in from
10433 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
10434 This hook is called to print any additional target-specific
10435 information in the @code{struct cl_target_option} structure for
10436 function-specific options.
10439 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
10440 This target hook parses the options for @code{#pragma GCC target}, which
10441 sets the target-specific options for functions that occur later in the
10442 input stream. The options accepted should be the same as those handled by the
10443 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
10446 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
10447 Sometimes certain combinations of command options do not make sense on
10448 a particular target machine. You can override the hook
10449 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
10450 once just after all the command options have been parsed.
10452 Don't use this hook to turn on various extra optimizations for
10453 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
10455 If you need to do something whenever the optimization level is
10456 changed via the optimize attribute or pragma, see
10457 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
10460 @deftypefn {Target Hook} bool TARGET_OPTION_FUNCTION_VERSIONS (tree @var{decl1}, tree @var{decl2})
10461 This target hook returns @code{true} if @var{DECL1} and @var{DECL2} are
10462 versions of the same function. @var{DECL1} and @var{DECL2} are function
10463 versions if and only if they have the same function signature and
10464 different target specific attributes, that is, they are compiled for
10465 different target machines.
10468 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
10469 This target hook returns @code{false} if the @var{caller} function
10470 cannot inline @var{callee}, based on target specific information. By
10471 default, inlining is not allowed if the callee function has function
10472 specific target options and the caller does not use the same options.
10475 @deftypefn {Target Hook} void TARGET_RELAYOUT_FUNCTION (tree @var{fndecl})
10476 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.
10480 @section Emulating TLS
10481 @cindex Emulated TLS
10483 For targets whose psABI does not provide Thread Local Storage via
10484 specific relocations and instruction sequences, an emulation layer is
10485 used. A set of target hooks allows this emulation layer to be
10486 configured for the requirements of a particular target. For instance
10487 the psABI may in fact specify TLS support in terms of an emulation
10490 The emulation layer works by creating a control object for every TLS
10491 object. To access the TLS object, a lookup function is provided
10492 which, when given the address of the control object, will return the
10493 address of the current thread's instance of the TLS object.
10495 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
10496 Contains the name of the helper function that uses a TLS control
10497 object to locate a TLS instance. The default causes libgcc's
10498 emulated TLS helper function to be used.
10501 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
10502 Contains the name of the helper function that should be used at
10503 program startup to register TLS objects that are implicitly
10504 initialized to zero. If this is @code{NULL}, all TLS objects will
10505 have explicit initializers. The default causes libgcc's emulated TLS
10506 registration function to be used.
10509 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
10510 Contains the name of the section in which TLS control variables should
10511 be placed. The default of @code{NULL} allows these to be placed in
10515 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
10516 Contains the name of the section in which TLS initializers should be
10517 placed. The default of @code{NULL} allows these to be placed in any
10521 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
10522 Contains the prefix to be prepended to TLS control variable names.
10523 The default of @code{NULL} uses a target-specific prefix.
10526 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
10527 Contains the prefix to be prepended to TLS initializer objects. The
10528 default of @code{NULL} uses a target-specific prefix.
10531 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
10532 Specifies a function that generates the FIELD_DECLs for a TLS control
10533 object type. @var{type} is the RECORD_TYPE the fields are for and
10534 @var{name} should be filled with the structure tag, if the default of
10535 @code{__emutls_object} is unsuitable. The default creates a type suitable
10536 for libgcc's emulated TLS function.
10539 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
10540 Specifies a function that generates the CONSTRUCTOR to initialize a
10541 TLS control object. @var{var} is the TLS control object, @var{decl}
10542 is the TLS object and @var{tmpl_addr} is the address of the
10543 initializer. The default initializes libgcc's emulated TLS control object.
10546 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
10547 Specifies whether the alignment of TLS control variable objects is
10548 fixed and should not be increased as some backends may do to optimize
10549 single objects. The default is false.
10552 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
10553 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
10554 may be used to describe emulated TLS control objects.
10557 @node MIPS Coprocessors
10558 @section Defining coprocessor specifics for MIPS targets.
10559 @cindex MIPS coprocessor-definition macros
10561 The MIPS specification allows MIPS implementations to have as many as 4
10562 coprocessors, each with as many as 32 private registers. GCC supports
10563 accessing these registers and transferring values between the registers
10564 and memory using asm-ized variables. For example:
10567 register unsigned int cp0count asm ("c0r1");
10573 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
10574 names may be added as described below, or the default names may be
10575 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
10577 Coprocessor registers are assumed to be epilogue-used; sets to them will
10578 be preserved even if it does not appear that the register is used again
10579 later in the function.
10581 Another note: according to the MIPS spec, coprocessor 1 (if present) is
10582 the FPU@. One accesses COP1 registers through standard mips
10583 floating-point support; they are not included in this mechanism.
10586 @section Parameters for Precompiled Header Validity Checking
10587 @cindex parameters, precompiled headers
10589 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
10590 This hook returns a pointer to the data needed by
10591 @code{TARGET_PCH_VALID_P} and sets
10592 @samp{*@var{sz}} to the size of the data in bytes.
10595 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
10596 This hook checks whether the options used to create a PCH file are
10597 compatible with the current settings. It returns @code{NULL}
10598 if so and a suitable error message if not. Error messages will
10599 be presented to the user and must be localized using @samp{_(@var{msg})}.
10601 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
10602 when the PCH file was created and @var{sz} is the size of that data in bytes.
10603 It's safe to assume that the data was created by the same version of the
10604 compiler, so no format checking is needed.
10606 The default definition of @code{default_pch_valid_p} should be
10607 suitable for most targets.
10610 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
10611 If this hook is nonnull, the default implementation of
10612 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
10613 of @code{target_flags}. @var{pch_flags} specifies the value that
10614 @code{target_flags} had when the PCH file was created. The return
10615 value is the same as for @code{TARGET_PCH_VALID_P}.
10618 @deftypefn {Target Hook} void TARGET_PREPARE_PCH_SAVE (void)
10619 Called before writing out a PCH file. If the target has some
10620 garbage-collected data that needs to be in a particular state on PCH loads,
10621 it can use this hook to enforce that state. Very few targets need
10622 to do anything here.
10626 @section C++ ABI parameters
10627 @cindex parameters, c++ abi
10629 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10630 Define this hook to override the integer type used for guard variables.
10631 These are used to implement one-time construction of static objects. The
10632 default is long_long_integer_type_node.
10635 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10636 This hook determines how guard variables are used. It should return
10637 @code{false} (the default) if the first byte should be used. A return value of
10638 @code{true} indicates that only the least significant bit should be used.
10641 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10642 This hook returns the size of the cookie to use when allocating an array
10643 whose elements have the indicated @var{type}. Assumes that it is already
10644 known that a cookie is needed. The default is
10645 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10646 IA64/Generic C++ ABI@.
10649 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10650 This hook should return @code{true} if the element size should be stored in
10651 array cookies. The default is to return @code{false}.
10654 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10655 If defined by a backend this hook allows the decision made to export
10656 class @var{type} to be overruled. Upon entry @var{import_export}
10657 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10658 to be imported and 0 otherwise. This function should return the
10659 modified value and perform any other actions necessary to support the
10660 backend's targeted operating system.
10663 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10664 This hook should return @code{true} if constructors and destructors return
10665 the address of the object created/destroyed. The default is to return
10669 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10670 This hook returns true if the key method for a class (i.e., the method
10671 which, if defined in the current translation unit, causes the virtual
10672 table to be emitted) may be an inline function. Under the standard
10673 Itanium C++ ABI the key method may be an inline function so long as
10674 the function is not declared inline in the class definition. Under
10675 some variants of the ABI, an inline function can never be the key
10676 method. The default is to return @code{true}.
10679 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10680 @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}.
10683 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10684 This hook returns true (the default) if virtual tables and other
10685 similar implicit class data objects are always COMDAT if they have
10686 external linkage. If this hook returns false, then class data for
10687 classes whose virtual table will be emitted in only one translation
10688 unit will not be COMDAT.
10691 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10692 This hook returns true (the default) if the RTTI information for
10693 the basic types which is defined in the C++ runtime should always
10694 be COMDAT, false if it should not be COMDAT.
10697 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10698 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10699 should be used to register static destructors when @option{-fuse-cxa-atexit}
10700 is in effect. The default is to return false to use @code{__cxa_atexit}.
10703 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10704 This hook returns true if the target @code{atexit} function can be used
10705 in the same manner as @code{__cxa_atexit} to register C++ static
10706 destructors. This requires that @code{atexit}-registered functions in
10707 shared libraries are run in the correct order when the libraries are
10708 unloaded. The default is to return false.
10711 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10712 @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).
10715 @deftypefn {Target Hook} tree TARGET_CXX_DECL_MANGLING_CONTEXT (const_tree @var{decl})
10716 Return target-specific mangling context of @var{decl} or @code{NULL_TREE}.
10719 @node D Language and ABI
10720 @section D ABI parameters
10721 @cindex parameters, d abi
10723 @deftypefn {D Target Hook} void TARGET_D_CPU_VERSIONS (void)
10724 Declare all environmental version identifiers relating to the target CPU
10725 using the function @code{builtin_version}, which takes a string representing
10726 the name of the version. Version identifiers predefined by this hook apply
10727 to all modules that are being compiled and imported.
10730 @deftypefn {D Target Hook} void TARGET_D_OS_VERSIONS (void)
10731 Similarly to @code{TARGET_D_CPU_VERSIONS}, but is used for versions
10732 relating to the target operating system.
10735 @deftypefn {D Target Hook} unsigned TARGET_D_CRITSEC_SIZE (void)
10736 Returns the size of the data structure used by the target operating system
10737 for critical sections and monitors. For example, on Microsoft Windows this
10738 would return the @code{sizeof(CRITICAL_SECTION)}, while other platforms that
10739 implement pthreads would return @code{sizeof(pthread_mutex_t)}.
10742 @node Named Address Spaces
10743 @section Adding support for named address spaces
10744 @cindex named address spaces
10746 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10747 standards committee, @cite{Programming Languages - C - Extensions to
10748 support embedded processors}, specifies a syntax for embedded
10749 processors to specify alternate address spaces. You can configure a
10750 GCC port to support section 5.1 of the draft report to add support for
10751 address spaces other than the default address space. These address
10752 spaces are new keywords that are similar to the @code{volatile} and
10753 @code{const} type attributes.
10755 Pointers to named address spaces can have a different size than
10756 pointers to the generic address space.
10758 For example, the SPU port uses the @code{__ea} address space to refer
10759 to memory in the host processor, rather than memory local to the SPU
10760 processor. Access to memory in the @code{__ea} address space involves
10761 issuing DMA operations to move data between the host processor and the
10762 local processor memory address space. Pointers in the @code{__ea}
10763 address space are either 32 bits or 64 bits based on the
10764 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10767 Internally, address spaces are represented as a small integer in the
10768 range 0 to 15 with address space 0 being reserved for the generic
10771 To register a named address space qualifier keyword with the C front end,
10772 the target may call the @code{c_register_addr_space} routine. For example,
10773 the SPU port uses the following to declare @code{__ea} as the keyword for
10774 named address space #1:
10776 #define ADDR_SPACE_EA 1
10777 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10780 @deftypefn {Target Hook} scalar_int_mode TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10781 Define this to return the machine mode to use for pointers to
10782 @var{address_space} if the target supports named address spaces.
10783 The default version of this hook returns @code{ptr_mode}.
10786 @deftypefn {Target Hook} scalar_int_mode TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10787 Define this to return the machine mode to use for addresses in
10788 @var{address_space} if the target supports named address spaces.
10789 The default version of this hook returns @code{Pmode}.
10792 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (scalar_int_mode @var{mode}, addr_space_t @var{as})
10793 Define this to return nonzero if the port can handle pointers
10794 with machine mode @var{mode} to address space @var{as}. This target
10795 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10796 except that it includes explicit named address space support. The default
10797 version of this hook returns true for the modes returned by either the
10798 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10799 target hooks for the given address space.
10802 @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})
10803 Define this to return true if @var{exp} is a valid address for mode
10804 @var{mode} in the named address space @var{as}. The @var{strict}
10805 parameter says whether strict addressing is in effect after reload has
10806 finished. This target hook is the same as the
10807 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10808 explicit named address space support.
10811 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode}, addr_space_t @var{as})
10812 Define this to modify an invalid address @var{x} to be a valid address
10813 with mode @var{mode} in the named address space @var{as}. This target
10814 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10815 except that it includes explicit named address space support.
10818 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{subset}, addr_space_t @var{superset})
10819 Define this to return whether the @var{subset} named address space is
10820 contained within the @var{superset} named address space. Pointers to
10821 a named address space that is a subset of another named address space
10822 will be converted automatically without a cast if used together in
10823 arithmetic operations. Pointers to a superset address space can be
10824 converted to pointers to a subset address space via explicit casts.
10827 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID (addr_space_t @var{as})
10828 Define this to modify the default handling of address 0 for the
10829 address space. Return true if 0 should be considered a valid address.
10832 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10833 Define this to convert the pointer expression represented by the RTL
10834 @var{op} with type @var{from_type} that points to a named address
10835 space to a new pointer expression with type @var{to_type} that points
10836 to a different named address space. When this hook it called, it is
10837 guaranteed that one of the two address spaces is a subset of the other,
10838 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10841 @deftypefn {Target Hook} int TARGET_ADDR_SPACE_DEBUG (addr_space_t @var{as})
10842 Define this to define how the address space is encoded in dwarf.
10843 The result is the value to be used with @code{DW_AT_address_class}.
10846 @deftypefn {Target Hook} void TARGET_ADDR_SPACE_DIAGNOSE_USAGE (addr_space_t @var{as}, location_t @var{loc})
10847 Define this hook if the availability of an address space depends on
10848 command line options and some diagnostics should be printed when the
10849 address space is used. This hook is called during parsing and allows
10850 to emit a better diagnostic compared to the case where the address space
10851 was not registered with @code{c_register_addr_space}. @var{as} is
10852 the address space as registered with @code{c_register_addr_space}.
10853 @var{loc} is the location of the address space qualifier token.
10854 The default implementation does nothing.
10858 @section Miscellaneous Parameters
10859 @cindex parameters, miscellaneous
10861 @c prevent bad page break with this line
10862 Here are several miscellaneous parameters.
10864 @defmac HAS_LONG_COND_BRANCH
10865 Define this boolean macro to indicate whether or not your architecture
10866 has conditional branches that can span all of memory. It is used in
10867 conjunction with an optimization that partitions hot and cold basic
10868 blocks into separate sections of the executable. If this macro is
10869 set to false, gcc will convert any conditional branches that attempt
10870 to cross between sections into unconditional branches or indirect jumps.
10873 @defmac HAS_LONG_UNCOND_BRANCH
10874 Define this boolean macro to indicate whether or not your architecture
10875 has unconditional branches that can span all of memory. It is used in
10876 conjunction with an optimization that partitions hot and cold basic
10877 blocks into separate sections of the executable. If this macro is
10878 set to false, gcc will convert any unconditional branches that attempt
10879 to cross between sections into indirect jumps.
10882 @defmac CASE_VECTOR_MODE
10883 An alias for a machine mode name. This is the machine mode that
10884 elements of a jump-table should have.
10887 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10888 Optional: return the preferred mode for an @code{addr_diff_vec}
10889 when the minimum and maximum offset are known. If you define this,
10890 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10891 To make this work, you also have to define @code{INSN_ALIGN} and
10892 make the alignment for @code{addr_diff_vec} explicit.
10893 The @var{body} argument is provided so that the offset_unsigned and scale
10894 flags can be updated.
10897 @defmac CASE_VECTOR_PC_RELATIVE
10898 Define this macro to be a C expression to indicate when jump-tables
10899 should contain relative addresses. You need not define this macro if
10900 jump-tables never contain relative addresses, or jump-tables should
10901 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10905 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10906 This function return the smallest number of different values for which it
10907 is best to use a jump-table instead of a tree of conditional branches.
10908 The default is four for machines with a @code{casesi} instruction and
10909 five otherwise. This is best for most machines.
10912 @defmac WORD_REGISTER_OPERATIONS
10913 Define this macro to 1 if operations between registers with integral mode
10914 smaller than a word are always performed on the entire register. To be
10915 more explicit, if you start with a pair of @code{word_mode} registers with
10916 known values and you do a subword, for example @code{QImode}, addition on
10917 the low part of the registers, then the compiler may consider that the
10918 result has a known value in @code{word_mode} too if the macro is defined
10919 to 1. Most RISC machines have this property and most CISC machines do not.
10922 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_ARITHMETIC_PRECISION (void)
10923 On some RISC architectures with 64-bit registers, the processor also
10924 maintains 32-bit condition codes that make it possible to do real 32-bit
10925 arithmetic, although the operations are performed on the full registers.
10927 On such architectures, defining this hook to 32 tells the compiler to try
10928 using 32-bit arithmetical operations setting the condition codes instead
10929 of doing full 64-bit arithmetic.
10931 More generally, define this hook on RISC architectures if you want the
10932 compiler to try using arithmetical operations setting the condition codes
10933 with a precision lower than the word precision.
10935 You need not define this hook if @code{WORD_REGISTER_OPERATIONS} is not
10939 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10940 Define this macro to be a C expression indicating when insns that read
10941 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10942 bits outside of @var{mem_mode} to be either the sign-extension or the
10943 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10944 of @var{mem_mode} for which the
10945 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10946 @code{UNKNOWN} for other modes.
10948 This macro is not called with @var{mem_mode} non-integral or with a width
10949 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10950 value in this case. Do not define this macro if it would always return
10951 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10952 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10954 You may return a non-@code{UNKNOWN} value even if for some hard registers
10955 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10956 of these hard registers @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false
10957 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10958 integral mode larger than this but not larger than @code{word_mode}.
10960 You must return @code{UNKNOWN} if for some hard registers that allow this
10961 mode, @code{TARGET_CAN_CHANGE_MODE_CLASS} says that they cannot change to
10962 @code{word_mode}, but that they can change to another integral mode that
10963 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10966 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10967 Define this macro to 1 if loading short immediate values into registers sign
10971 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (machine_mode @var{mode})
10972 When @option{-ffast-math} is in effect, GCC tries to optimize
10973 divisions by the same divisor, by turning them into multiplications by
10974 the reciprocal. This target hook specifies the minimum number of divisions
10975 that should be there for GCC to perform the optimization for a variable
10976 of mode @var{mode}. The default implementation returns 3 if the machine
10977 has an instruction for the division, and 2 if it does not.
10981 The maximum number of bytes that a single instruction can move quickly
10982 between memory and registers or between two memory locations.
10985 @defmac MAX_MOVE_MAX
10986 The maximum number of bytes that a single instruction can move quickly
10987 between memory and registers or between two memory locations. If this
10988 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10989 constant value that is the largest value that @code{MOVE_MAX} can have
10993 @defmac SHIFT_COUNT_TRUNCATED
10994 A C expression that is nonzero if on this machine the number of bits
10995 actually used for the count of a shift operation is equal to the number
10996 of bits needed to represent the size of the object being shifted. When
10997 this macro is nonzero, the compiler will assume that it is safe to omit
10998 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10999 truncates the count of a shift operation. On machines that have
11000 instructions that act on bit-fields at variable positions, which may
11001 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
11002 also enables deletion of truncations of the values that serve as
11003 arguments to bit-field instructions.
11005 If both types of instructions truncate the count (for shifts) and
11006 position (for bit-field operations), or if no variable-position bit-field
11007 instructions exist, you should define this macro.
11009 However, on some machines, such as the 80386 and the 680x0, truncation
11010 only applies to shift operations and not the (real or pretended)
11011 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
11012 such machines. Instead, add patterns to the @file{md} file that include
11013 the implied truncation of the shift instructions.
11015 You need not define this macro if it would always have the value of zero.
11018 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
11019 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (machine_mode @var{mode})
11020 This function describes how the standard shift patterns for @var{mode}
11021 deal with shifts by negative amounts or by more than the width of the mode.
11022 @xref{shift patterns}.
11024 On many machines, the shift patterns will apply a mask @var{m} to the
11025 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
11026 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
11027 this is true for mode @var{mode}, the function should return @var{m},
11028 otherwise it should return 0. A return value of 0 indicates that no
11029 particular behavior is guaranteed.
11031 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
11032 @emph{not} apply to general shift rtxes; it applies only to instructions
11033 that are generated by the named shift patterns.
11035 The default implementation of this function returns
11036 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
11037 and 0 otherwise. This definition is always safe, but if
11038 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
11039 nevertheless truncate the shift count, you may get better code
11043 @deftypefn {Target Hook} bool TARGET_TRULY_NOOP_TRUNCATION (poly_uint64 @var{outprec}, poly_uint64 @var{inprec})
11044 This hook returns true if it is safe to ``convert'' a value of
11045 @var{inprec} bits to one of @var{outprec} bits (where @var{outprec} is
11046 smaller than @var{inprec}) by merely operating on it as if it had only
11047 @var{outprec} bits. The default returns true unconditionally, which
11048 is correct for most machines.
11050 If @code{TARGET_MODES_TIEABLE_P} returns false for a pair of modes,
11051 suboptimal code can result if this hook returns true for the corresponding
11052 mode sizes. Making this hook return false in such cases may improve things.
11055 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (scalar_int_mode @var{mode}, scalar_int_mode @var{rep_mode})
11056 The representation of an integral mode can be such that the values
11057 are always extended to a wider integral mode. Return
11058 @code{SIGN_EXTEND} if values of @var{mode} are represented in
11059 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
11060 otherwise. (Currently, none of the targets use zero-extended
11061 representation this way so unlike @code{LOAD_EXTEND_OP},
11062 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
11063 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
11064 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
11065 widest integral mode and currently we take advantage of this fact.)
11067 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
11068 value even if the extension is not performed on certain hard registers
11069 as long as for the @code{REGNO_REG_CLASS} of these hard registers
11070 @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false.
11072 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
11073 describe two related properties. If you define
11074 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
11075 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
11078 In order to enforce the representation of @code{mode},
11079 @code{TARGET_TRULY_NOOP_TRUNCATION} should return false when truncating to
11083 @deftypefn {Target Hook} bool TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P (void)
11084 On some targets, it is assumed that the compiler will spill all pseudos
11085 that are live across a call to @code{setjmp}, while other targets treat
11086 @code{setjmp} calls as normal function calls.
11088 This hook returns false if @code{setjmp} calls do not preserve all
11089 non-volatile registers so that gcc that must spill all pseudos that are
11090 live across @code{setjmp} calls. Define this to return true if the
11091 target does not need to spill all pseudos live across @code{setjmp} calls.
11092 The default implementation conservatively assumes all pseudos must be
11093 spilled across @code{setjmp} calls.
11096 @defmac STORE_FLAG_VALUE
11097 A C expression describing the value returned by a comparison operator
11098 with an integral mode and stored by a store-flag instruction
11099 (@samp{cstore@var{mode}4}) when the condition is true. This description must
11100 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
11101 comparison operators whose results have a @code{MODE_INT} mode.
11103 A value of 1 or @minus{}1 means that the instruction implementing the
11104 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
11105 and 0 when the comparison is false. Otherwise, the value indicates
11106 which bits of the result are guaranteed to be 1 when the comparison is
11107 true. This value is interpreted in the mode of the comparison
11108 operation, which is given by the mode of the first operand in the
11109 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
11110 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
11113 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
11114 generate code that depends only on the specified bits. It can also
11115 replace comparison operators with equivalent operations if they cause
11116 the required bits to be set, even if the remaining bits are undefined.
11117 For example, on a machine whose comparison operators return an
11118 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
11119 @samp{0x80000000}, saying that just the sign bit is relevant, the
11123 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
11127 can be converted to
11130 (ashift:SI @var{x} (const_int @var{n}))
11134 where @var{n} is the appropriate shift count to move the bit being
11135 tested into the sign bit.
11137 There is no way to describe a machine that always sets the low-order bit
11138 for a true value, but does not guarantee the value of any other bits,
11139 but we do not know of any machine that has such an instruction. If you
11140 are trying to port GCC to such a machine, include an instruction to
11141 perform a logical-and of the result with 1 in the pattern for the
11142 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
11144 Often, a machine will have multiple instructions that obtain a value
11145 from a comparison (or the condition codes). Here are rules to guide the
11146 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
11151 Use the shortest sequence that yields a valid definition for
11152 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
11153 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
11154 comparison operators to do so because there may be opportunities to
11155 combine the normalization with other operations.
11158 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
11159 slightly preferred on machines with expensive jumps and 1 preferred on
11163 As a second choice, choose a value of @samp{0x80000001} if instructions
11164 exist that set both the sign and low-order bits but do not define the
11168 Otherwise, use a value of @samp{0x80000000}.
11171 Many machines can produce both the value chosen for
11172 @code{STORE_FLAG_VALUE} and its negation in the same number of
11173 instructions. On those machines, you should also define a pattern for
11174 those cases, e.g., one matching
11177 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
11180 Some machines can also perform @code{and} or @code{plus} operations on
11181 condition code values with less instructions than the corresponding
11182 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
11183 machines, define the appropriate patterns. Use the names @code{incscc}
11184 and @code{decscc}, respectively, for the patterns which perform
11185 @code{plus} or @code{minus} operations on condition code values. See
11186 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
11187 find such instruction sequences on other machines.
11189 If this macro is not defined, the default value, 1, is used. You need
11190 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
11191 instructions, or if the value generated by these instructions is 1.
11194 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
11195 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
11196 returned when comparison operators with floating-point results are true.
11197 Define this macro on machines that have comparison operations that return
11198 floating-point values. If there are no such operations, do not define
11202 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
11203 A C expression that gives a rtx representing the nonzero true element
11204 for vector comparisons. The returned rtx should be valid for the inner
11205 mode of @var{mode} which is guaranteed to be a vector mode. Define
11206 this macro on machines that have vector comparison operations that
11207 return a vector result. If there are no such operations, do not define
11208 this macro. Typically, this macro is defined as @code{const1_rtx} or
11209 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
11210 the compiler optimizing such vector comparison operations for the
11214 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
11215 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
11216 A C expression that indicates whether the architecture defines a value
11217 for @code{clz} or @code{ctz} with a zero operand.
11218 A result of @code{0} indicates the value is undefined.
11219 If the value is defined for only the RTL expression, the macro should
11220 evaluate to @code{1}; if the value applies also to the corresponding optab
11221 entry (which is normally the case if it expands directly into
11222 the corresponding RTL), then the macro should evaluate to @code{2}.
11223 In the cases where the value is defined, @var{value} should be set to
11226 If this macro is not defined, the value of @code{clz} or
11227 @code{ctz} at zero is assumed to be undefined.
11229 This macro must be defined if the target's expansion for @code{ffs}
11230 relies on a particular value to get correct results. Otherwise it
11231 is not necessary, though it may be used to optimize some corner cases, and
11232 to provide a default expansion for the @code{ffs} optab.
11234 Note that regardless of this macro the ``definedness'' of @code{clz}
11235 and @code{ctz} at zero do @emph{not} extend to the builtin functions
11236 visible to the user. Thus one may be free to adjust the value at will
11237 to match the target expansion of these operations without fear of
11242 An alias for the machine mode for pointers. On most machines, define
11243 this to be the integer mode corresponding to the width of a hardware
11244 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
11245 On some machines you must define this to be one of the partial integer
11246 modes, such as @code{PSImode}.
11248 The width of @code{Pmode} must be at least as large as the value of
11249 @code{POINTER_SIZE}. If it is not equal, you must define the macro
11250 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
11254 @defmac FUNCTION_MODE
11255 An alias for the machine mode used for memory references to functions
11256 being called, in @code{call} RTL expressions. On most CISC machines,
11257 where an instruction can begin at any byte address, this should be
11258 @code{QImode}. On most RISC machines, where all instructions have fixed
11259 size and alignment, this should be a mode with the same size and alignment
11260 as the machine instruction words - typically @code{SImode} or @code{HImode}.
11263 @defmac STDC_0_IN_SYSTEM_HEADERS
11264 In normal operation, the preprocessor expands @code{__STDC__} to the
11265 constant 1, to signify that GCC conforms to ISO Standard C@. On some
11266 hosts, like Solaris, the system compiler uses a different convention,
11267 where @code{__STDC__} is normally 0, but is 1 if the user specifies
11268 strict conformance to the C Standard.
11270 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
11271 convention when processing system header files, but when processing user
11272 files @code{__STDC__} will always expand to 1.
11275 @deftypefn {C Target Hook} {const char *} TARGET_C_PREINCLUDE (void)
11276 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.
11278 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.
11281 @deftypefn {C Target Hook} bool TARGET_CXX_IMPLICIT_EXTERN_C (const char*@var{})
11282 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.
11285 @defmac SYSTEM_IMPLICIT_EXTERN_C
11286 Define this macro if the system header files do not support C++@.
11287 This macro handles system header files by pretending that system
11288 header files are enclosed in @samp{extern "C" @{@dots{}@}}.
11293 @defmac REGISTER_TARGET_PRAGMAS ()
11294 Define this macro if you want to implement any target-specific pragmas.
11295 If defined, it is a C expression which makes a series of calls to
11296 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
11297 for each pragma. The macro may also do any
11298 setup required for the pragmas.
11300 The primary reason to define this macro is to provide compatibility with
11301 other compilers for the same target. In general, we discourage
11302 definition of target-specific pragmas for GCC@.
11304 If the pragma can be implemented by attributes then you should consider
11305 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
11307 Preprocessor macros that appear on pragma lines are not expanded. All
11308 @samp{#pragma} directives that do not match any registered pragma are
11309 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
11312 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
11313 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
11315 Each call to @code{c_register_pragma} or
11316 @code{c_register_pragma_with_expansion} establishes one pragma. The
11317 @var{callback} routine will be called when the preprocessor encounters a
11321 #pragma [@var{space}] @var{name} @dots{}
11324 @var{space} is the case-sensitive namespace of the pragma, or
11325 @code{NULL} to put the pragma in the global namespace. The callback
11326 routine receives @var{pfile} as its first argument, which can be passed
11327 on to cpplib's functions if necessary. You can lex tokens after the
11328 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
11329 callback will be silently ignored. The end of the line is indicated by
11330 a token of type @code{CPP_EOF}. Macro expansion occurs on the
11331 arguments of pragmas registered with
11332 @code{c_register_pragma_with_expansion} but not on the arguments of
11333 pragmas registered with @code{c_register_pragma}.
11335 Note that the use of @code{pragma_lex} is specific to the C and C++
11336 compilers. It will not work in the Java or Fortran compilers, or any
11337 other language compilers for that matter. Thus if @code{pragma_lex} is going
11338 to be called from target-specific code, it must only be done so when
11339 building the C and C++ compilers. This can be done by defining the
11340 variables @code{c_target_objs} and @code{cxx_target_objs} in the
11341 target entry in the @file{config.gcc} file. These variables should name
11342 the target-specific, language-specific object file which contains the
11343 code that uses @code{pragma_lex}. Note it will also be necessary to add a
11344 rule to the makefile fragment pointed to by @code{tmake_file} that shows
11345 how to build this object file.
11348 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
11349 Define this macro if macros should be expanded in the
11350 arguments of @samp{#pragma pack}.
11353 @defmac TARGET_DEFAULT_PACK_STRUCT
11354 If your target requires a structure packing default other than 0 (meaning
11355 the machine default), define this macro to the necessary value (in bytes).
11356 This must be a value that would also be valid to use with
11357 @samp{#pragma pack()} (that is, a small power of two).
11360 @defmac DOLLARS_IN_IDENTIFIERS
11361 Define this macro to control use of the character @samp{$} in
11362 identifier names for the C family of languages. 0 means @samp{$} is
11363 not allowed by default; 1 means it is allowed. 1 is the default;
11364 there is no need to define this macro in that case.
11367 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
11368 Define this macro as a C expression that is nonzero if it is safe for the
11369 delay slot scheduler to place instructions in the delay slot of @var{insn},
11370 even if they appear to use a resource set or clobbered in @var{insn}.
11371 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
11372 every @code{call_insn} has this behavior. On machines where some @code{insn}
11373 or @code{jump_insn} is really a function call and hence has this behavior,
11374 you should define this macro.
11376 You need not define this macro if it would always return zero.
11379 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
11380 Define this macro as a C expression that is nonzero if it is safe for the
11381 delay slot scheduler to place instructions in the delay slot of @var{insn},
11382 even if they appear to set or clobber a resource referenced in @var{insn}.
11383 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
11384 some @code{insn} or @code{jump_insn} is really a function call and its operands
11385 are registers whose use is actually in the subroutine it calls, you should
11386 define this macro. Doing so allows the delay slot scheduler to move
11387 instructions which copy arguments into the argument registers into the delay
11388 slot of @var{insn}.
11390 You need not define this macro if it would always return zero.
11393 @defmac MULTIPLE_SYMBOL_SPACES
11394 Define this macro as a C expression that is nonzero if, in some cases,
11395 global symbols from one translation unit may not be bound to undefined
11396 symbols in another translation unit without user intervention. For
11397 instance, under Microsoft Windows symbols must be explicitly imported
11398 from shared libraries (DLLs).
11400 You need not define this macro if it would always evaluate to zero.
11403 @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})
11404 This target hook may add @dfn{clobbers} to @var{clobbers} and
11405 @var{clobbered_regs} for any hard regs the port wishes to automatically
11406 clobber for an asm. The @var{outputs} and @var{inputs} may be inspected
11407 to avoid clobbering a register that is already used by the asm.
11409 It may modify the @var{outputs}, @var{inputs}, and @var{constraints}
11410 as necessary for other pre-processing. In this case the return value is
11411 a sequence of insns to emit after the asm.
11414 @defmac MATH_LIBRARY
11415 Define this macro as a C string constant for the linker argument to link
11416 in the system math library, minus the initial @samp{"-l"}, or
11417 @samp{""} if the target does not have a
11418 separate math library.
11420 You need only define this macro if the default of @samp{"m"} is wrong.
11423 @defmac LIBRARY_PATH_ENV
11424 Define this macro as a C string constant for the environment variable that
11425 specifies where the linker should look for libraries.
11427 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
11431 @defmac TARGET_POSIX_IO
11432 Define this macro if the target supports the following POSIX@ file
11433 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
11434 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
11435 to use file locking when exiting a program, which avoids race conditions
11436 if the program has forked. It will also create directories at run-time
11437 for cross-profiling.
11440 @defmac MAX_CONDITIONAL_EXECUTE
11442 A C expression for the maximum number of instructions to execute via
11443 conditional execution instructions instead of a branch. A value of
11444 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
11445 1 if it does use cc0.
11448 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
11449 Used if the target needs to perform machine-dependent modifications on the
11450 conditionals used for turning basic blocks into conditionally executed code.
11451 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
11452 contains information about the currently processed blocks. @var{true_expr}
11453 and @var{false_expr} are the tests that are used for converting the
11454 then-block and the else-block, respectively. Set either @var{true_expr} or
11455 @var{false_expr} to a null pointer if the tests cannot be converted.
11458 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
11459 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
11460 if-statements into conditions combined by @code{and} and @code{or} operations.
11461 @var{bb} contains the basic block that contains the test that is currently
11462 being processed and about to be turned into a condition.
11465 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
11466 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
11467 be converted to conditional execution format. @var{ce_info} points to
11468 a data structure, @code{struct ce_if_block}, which contains information
11469 about the currently processed blocks.
11472 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
11473 A C expression to perform any final machine dependent modifications in
11474 converting code to conditional execution. The involved basic blocks
11475 can be found in the @code{struct ce_if_block} structure that is pointed
11476 to by @var{ce_info}.
11479 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
11480 A C expression to cancel any machine dependent modifications in
11481 converting code to conditional execution. The involved basic blocks
11482 can be found in the @code{struct ce_if_block} structure that is pointed
11483 to by @var{ce_info}.
11486 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
11487 A C expression to initialize any machine specific data for if-conversion
11488 of the if-block in the @code{struct ce_if_block} structure that is pointed
11489 to by @var{ce_info}.
11492 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
11493 If non-null, this hook performs a target-specific pass over the
11494 instruction stream. The compiler will run it at all optimization levels,
11495 just before the point at which it normally does delayed-branch scheduling.
11497 The exact purpose of the hook varies from target to target. Some use
11498 it to do transformations that are necessary for correctness, such as
11499 laying out in-function constant pools or avoiding hardware hazards.
11500 Others use it as an opportunity to do some machine-dependent optimizations.
11502 You need not implement the hook if it has nothing to do. The default
11503 definition is null.
11506 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
11507 Define this hook if you have any machine-specific built-in functions
11508 that need to be defined. It should be a function that performs the
11511 Machine specific built-in functions can be useful to expand special machine
11512 instructions that would otherwise not normally be generated because
11513 they have no equivalent in the source language (for example, SIMD vector
11514 instructions or prefetch instructions).
11516 To create a built-in function, call the function
11517 @code{lang_hooks.builtin_function}
11518 which is defined by the language front end. You can use any type nodes set
11519 up by @code{build_common_tree_nodes};
11520 only language front ends that use those two functions will call
11521 @samp{TARGET_INIT_BUILTINS}.
11524 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
11525 Define this hook if you have any machine-specific built-in functions
11526 that need to be defined. It should be a function that returns the
11527 builtin function declaration for the builtin function code @var{code}.
11528 If there is no such builtin and it cannot be initialized at this time
11529 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
11530 If @var{code} is out of range the function should return
11531 @code{error_mark_node}.
11534 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, machine_mode @var{mode}, int @var{ignore})
11536 Expand a call to a machine specific built-in function that was set up by
11537 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
11538 function call; the result should go to @var{target} if that is
11539 convenient, and have mode @var{mode} if that is convenient.
11540 @var{subtarget} may be used as the target for computing one of
11541 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
11542 ignored. This function should return the result of the call to the
11546 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
11547 Select a replacement for a machine specific built-in function that
11548 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
11549 @emph{before} regular type checking, and so allows the target to
11550 implement a crude form of function overloading. @var{fndecl} is the
11551 declaration of the built-in function. @var{arglist} is the list of
11552 arguments passed to the built-in function. The result is a
11553 complete expression that implements the operation, usually
11554 another @code{CALL_EXPR}.
11555 @var{arglist} really has type @samp{VEC(tree,gc)*}
11558 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
11559 Fold a call to a machine specific built-in function that was set up by
11560 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
11561 built-in function. @var{n_args} is the number of arguments passed to
11562 the function; the arguments themselves are pointed to by @var{argp}.
11563 The result is another tree, valid for both GIMPLE and GENERIC,
11564 containing a simplified expression for the call's result. If
11565 @var{ignore} is true the value will be ignored.
11568 @deftypefn {Target Hook} bool TARGET_GIMPLE_FOLD_BUILTIN (gimple_stmt_iterator *@var{gsi})
11569 Fold a call to a machine specific built-in function that was set up
11570 by @samp{TARGET_INIT_BUILTINS}. @var{gsi} points to the gimple
11571 statement holding the function call. Returns true if any change
11572 was made to the GIMPLE stream.
11575 @deftypefn {Target Hook} int TARGET_COMPARE_VERSION_PRIORITY (tree @var{decl1}, tree @var{decl2})
11576 This hook is used to compare the target attributes in two functions to
11577 determine which function's features get higher priority. This is used
11578 during function multi-versioning to figure out the order in which two
11579 versions must be dispatched. A function version with a higher priority
11580 is checked for dispatching earlier. @var{decl1} and @var{decl2} are
11581 the two function decls that will be compared.
11584 @deftypefn {Target Hook} tree TARGET_GET_FUNCTION_VERSIONS_DISPATCHER (void *@var{decl})
11585 This hook is used to get the dispatcher function for a set of function
11586 versions. The dispatcher function is called to invoke the right function
11587 version at run-time. @var{decl} is one version from a set of semantically
11588 identical versions.
11591 @deftypefn {Target Hook} tree TARGET_GENERATE_VERSION_DISPATCHER_BODY (void *@var{arg})
11592 This hook is used to generate the dispatcher logic to invoke the right
11593 function version at run-time for a given set of function versions.
11594 @var{arg} points to the callgraph node of the dispatcher function whose
11595 body must be generated.
11598 @deftypefn {Target Hook} bool TARGET_PREDICT_DOLOOP_P (class loop *@var{loop})
11599 Return true if we can predict it is possible to use a low-overhead loop
11600 for a particular loop. The parameter @var{loop} is a pointer to the loop.
11601 This target hook is required only when the target supports low-overhead
11602 loops, and will help ivopts to make some decisions.
11603 The default version of this hook returns false.
11606 @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})
11607 Return true if it is possible to use low-overhead loops (@code{doloop_end}
11608 and @code{doloop_begin}) for a particular loop. @var{iterations} gives the
11609 exact number of iterations, or 0 if not known. @var{iterations_max} gives
11610 the maximum number of iterations, or 0 if not known. @var{loop_depth} is
11611 the nesting depth of the loop, with 1 for innermost loops, 2 for loops that
11612 contain innermost loops, and so on. @var{entered_at_top} is true if the
11613 loop is only entered from the top.
11615 This hook is only used if @code{doloop_end} is available. The default
11616 implementation returns true. You can use @code{can_use_doloop_if_innermost}
11617 if the loop must be the innermost, and if there are no other restrictions.
11620 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const rtx_insn *@var{insn})
11622 Take an instruction in @var{insn} and return NULL if it is valid within a
11623 low-overhead loop, otherwise return a string explaining why doloop
11624 could not be applied.
11626 Many targets use special registers for low-overhead looping. For any
11627 instruction that clobbers these this function should return a string indicating
11628 the reason why the doloop could not be applied.
11629 By default, the RTL loop optimizer does not use a present doloop pattern for
11630 loops containing function calls or branch on table instructions.
11633 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_COMBINED_INSN (rtx_insn *@var{insn})
11634 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.
11637 @deftypefn {Target Hook} bool TARGET_CAN_FOLLOW_JUMP (const rtx_insn *@var{follower}, const rtx_insn *@var{followee})
11638 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.
11641 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
11642 This target hook returns @code{true} if @var{x} is considered to be commutative.
11643 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
11644 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
11645 of the enclosing rtl, if known, otherwise it is UNKNOWN.
11648 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
11650 When the initial value of a hard register has been copied in a pseudo
11651 register, it is often not necessary to actually allocate another register
11652 to this pseudo register, because the original hard register or a stack slot
11653 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
11654 is called at the start of register allocation once for each hard register
11655 that had its initial value copied by using
11656 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
11657 Possible values are @code{NULL_RTX}, if you don't want
11658 to do any special allocation, a @code{REG} rtx---that would typically be
11659 the hard register itself, if it is known not to be clobbered---or a
11661 If you are returning a @code{MEM}, this is only a hint for the allocator;
11662 it might decide to use another register anyways.
11663 You may use @code{current_function_is_leaf} or
11664 @code{REG_N_SETS} in the hook to determine if the hard
11665 register in question will not be clobbered.
11666 The default value of this hook is @code{NULL}, which disables any special
11670 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
11671 This target hook returns nonzero if @var{x}, an @code{unspec} or
11672 @code{unspec_volatile} operation, might cause a trap. Targets can use
11673 this hook to enhance precision of analysis for @code{unspec} and
11674 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
11675 to analyze inner elements of @var{x} in which case @var{flags} should be
11679 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
11680 The compiler invokes this hook whenever it changes its current function
11681 context (@code{cfun}). You can define this function if
11682 the back end needs to perform any initialization or reset actions on a
11683 per-function basis. For example, it may be used to implement function
11684 attributes that affect register usage or code generation patterns.
11685 The argument @var{decl} is the declaration for the new function context,
11686 and may be null to indicate that the compiler has left a function context
11687 and is returning to processing at the top level.
11688 The default hook function does nothing.
11690 GCC sets @code{cfun} to a dummy function context during initialization of
11691 some parts of the back end. The hook function is not invoked in this
11692 situation; you need not worry about the hook being invoked recursively,
11693 or when the back end is in a partially-initialized state.
11694 @code{cfun} might be @code{NULL} to indicate processing at top level,
11695 outside of any function scope.
11698 @defmac TARGET_OBJECT_SUFFIX
11699 Define this macro to be a C string representing the suffix for object
11700 files on your target machine. If you do not define this macro, GCC will
11701 use @samp{.o} as the suffix for object files.
11704 @defmac TARGET_EXECUTABLE_SUFFIX
11705 Define this macro to be a C string representing the suffix to be
11706 automatically added to executable files on your target machine. If you
11707 do not define this macro, GCC will use the null string as the suffix for
11711 @defmac COLLECT_EXPORT_LIST
11712 If defined, @code{collect2} will scan the individual object files
11713 specified on its command line and create an export list for the linker.
11714 Define this macro for systems like AIX, where the linker discards
11715 object files that are not referenced from @code{main} and uses export
11719 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11720 Define this macro to a C expression representing a variant of the
11721 method call @var{mdecl}, if Java Native Interface (JNI) methods
11722 must be invoked differently from other methods on your target.
11723 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11724 the @code{stdcall} calling convention and this macro is then
11725 defined as this expression:
11728 build_type_attribute_variant (@var{mdecl},
11730 (get_identifier ("stdcall"),
11735 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11736 This target hook returns @code{true} past the point in which new jump
11737 instructions could be created. On machines that require a register for
11738 every jump such as the SHmedia ISA of SH5, this point would typically be
11739 reload, so this target hook should be defined to a function such as:
11743 cannot_modify_jumps_past_reload_p ()
11745 return (reload_completed || reload_in_progress);
11750 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11751 This target hook returns a register class for which branch target register
11752 optimizations should be applied. All registers in this class should be
11753 usable interchangeably. After reload, registers in this class will be
11754 re-allocated and loads will be hoisted out of loops and be subjected
11755 to inter-block scheduling.
11758 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11759 Branch target register optimization will by default exclude callee-saved
11761 that are not already live during the current function; if this target hook
11762 returns true, they will be included. The target code must than make sure
11763 that all target registers in the class returned by
11764 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11765 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11766 epilogues have already been generated. Note, even if you only return
11767 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11768 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11769 to reserve space for caller-saved target registers.
11772 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11773 This target hook returns true if the target supports conditional execution.
11774 This target hook is required only when the target has several different
11775 modes and they have different conditional execution capability, such as ARM.
11778 @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})
11779 This function prepares to emit a comparison insn for the first compare in a
11780 sequence of conditional comparisions. It returns an appropriate comparison
11781 with @code{CC} for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11782 The insns to prepare the compare are saved in @var{prep_seq} and the compare
11783 insns are saved in @var{gen_seq}. They will be emitted when all the
11784 compares in the the conditional comparision are generated without error.
11785 @var{code} is the @code{rtx_code} of the compare for @var{op0} and @var{op1}.
11788 @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})
11789 This function prepares to emit a conditional comparison within a sequence
11790 of conditional comparisons. It returns an appropriate comparison with
11791 @code{CC} for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11792 The insns to prepare the compare are saved in @var{prep_seq} and the compare
11793 insns are saved in @var{gen_seq}. They will be emitted when all the
11794 compares in the conditional comparision are generated without error. The
11795 @var{prev} expression is the result of a prior call to @code{gen_ccmp_first}
11796 or @code{gen_ccmp_next}. It may return @code{NULL} if the combination of
11797 @var{prev} and this comparison is not supported, otherwise the result must
11798 be appropriate for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11799 @var{code} is the @code{rtx_code} of the compare for @var{op0} and @var{op1}.
11800 @var{bit_code} is @code{AND} or @code{IOR}, which is the op on the compares.
11803 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, class loop *@var{loop})
11804 This target hook returns a new value for the number of times @var{loop}
11805 should be unrolled. The parameter @var{nunroll} is the number of times
11806 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11807 the loop, which is going to be checked for unrolling. This target hook
11808 is required only when the target has special constraints like maximum
11809 number of memory accesses.
11812 @defmac POWI_MAX_MULTS
11813 If defined, this macro is interpreted as a signed integer C expression
11814 that specifies the maximum number of floating point multiplications
11815 that should be emitted when expanding exponentiation by an integer
11816 constant inline. When this value is defined, exponentiation requiring
11817 more than this number of multiplications is implemented by calling the
11818 system library's @code{pow}, @code{powf} or @code{powl} routines.
11819 The default value places no upper bound on the multiplication count.
11822 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11823 This target hook should register any extra include files for the
11824 target. The parameter @var{stdinc} indicates if normal include files
11825 are present. The parameter @var{sysroot} is the system root directory.
11826 The parameter @var{iprefix} is the prefix for the gcc directory.
11829 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11830 This target hook should register any extra include files for the
11831 target before any standard headers. The parameter @var{stdinc}
11832 indicates if normal include files are present. The parameter
11833 @var{sysroot} is the system root directory. The parameter
11834 @var{iprefix} is the prefix for the gcc directory.
11837 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11838 This target hook should register special include paths for the target.
11839 The parameter @var{path} is the include to register. On Darwin
11840 systems, this is used for Framework includes, which have semantics
11841 that are different from @option{-I}.
11844 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11845 This target macro returns @code{true} if it is safe to use a local alias
11846 for a virtual function @var{fndecl} when constructing thunks,
11847 @code{false} otherwise. By default, the macro returns @code{true} for all
11848 functions, if a target supports aliases (i.e.@: defines
11849 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11852 @defmac TARGET_FORMAT_TYPES
11853 If defined, this macro is the name of a global variable containing
11854 target-specific format checking information for the @option{-Wformat}
11855 option. The default is to have no target-specific format checks.
11858 @defmac TARGET_N_FORMAT_TYPES
11859 If defined, this macro is the number of entries in
11860 @code{TARGET_FORMAT_TYPES}.
11863 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11864 If defined, this macro is the name of a global variable containing
11865 target-specific format overrides for the @option{-Wformat} option. The
11866 default is to have no target-specific format overrides. If defined,
11867 @code{TARGET_FORMAT_TYPES} must be defined, too.
11870 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11871 If defined, this macro specifies the number of entries in
11872 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11875 @defmac TARGET_OVERRIDES_FORMAT_INIT
11876 If defined, this macro specifies the optional initialization
11877 routine for target specific customizations of the system printf
11878 and scanf formatter settings.
11881 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11882 If defined, this macro returns the diagnostic message when it is
11883 illegal to pass argument @var{val} to function @var{funcdecl}
11884 with prototype @var{typelist}.
11887 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11888 If defined, this macro returns the diagnostic message when it is
11889 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11890 if validity should be determined by the front end.
11893 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11894 If defined, this macro returns the diagnostic message when it is
11895 invalid to apply operation @var{op} (where unary plus is denoted by
11896 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11897 if validity should be determined by the front end.
11900 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11901 If defined, this macro returns the diagnostic message when it is
11902 invalid to apply operation @var{op} to operands of types @var{type1}
11903 and @var{type2}, or @code{NULL} if validity should be determined by
11907 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11908 If defined, this target hook returns the type to which values of
11909 @var{type} should be promoted when they appear in expressions,
11910 analogous to the integer promotions, or @code{NULL_TREE} to use the
11911 front end's normal promotion rules. This hook is useful when there are
11912 target-specific types with special promotion rules.
11913 This is currently used only by the C and C++ front ends.
11916 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11917 If defined, this hook returns the result of converting @var{expr} to
11918 @var{type}. It should return the converted expression,
11919 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11920 This hook is useful when there are target-specific types with special
11922 This is currently used only by the C and C++ front ends.
11926 This macro determines the size of the objective C jump buffer for the
11927 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11930 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11931 Define this macro if any target-specific attributes need to be attached
11932 to the functions in @file{libgcc} that provide low-level support for
11933 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11934 and the associated definitions of those functions.
11937 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11938 Define this macro to update the current function stack boundary if
11942 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11943 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11944 different argument pointer register is needed to access the function's
11945 argument list due to stack realignment. Return @code{NULL} if no DRAP
11949 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11950 When optimization is disabled, this hook indicates whether or not
11951 arguments should be allocated to stack slots. Normally, GCC allocates
11952 stacks slots for arguments when not optimizing in order to make
11953 debugging easier. However, when a function is declared with
11954 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11955 cannot safely move arguments from the registers in which they are passed
11956 to the stack. Therefore, this hook should return true in general, but
11957 false for naked functions. The default implementation always returns true.
11960 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11961 On some architectures it can take multiple instructions to synthesize
11962 a constant. If there is another constant already in a register that
11963 is close enough in value then it is preferable that the new constant
11964 is computed from this register using immediate addition or
11965 subtraction. We accomplish this through CSE. Besides the value of
11966 the constant we also add a lower and an upper constant anchor to the
11967 available expressions. These are then queried when encountering new
11968 constants. The anchors are computed by rounding the constant up and
11969 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11970 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11971 accepted by immediate-add plus one. We currently assume that the
11972 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11973 MIPS, where add-immediate takes a 16-bit signed value,
11974 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11975 is zero, which disables this optimization.
11978 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_ASAN_SHADOW_OFFSET (void)
11979 Return the offset bitwise ored into shifted address to get corresponding
11980 Address Sanitizer shadow memory address. NULL if Address Sanitizer is not
11981 supported by the target.
11984 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_MEMMODEL_CHECK (unsigned HOST_WIDE_INT @var{val})
11985 Validate target specific memory model mask bits. When NULL no target specific
11986 memory model bits are allowed.
11989 @deftypevr {Target Hook} {unsigned char} TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
11990 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}.
11993 @deftypefn {Target Hook} bool TARGET_HAS_IFUNC_P (void)
11994 It returns true if the target supports GNU indirect functions.
11995 The support includes the assembler, linker and dynamic linker.
11996 The default value of this hook is based on target's libc.
11999 @deftypefn {Target Hook} {unsigned int} TARGET_ATOMIC_ALIGN_FOR_MODE (machine_mode @var{mode})
12000 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.
12003 @deftypefn {Target Hook} void TARGET_ATOMIC_ASSIGN_EXPAND_FENV (tree *@var{hold}, tree *@var{clear}, tree *@var{update})
12004 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}}.
12007 @deftypefn {Target Hook} void TARGET_RECORD_OFFLOAD_SYMBOL (tree)
12008 Used when offloaded functions are seen in the compilation unit and no named
12009 sections are available. It is called once for each symbol that must be
12010 recorded in the offload function and variable table.
12013 @deftypefn {Target Hook} {char *} TARGET_OFFLOAD_OPTIONS (void)
12014 Used when writing out the list of options into an LTO file. It should
12015 translate any relevant target-specific options (such as the ABI in use)
12016 into one of the @option{-foffload} options that exist as a common interface
12017 to express such options. It should return a string containing these options,
12018 separated by spaces, which the caller will free.
12022 @defmac TARGET_SUPPORTS_WIDE_INT
12024 On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
12025 objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
12026 to indicate that large integers are stored in
12027 @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows
12028 very large integer constants to be represented. @code{CONST_DOUBLE}
12029 is limited to twice the size of the host's @code{HOST_WIDE_INT}
12032 Converting a port mostly requires looking for the places where
12033 @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
12034 code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i
12035 const_double"} at the port level gets you to 95% of the changes that
12036 need to be made. There are a few places that require a deeper look.
12040 There is no equivalent to @code{hval} and @code{lval} for
12041 @code{CONST_WIDE_INT}s. This would be difficult to express in the md
12042 language since there are a variable number of elements.
12044 Most ports only check that @code{hval} is either 0 or -1 to see if the
12045 value is small. As mentioned above, this will no longer be necessary
12046 since small constants are always @code{CONST_INT}. Of course there
12047 are still a few exceptions, the alpha's constraint used by the zap
12048 instruction certainly requires careful examination by C code.
12049 However, all the current code does is pass the hval and lval to C
12050 code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
12051 not really a large change.
12054 Because there is no standard template that ports use to materialize
12055 constants, there is likely to be some futzing that is unique to each
12059 The rtx costs may have to be adjusted to properly account for larger
12060 constants that are represented as @code{CONST_WIDE_INT}.
12063 All and all it does not take long to convert ports that the
12064 maintainer is familiar with.
12068 @deftypefn {Target Hook} bool TARGET_HAVE_SPECULATION_SAFE_VALUE (bool @var{active})
12069 This hook is used to determine the level of target support for
12070 @code{__builtin_speculation_safe_value}. If called with an argument
12071 of false, it returns true if the target has been modified to support
12072 this builtin. If called with an argument of true, it returns true
12073 if the target requires active mitigation execution might be speculative.
12075 The default implementation returns false if the target does not define
12076 a pattern named @code{speculation_barrier}. Else it returns true
12077 for the first case and whether the pattern is enabled for the current
12078 compilation for the second case.
12080 For targets that have no processors that can execute instructions
12081 speculatively an alternative implemenation of this hook is available:
12082 simply redefine this hook to @code{speculation_safe_value_not_needed}
12083 along with your other target hooks.
12086 @deftypefn {Target Hook} rtx TARGET_SPECULATION_SAFE_VALUE (machine_mode @var{mode}, rtx @var{result}, rtx @var{val}, rtx @var{failval})
12087 This target hook can be used to generate a target-specific code
12088 sequence that implements the @code{__builtin_speculation_safe_value}
12089 built-in function. The function must always return @var{val} in
12090 @var{result} in mode @var{mode} when the cpu is not executing
12091 speculatively, but must never return that when speculating until it
12092 is known that the speculation will not be unwound. The hook supports
12093 two primary mechanisms for implementing the requirements. The first
12094 is to emit a speculation barrier which forces the processor to wait
12095 until all prior speculative operations have been resolved; the second
12096 is to use a target-specific mechanism that can track the speculation
12097 state and to return @var{failval} if it can determine that
12098 speculation must be unwound at a later time.
12100 The default implementation simply copies @var{val} to @var{result} and
12101 emits a @code{speculation_barrier} instruction if that is defined.
12104 @deftypefn {Target Hook} void TARGET_RUN_TARGET_SELFTESTS (void)
12105 If selftests are enabled, run any selftests for this target.