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 @defmac SWITCHABLE_TARGET
762 Some targets need to switch between substantially different subtargets
763 during compilation. For example, the MIPS target has one subtarget for
764 the traditional MIPS architecture and another for MIPS16. Source code
765 can switch between these two subarchitectures using the @code{mips16}
766 and @code{nomips16} attributes.
768 Such subtargets can differ in things like the set of available
769 registers, the set of available instructions, the costs of various
770 operations, and so on. GCC caches a lot of this type of information
771 in global variables, and recomputing them for each subtarget takes a
772 significant amount of time. The compiler therefore provides a facility
773 for maintaining several versions of the global variables and quickly
774 switching between them; see @file{target-globals.h} for details.
776 Define this macro to 1 if your target needs this facility. The default
780 @deftypefn {Target Hook} bool TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P (void)
781 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.
784 @node Per-Function Data
785 @section Defining data structures for per-function information.
786 @cindex per-function data
787 @cindex data structures
789 If the target needs to store information on a per-function basis, GCC
790 provides a macro and a couple of variables to allow this. Note, just
791 using statics to store the information is a bad idea, since GCC supports
792 nested functions, so you can be halfway through encoding one function
793 when another one comes along.
795 GCC defines a data structure called @code{struct function} which
796 contains all of the data specific to an individual function. This
797 structure contains a field called @code{machine} whose type is
798 @code{struct machine_function *}, which can be used by targets to point
799 to their own specific data.
801 If a target needs per-function specific data it should define the type
802 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
803 This macro should be used to initialize the function pointer
804 @code{init_machine_status}. This pointer is explained below.
806 One typical use of per-function, target specific data is to create an
807 RTX to hold the register containing the function's return address. This
808 RTX can then be used to implement the @code{__builtin_return_address}
809 function, for level 0.
811 Note---earlier implementations of GCC used a single data area to hold
812 all of the per-function information. Thus when processing of a nested
813 function began the old per-function data had to be pushed onto a
814 stack, and when the processing was finished, it had to be popped off the
815 stack. GCC used to provide function pointers called
816 @code{save_machine_status} and @code{restore_machine_status} to handle
817 the saving and restoring of the target specific information. Since the
818 single data area approach is no longer used, these pointers are no
821 @defmac INIT_EXPANDERS
822 Macro called to initialize any target specific information. This macro
823 is called once per function, before generation of any RTL has begun.
824 The intention of this macro is to allow the initialization of the
825 function pointer @code{init_machine_status}.
828 @deftypevar {void (*)(struct function *)} init_machine_status
829 If this function pointer is non-@code{NULL} it will be called once per
830 function, before function compilation starts, in order to allow the
831 target to perform any target specific initialization of the
832 @code{struct function} structure. It is intended that this would be
833 used to initialize the @code{machine} of that structure.
835 @code{struct machine_function} structures are expected to be freed by GC@.
836 Generally, any memory that they reference must be allocated by using
837 GC allocation, including the structure itself.
841 @section Storage Layout
842 @cindex storage layout
844 Note that the definitions of the macros in this table which are sizes or
845 alignments measured in bits do not need to be constant. They can be C
846 expressions that refer to static variables, such as the @code{target_flags}.
847 @xref{Run-time Target}.
849 @defmac BITS_BIG_ENDIAN
850 Define this macro to have the value 1 if the most significant bit in a
851 byte has the lowest number; otherwise define it to have the value zero.
852 This means that bit-field instructions count from the most significant
853 bit. If the machine has no bit-field instructions, then this must still
854 be defined, but it doesn't matter which value it is defined to. This
855 macro need not be a constant.
857 This macro does not affect the way structure fields are packed into
858 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
861 @defmac BYTES_BIG_ENDIAN
862 Define this macro to have the value 1 if the most significant byte in a
863 word has the lowest number. This macro need not be a constant.
866 @defmac WORDS_BIG_ENDIAN
867 Define this macro to have the value 1 if, in a multiword object, the
868 most significant word has the lowest number. This applies to both
869 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
870 order of words in memory is not the same as the order in registers. This
871 macro need not be a constant.
874 @defmac REG_WORDS_BIG_ENDIAN
875 On some machines, the order of words in a multiword object differs between
876 registers in memory. In such a situation, define this macro to describe
877 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
878 the order of words in memory.
881 @defmac FLOAT_WORDS_BIG_ENDIAN
882 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
883 @code{TFmode} floating point numbers are stored in memory with the word
884 containing the sign bit at the lowest address; otherwise define it to
885 have the value 0. This macro need not be a constant.
887 You need not define this macro if the ordering is the same as for
891 @defmac BITS_PER_WORD
892 Number of bits in a word. If you do not define this macro, the default
893 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
896 @defmac MAX_BITS_PER_WORD
897 Maximum number of bits in a word. If this is undefined, the default is
898 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
899 largest value that @code{BITS_PER_WORD} can have at run-time.
902 @defmac UNITS_PER_WORD
903 Number of storage units in a word; normally the size of a general-purpose
904 register, a power of two from 1 or 8.
907 @defmac MIN_UNITS_PER_WORD
908 Minimum number of units in a word. If this is undefined, the default is
909 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
910 smallest value that @code{UNITS_PER_WORD} can have at run-time.
914 Width of a pointer, in bits. You must specify a value no wider than the
915 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
916 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
917 a value the default is @code{BITS_PER_WORD}.
920 @defmac POINTERS_EXTEND_UNSIGNED
921 A C expression that determines how pointers should be extended from
922 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
923 greater than zero if pointers should be zero-extended, zero if they
924 should be sign-extended, and negative if some other sort of conversion
925 is needed. In the last case, the extension is done by the target's
926 @code{ptr_extend} instruction.
928 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
929 and @code{word_mode} are all the same width.
932 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
933 A macro to update @var{m} and @var{unsignedp} when an object whose type
934 is @var{type} and which has the specified mode and signedness is to be
935 stored in a register. This macro is only called when @var{type} is a
938 On most RISC machines, which only have operations that operate on a full
939 register, define this macro to set @var{m} to @code{word_mode} if
940 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
941 cases, only integer modes should be widened because wider-precision
942 floating-point operations are usually more expensive than their narrower
945 For most machines, the macro definition does not change @var{unsignedp}.
946 However, some machines, have instructions that preferentially handle
947 either signed or unsigned quantities of certain modes. For example, on
948 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
949 sign-extend the result to 64 bits. On such machines, set
950 @var{unsignedp} according to which kind of extension is more efficient.
952 Do not define this macro if it would never modify @var{m}.
955 @deftypefn {Target Hook} {enum flt_eval_method} TARGET_C_EXCESS_PRECISION (enum excess_precision_type @var{type})
956 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}.
959 @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})
960 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
961 function return values. The target hook should return the new mode
962 and possibly change @code{*@var{punsignedp}} if the promotion should
963 change signedness. This function is called only for scalar @emph{or
966 @var{for_return} allows to distinguish the promotion of arguments and
967 return values. If it is @code{1}, a return value is being promoted and
968 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
969 If it is @code{2}, the returned mode should be that of the register in
970 which an incoming parameter is copied, or the outgoing result is computed;
971 then the hook should return the same mode as @code{promote_mode}, though
972 the signedness may be different.
974 @var{type} can be NULL when promoting function arguments of libcalls.
976 The default is to not promote arguments and return values. You can
977 also define the hook to @code{default_promote_function_mode_always_promote}
978 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
981 @defmac PARM_BOUNDARY
982 Normal alignment required for function parameters on the stack, in
983 bits. All stack parameters receive at least this much alignment
984 regardless of data type. On most machines, this is the same as the
988 @defmac STACK_BOUNDARY
989 Define this macro to the minimum alignment enforced by hardware for the
990 stack pointer on this machine. The definition is a C expression for the
991 desired alignment (measured in bits). This value is used as a default
992 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
993 this should be the same as @code{PARM_BOUNDARY}.
996 @defmac PREFERRED_STACK_BOUNDARY
997 Define this macro if you wish to preserve a certain alignment for the
998 stack pointer, greater than what the hardware enforces. The definition
999 is a C expression for the desired alignment (measured in bits). This
1000 macro must evaluate to a value equal to or larger than
1001 @code{STACK_BOUNDARY}.
1004 @defmac INCOMING_STACK_BOUNDARY
1005 Define this macro if the incoming stack boundary may be different
1006 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
1007 to a value equal to or larger than @code{STACK_BOUNDARY}.
1010 @defmac FUNCTION_BOUNDARY
1011 Alignment required for a function entry point, in bits.
1014 @defmac BIGGEST_ALIGNMENT
1015 Biggest alignment that any data type can require on this machine, in
1016 bits. Note that this is not the biggest alignment that is supported,
1017 just the biggest alignment that, when violated, may cause a fault.
1020 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_ABSOLUTE_BIGGEST_ALIGNMENT
1021 If defined, this target hook specifies the absolute biggest alignment
1022 that a type or variable can have on this machine, otherwise,
1023 @code{BIGGEST_ALIGNMENT} is used.
1026 @defmac MALLOC_ABI_ALIGNMENT
1027 Alignment, in bits, a C conformant malloc implementation has to
1028 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1031 @defmac ATTRIBUTE_ALIGNED_VALUE
1032 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1033 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1036 @defmac MINIMUM_ATOMIC_ALIGNMENT
1037 If defined, the smallest alignment, in bits, that can be given to an
1038 object that can be referenced in one operation, without disturbing any
1039 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1040 on machines that don't have byte or half-word store operations.
1043 @defmac BIGGEST_FIELD_ALIGNMENT
1044 Biggest alignment that any structure or union field can require on this
1045 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1046 structure and union fields only, unless the field alignment has been set
1047 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1050 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{type}, @var{computed})
1051 An expression for the alignment of a structure field @var{field} of
1052 type @var{type} if the alignment computed in the usual way (including
1053 applying of @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1054 alignment) is @var{computed}. It overrides alignment only if the
1055 field alignment has not been set by the
1056 @code{__attribute__ ((aligned (@var{n})))} construct. Note that @var{field}
1057 may be @code{NULL_TREE} in case we just query for the minimum alignment
1058 of a field of type @var{type} in structure context.
1061 @defmac MAX_STACK_ALIGNMENT
1062 Biggest stack alignment guaranteed by the backend. Use this macro
1063 to specify the maximum alignment of a variable on stack.
1065 If not defined, the default value is @code{STACK_BOUNDARY}.
1067 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1068 @c But the fix for PR 32893 indicates that we can only guarantee
1069 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1070 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1073 @defmac MAX_OFILE_ALIGNMENT
1074 Biggest alignment supported by the object file format of this machine.
1075 Use this macro to limit the alignment which can be specified using the
1076 @code{__attribute__ ((aligned (@var{n})))} construct for functions and
1077 objects with static storage duration. The alignment of automatic
1078 objects may exceed the object file format maximum up to the maximum
1079 supported by GCC. If not defined, the default value is
1080 @code{BIGGEST_ALIGNMENT}.
1082 On systems that use ELF, the default (in @file{config/elfos.h}) is
1083 the largest supported 32-bit ELF section alignment representable on
1084 a 32-bit host e.g.@: @samp{(((uint64_t) 1 << 28) * 8)}.
1085 On 32-bit ELF the largest supported section alignment in bits is
1086 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1089 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_STATIC_RTX_ALIGNMENT (machine_mode @var{mode})
1090 This hook returns the preferred alignment in bits for a
1091 statically-allocated rtx, such as a constant pool entry. @var{mode}
1092 is the mode of the rtx. The default implementation returns
1093 @samp{GET_MODE_ALIGNMENT (@var{mode})}.
1096 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1097 If defined, a C expression to compute the alignment for a variable in
1098 the static store. @var{type} is the data type, and @var{basic-align} is
1099 the alignment that the object would ordinarily have. The value of this
1100 macro is used instead of that alignment to align the object.
1102 If this macro is not defined, then @var{basic-align} is used.
1105 One use of this macro is to increase alignment of medium-size data to
1106 make it all fit in fewer cache lines. Another is to cause character
1107 arrays to be word-aligned so that @code{strcpy} calls that copy
1108 constants to character arrays can be done inline.
1111 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1112 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1113 some alignment increase, instead of optimization only purposes. E.g.@
1114 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1115 must be aligned to 16 byte boundaries.
1117 If this macro is not defined, then @var{basic-align} is used.
1120 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_CONSTANT_ALIGNMENT (const_tree @var{constant}, HOST_WIDE_INT @var{basic_align})
1121 This hook returns the alignment in bits of a constant that is being
1122 placed in memory. @var{constant} is the constant and @var{basic_align}
1123 is the alignment that the object would ordinarily have.
1125 The default definition just returns @var{basic_align}.
1127 The typical use of this hook is to increase alignment for string
1128 constants to be word aligned so that @code{strcpy} calls that copy
1129 constants can be done inline. The function
1130 @code{constant_alignment_word_strings} provides such a definition.
1133 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1134 If defined, a C expression to compute the alignment for a variable in
1135 the local store. @var{type} is the data type, and @var{basic-align} is
1136 the alignment that the object would ordinarily have. The value of this
1137 macro is used instead of that alignment to align the object.
1139 If this macro is not defined, then @var{basic-align} is used.
1141 One use of this macro is to increase alignment of medium-size data to
1142 make it all fit in fewer cache lines.
1144 If the value of this macro has a type, it should be an unsigned type.
1147 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_VECTOR_ALIGNMENT (const_tree @var{type})
1148 This hook can be used to define the alignment for a vector of type
1149 @var{type}, in order to comply with a platform ABI. The default is to
1150 require natural alignment for vector types. The alignment returned by
1151 this hook must be a power-of-two multiple of the default alignment of
1152 the vector element type.
1155 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1156 If defined, a C expression to compute the alignment for stack slot.
1157 @var{type} is the data type, @var{mode} is the widest mode available,
1158 and @var{basic-align} is the alignment that the slot would ordinarily
1159 have. The value of this macro is used instead of that alignment to
1162 If this macro is not defined, then @var{basic-align} is used when
1163 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1166 This macro is to set alignment of stack slot to the maximum alignment
1167 of all possible modes which the slot may have.
1169 If the value of this macro has a type, it should be an unsigned type.
1172 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1173 If defined, a C expression to compute the alignment for a local
1174 variable @var{decl}.
1176 If this macro is not defined, then
1177 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1180 One use of this macro is to increase alignment of medium-size data to
1181 make it all fit in fewer cache lines.
1183 If the value of this macro has a type, it should be an unsigned type.
1186 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1187 If defined, a C expression to compute the minimum required alignment
1188 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1189 @var{mode}, assuming normal alignment @var{align}.
1191 If this macro is not defined, then @var{align} will be used.
1194 @defmac EMPTY_FIELD_BOUNDARY
1195 Alignment in bits to be given to a structure bit-field that follows an
1196 empty field such as @code{int : 0;}.
1198 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1201 @defmac STRUCTURE_SIZE_BOUNDARY
1202 Number of bits which any structure or union's size must be a multiple of.
1203 Each structure or union's size is rounded up to a multiple of this.
1205 If you do not define this macro, the default is the same as
1206 @code{BITS_PER_UNIT}.
1209 @defmac STRICT_ALIGNMENT
1210 Define this macro to be the value 1 if instructions will fail to work
1211 if given data not on the nominal alignment. If instructions will merely
1212 go slower in that case, define this macro as 0.
1215 @defmac PCC_BITFIELD_TYPE_MATTERS
1216 Define this if you wish to imitate the way many other C compilers handle
1217 alignment of bit-fields and the structures that contain them.
1219 The behavior is that the type written for a named bit-field (@code{int},
1220 @code{short}, or other integer type) imposes an alignment for the entire
1221 structure, as if the structure really did contain an ordinary field of
1222 that type. In addition, the bit-field is placed within the structure so
1223 that it would fit within such a field, not crossing a boundary for it.
1225 Thus, on most machines, a named bit-field whose type is written as
1226 @code{int} would not cross a four-byte boundary, and would force
1227 four-byte alignment for the whole structure. (The alignment used may
1228 not be four bytes; it is controlled by the other alignment parameters.)
1230 An unnamed bit-field will not affect the alignment of the containing
1233 If the macro is defined, its definition should be a C expression;
1234 a nonzero value for the expression enables this behavior.
1236 Note that if this macro is not defined, or its value is zero, some
1237 bit-fields may cross more than one alignment boundary. The compiler can
1238 support such references if there are @samp{insv}, @samp{extv}, and
1239 @samp{extzv} insns that can directly reference memory.
1241 The other known way of making bit-fields work is to define
1242 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1243 Then every structure can be accessed with fullwords.
1245 Unless the machine has bit-field instructions or you define
1246 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1247 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1249 If your aim is to make GCC use the same conventions for laying out
1250 bit-fields as are used by another compiler, here is how to investigate
1251 what the other compiler does. Compile and run this program:
1270 printf ("Size of foo1 is %d\n",
1271 sizeof (struct foo1));
1272 printf ("Size of foo2 is %d\n",
1273 sizeof (struct foo2));
1278 If this prints 2 and 5, then the compiler's behavior is what you would
1279 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1282 @defmac BITFIELD_NBYTES_LIMITED
1283 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1284 to aligning a bit-field within the structure.
1287 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1288 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1289 whether unnamed bitfields affect the alignment of the containing
1290 structure. The hook should return true if the structure should inherit
1291 the alignment requirements of an unnamed bitfield's type.
1294 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1295 This target hook should return @code{true} if accesses to volatile bitfields
1296 should use the narrowest mode possible. It should return @code{false} if
1297 these accesses should use the bitfield container type.
1299 The default is @code{false}.
1302 @deftypefn {Target Hook} bool TARGET_MEMBER_TYPE_FORCES_BLK (const_tree @var{field}, machine_mode @var{mode})
1303 Return true if a structure, union or array containing @var{field} should
1304 be accessed using @code{BLKMODE}.
1306 If @var{field} is the only field in the structure, @var{mode} is its
1307 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1308 case where structures of one field would require the structure's mode to
1309 retain the field's mode.
1311 Normally, this is not needed.
1314 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1315 Define this macro as an expression for the alignment of a type (given
1316 by @var{type} as a tree node) if the alignment computed in the usual
1317 way is @var{computed} and the alignment explicitly specified was
1320 The default is to use @var{specified} if it is larger; otherwise, use
1321 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1324 @defmac MAX_FIXED_MODE_SIZE
1325 An integer expression for the size in bits of the largest integer
1326 machine mode that should actually be used. All integer machine modes of
1327 this size or smaller can be used for structures and unions with the
1328 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1329 (DImode)} is assumed.
1332 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1333 If defined, an expression of type @code{machine_mode} that
1334 specifies the mode of the save area operand of a
1335 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1336 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1337 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1338 having its mode specified.
1340 You need not define this macro if it always returns @code{Pmode}. You
1341 would most commonly define this macro if the
1342 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1346 @defmac STACK_SIZE_MODE
1347 If defined, an expression of type @code{machine_mode} that
1348 specifies the mode of the size increment operand of an
1349 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1351 You need not define this macro if it always returns @code{word_mode}.
1352 You would most commonly define this macro if the @code{allocate_stack}
1353 pattern needs to support both a 32- and a 64-bit mode.
1356 @deftypefn {Target Hook} scalar_int_mode TARGET_LIBGCC_CMP_RETURN_MODE (void)
1357 This target hook should return the mode to be used for the return value
1358 of compare instructions expanded to libgcc calls. If not defined
1359 @code{word_mode} is returned which is the right choice for a majority of
1363 @deftypefn {Target Hook} scalar_int_mode TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1364 This target hook should return the mode to be used for the shift count operand
1365 of shift instructions expanded to libgcc calls. If not defined
1366 @code{word_mode} is returned which is the right choice for a majority of
1370 @deftypefn {Target Hook} scalar_int_mode TARGET_UNWIND_WORD_MODE (void)
1371 Return machine mode to be used for @code{_Unwind_Word} type.
1372 The default is to use @code{word_mode}.
1375 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1376 This target hook returns @code{true} if bit-fields in the given
1377 @var{record_type} are to be laid out following the rules of Microsoft
1378 Visual C/C++, namely: (i) a bit-field won't share the same storage
1379 unit with the previous bit-field if their underlying types have
1380 different sizes, and the bit-field will be aligned to the highest
1381 alignment of the underlying types of itself and of the previous
1382 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1383 the whole enclosing structure, even if it is unnamed; except that
1384 (iii) a zero-sized bit-field will be disregarded unless it follows
1385 another bit-field of nonzero size. If this hook returns @code{true},
1386 other macros that control bit-field layout are ignored.
1388 When a bit-field is inserted into a packed record, the whole size
1389 of the underlying type is used by one or more same-size adjacent
1390 bit-fields (that is, if its long:3, 32 bits is used in the record,
1391 and any additional adjacent long bit-fields are packed into the same
1392 chunk of 32 bits. However, if the size changes, a new field of that
1393 size is allocated). In an unpacked record, this is the same as using
1394 alignment, but not equivalent when packing.
1396 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1397 the latter will take precedence. If @samp{__attribute__((packed))} is
1398 used on a single field when MS bit-fields are in use, it will take
1399 precedence for that field, but the alignment of the rest of the structure
1400 may affect its placement.
1403 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1404 Returns true if the target supports decimal floating point.
1407 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1408 Returns true if the target supports fixed-point arithmetic.
1411 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1412 This hook is called just before expansion into rtl, allowing the target
1413 to perform additional initializations or analysis before the expansion.
1414 For example, the rs6000 port uses it to allocate a scratch stack slot
1415 for use in copying SDmode values between memory and floating point
1416 registers whenever the function being expanded has any SDmode
1420 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1421 This hook allows the backend to perform additional instantiations on rtl
1422 that are not actually in any insns yet, but will be later.
1425 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1426 If your target defines any fundamental types, or any types your target
1427 uses should be mangled differently from the default, define this hook
1428 to return the appropriate encoding for these types as part of a C++
1429 mangled name. The @var{type} argument is the tree structure representing
1430 the type to be mangled. The hook may be applied to trees which are
1431 not target-specific fundamental types; it should return @code{NULL}
1432 for all such types, as well as arguments it does not recognize. If the
1433 return value is not @code{NULL}, it must point to a statically-allocated
1436 Target-specific fundamental types might be new fundamental types or
1437 qualified versions of ordinary fundamental types. Encode new
1438 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1439 is the name used for the type in source code, and @var{n} is the
1440 length of @var{name} in decimal. Encode qualified versions of
1441 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1442 @var{name} is the name used for the type qualifier in source code,
1443 @var{n} is the length of @var{name} as above, and @var{code} is the
1444 code used to represent the unqualified version of this type. (See
1445 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1446 codes.) In both cases the spaces are for clarity; do not include any
1447 spaces in your string.
1449 This hook is applied to types prior to typedef resolution. If the mangled
1450 name for a particular type depends only on that type's main variant, you
1451 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1454 The default version of this hook always returns @code{NULL}, which is
1455 appropriate for a target that does not define any new fundamental
1460 @section Layout of Source Language Data Types
1462 These macros define the sizes and other characteristics of the standard
1463 basic data types used in programs being compiled. Unlike the macros in
1464 the previous section, these apply to specific features of C and related
1465 languages, rather than to fundamental aspects of storage layout.
1467 @defmac INT_TYPE_SIZE
1468 A C expression for the size in bits of the type @code{int} on the
1469 target machine. If you don't define this, the default is one word.
1472 @defmac SHORT_TYPE_SIZE
1473 A C expression for the size in bits of the type @code{short} on the
1474 target machine. If you don't define this, the default is half a word.
1475 (If this would be less than one storage unit, it is rounded up to one
1479 @defmac LONG_TYPE_SIZE
1480 A C expression for the size in bits of the type @code{long} on the
1481 target machine. If you don't define this, the default is one word.
1484 @defmac ADA_LONG_TYPE_SIZE
1485 On some machines, the size used for the Ada equivalent of the type
1486 @code{long} by a native Ada compiler differs from that used by C@. In
1487 that situation, define this macro to be a C expression to be used for
1488 the size of that type. If you don't define this, the default is the
1489 value of @code{LONG_TYPE_SIZE}.
1492 @defmac LONG_LONG_TYPE_SIZE
1493 A C expression for the size in bits of the type @code{long long} on the
1494 target machine. If you don't define this, the default is two
1495 words. If you want to support GNU Ada on your machine, the value of this
1496 macro must be at least 64.
1499 @defmac CHAR_TYPE_SIZE
1500 A C expression for the size in bits of the type @code{char} on the
1501 target machine. If you don't define this, the default is
1502 @code{BITS_PER_UNIT}.
1505 @defmac BOOL_TYPE_SIZE
1506 A C expression for the size in bits of the C++ type @code{bool} and
1507 C99 type @code{_Bool} on the target machine. If you don't define
1508 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1511 @defmac FLOAT_TYPE_SIZE
1512 A C expression for the size in bits of the type @code{float} on the
1513 target machine. If you don't define this, the default is one word.
1516 @defmac DOUBLE_TYPE_SIZE
1517 A C expression for the size in bits of the type @code{double} on the
1518 target machine. If you don't define this, the default is two
1522 @defmac LONG_DOUBLE_TYPE_SIZE
1523 A C expression for the size in bits of the type @code{long double} on
1524 the target machine. If you don't define this, the default is two
1528 @defmac SHORT_FRACT_TYPE_SIZE
1529 A C expression for the size in bits of the type @code{short _Fract} on
1530 the target machine. If you don't define this, the default is
1531 @code{BITS_PER_UNIT}.
1534 @defmac FRACT_TYPE_SIZE
1535 A C expression for the size in bits of the type @code{_Fract} on
1536 the target machine. If you don't define this, the default is
1537 @code{BITS_PER_UNIT * 2}.
1540 @defmac LONG_FRACT_TYPE_SIZE
1541 A C expression for the size in bits of the type @code{long _Fract} on
1542 the target machine. If you don't define this, the default is
1543 @code{BITS_PER_UNIT * 4}.
1546 @defmac LONG_LONG_FRACT_TYPE_SIZE
1547 A C expression for the size in bits of the type @code{long long _Fract} on
1548 the target machine. If you don't define this, the default is
1549 @code{BITS_PER_UNIT * 8}.
1552 @defmac SHORT_ACCUM_TYPE_SIZE
1553 A C expression for the size in bits of the type @code{short _Accum} on
1554 the target machine. If you don't define this, the default is
1555 @code{BITS_PER_UNIT * 2}.
1558 @defmac ACCUM_TYPE_SIZE
1559 A C expression for the size in bits of the type @code{_Accum} on
1560 the target machine. If you don't define this, the default is
1561 @code{BITS_PER_UNIT * 4}.
1564 @defmac LONG_ACCUM_TYPE_SIZE
1565 A C expression for the size in bits of the type @code{long _Accum} on
1566 the target machine. If you don't define this, the default is
1567 @code{BITS_PER_UNIT * 8}.
1570 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1571 A C expression for the size in bits of the type @code{long long _Accum} on
1572 the target machine. If you don't define this, the default is
1573 @code{BITS_PER_UNIT * 16}.
1576 @defmac LIBGCC2_GNU_PREFIX
1577 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1578 hook and should be defined if that hook is overriden to be true. It
1579 causes function names in libgcc to be changed to use a @code{__gnu_}
1580 prefix for their name rather than the default @code{__}. A port which
1581 uses this macro should also arrange to use @file{t-gnu-prefix} in
1582 the libgcc @file{config.host}.
1585 @defmac WIDEST_HARDWARE_FP_SIZE
1586 A C expression for the size in bits of the widest floating-point format
1587 supported by the hardware. If you define this macro, you must specify a
1588 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1589 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1593 @defmac DEFAULT_SIGNED_CHAR
1594 An expression whose value is 1 or 0, according to whether the type
1595 @code{char} should be signed or unsigned by default. The user can
1596 always override this default with the options @option{-fsigned-char}
1597 and @option{-funsigned-char}.
1600 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1601 This target hook should return true if the compiler should give an
1602 @code{enum} type only as many bytes as it takes to represent the range
1603 of possible values of that type. It should return false if all
1604 @code{enum} types should be allocated like @code{int}.
1606 The default is to return false.
1610 A C expression for a string describing the name of the data type to use
1611 for size values. The typedef name @code{size_t} is defined using the
1612 contents of the string.
1614 The string can contain more than one keyword. If so, separate them with
1615 spaces, and write first any length keyword, then @code{unsigned} if
1616 appropriate, and finally @code{int}. The string must exactly match one
1617 of the data type names defined in the function
1618 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1619 You may not omit @code{int} or change the order---that would cause the
1620 compiler to crash on startup.
1622 If you don't define this macro, the default is @code{"long unsigned
1627 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1628 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1629 dealing with size. This macro is a C expression for a string describing
1630 the name of the data type from which the precision of @code{sizetype}
1633 The string has the same restrictions as @code{SIZE_TYPE} string.
1635 If you don't define this macro, the default is @code{SIZE_TYPE}.
1638 @defmac PTRDIFF_TYPE
1639 A C expression for a string describing the name of the data type to use
1640 for the result of subtracting two pointers. The typedef name
1641 @code{ptrdiff_t} is defined using the contents of the string. See
1642 @code{SIZE_TYPE} above for more information.
1644 If you don't define this macro, the default is @code{"long int"}.
1648 A C expression for a string describing the name of the data type to use
1649 for wide characters. The typedef name @code{wchar_t} is defined using
1650 the contents of the string. See @code{SIZE_TYPE} above for more
1653 If you don't define this macro, the default is @code{"int"}.
1656 @defmac WCHAR_TYPE_SIZE
1657 A C expression for the size in bits of the data type for wide
1658 characters. This is used in @code{cpp}, which cannot make use of
1663 A C expression for a string describing the name of the data type to
1664 use for wide characters passed to @code{printf} and returned from
1665 @code{getwc}. The typedef name @code{wint_t} is defined using the
1666 contents of the string. See @code{SIZE_TYPE} above for more
1669 If you don't define this macro, the default is @code{"unsigned int"}.
1673 A C expression for a string describing the name of the data type that
1674 can represent any value of any standard or extended signed integer type.
1675 The typedef name @code{intmax_t} is defined using the contents of the
1676 string. See @code{SIZE_TYPE} above for more information.
1678 If you don't define this macro, the default is the first of
1679 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1680 much precision as @code{long long int}.
1683 @defmac UINTMAX_TYPE
1684 A C expression for a string describing the name of the data type that
1685 can represent any value of any standard or extended unsigned integer
1686 type. The typedef name @code{uintmax_t} is defined using the contents
1687 of the string. See @code{SIZE_TYPE} above for more information.
1689 If you don't define this macro, the default is the first of
1690 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1691 unsigned int"} that has as much precision as @code{long long unsigned
1695 @defmac SIG_ATOMIC_TYPE
1701 @defmacx UINT16_TYPE
1702 @defmacx UINT32_TYPE
1703 @defmacx UINT64_TYPE
1704 @defmacx INT_LEAST8_TYPE
1705 @defmacx INT_LEAST16_TYPE
1706 @defmacx INT_LEAST32_TYPE
1707 @defmacx INT_LEAST64_TYPE
1708 @defmacx UINT_LEAST8_TYPE
1709 @defmacx UINT_LEAST16_TYPE
1710 @defmacx UINT_LEAST32_TYPE
1711 @defmacx UINT_LEAST64_TYPE
1712 @defmacx INT_FAST8_TYPE
1713 @defmacx INT_FAST16_TYPE
1714 @defmacx INT_FAST32_TYPE
1715 @defmacx INT_FAST64_TYPE
1716 @defmacx UINT_FAST8_TYPE
1717 @defmacx UINT_FAST16_TYPE
1718 @defmacx UINT_FAST32_TYPE
1719 @defmacx UINT_FAST64_TYPE
1720 @defmacx INTPTR_TYPE
1721 @defmacx UINTPTR_TYPE
1722 C expressions for the standard types @code{sig_atomic_t},
1723 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1724 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1725 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1726 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1727 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1728 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1729 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1730 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1731 @code{SIZE_TYPE} above for more information.
1733 If any of these macros evaluates to a null pointer, the corresponding
1734 type is not supported; if GCC is configured to provide
1735 @code{<stdint.h>} in such a case, the header provided may not conform
1736 to C99, depending on the type in question. The defaults for all of
1737 these macros are null pointers.
1740 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1741 The C++ compiler represents a pointer-to-member-function with a struct
1748 ptrdiff_t vtable_index;
1755 The C++ compiler must use one bit to indicate whether the function that
1756 will be called through a pointer-to-member-function is virtual.
1757 Normally, we assume that the low-order bit of a function pointer must
1758 always be zero. Then, by ensuring that the vtable_index is odd, we can
1759 distinguish which variant of the union is in use. But, on some
1760 platforms function pointers can be odd, and so this doesn't work. In
1761 that case, we use the low-order bit of the @code{delta} field, and shift
1762 the remainder of the @code{delta} field to the left.
1764 GCC will automatically make the right selection about where to store
1765 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1766 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1767 set such that functions always start at even addresses, but the lowest
1768 bit of pointers to functions indicate whether the function at that
1769 address is in ARM or Thumb mode. If this is the case of your
1770 architecture, you should define this macro to
1771 @code{ptrmemfunc_vbit_in_delta}.
1773 In general, you should not have to define this macro. On architectures
1774 in which function addresses are always even, according to
1775 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1776 @code{ptrmemfunc_vbit_in_pfn}.
1779 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1780 Normally, the C++ compiler uses function pointers in vtables. This
1781 macro allows the target to change to use ``function descriptors''
1782 instead. Function descriptors are found on targets for whom a
1783 function pointer is actually a small data structure. Normally the
1784 data structure consists of the actual code address plus a data
1785 pointer to which the function's data is relative.
1787 If vtables are used, the value of this macro should be the number
1788 of words that the function descriptor occupies.
1791 @defmac TARGET_VTABLE_ENTRY_ALIGN
1792 By default, the vtable entries are void pointers, the so the alignment
1793 is the same as pointer alignment. The value of this macro specifies
1794 the alignment of the vtable entry in bits. It should be defined only
1795 when special alignment is necessary. */
1798 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1799 There are a few non-descriptor entries in the vtable at offsets below
1800 zero. If these entries must be padded (say, to preserve the alignment
1801 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1802 of words in each data entry.
1806 @section Register Usage
1807 @cindex register usage
1809 This section explains how to describe what registers the target machine
1810 has, and how (in general) they can be used.
1812 The description of which registers a specific instruction can use is
1813 done with register classes; see @ref{Register Classes}. For information
1814 on using registers to access a stack frame, see @ref{Frame Registers}.
1815 For passing values in registers, see @ref{Register Arguments}.
1816 For returning values in registers, see @ref{Scalar Return}.
1819 * Register Basics:: Number and kinds of registers.
1820 * Allocation Order:: Order in which registers are allocated.
1821 * Values in Registers:: What kinds of values each reg can hold.
1822 * Leaf Functions:: Renumbering registers for leaf functions.
1823 * Stack Registers:: Handling a register stack such as 80387.
1826 @node Register Basics
1827 @subsection Basic Characteristics of Registers
1829 @c prevent bad page break with this line
1830 Registers have various characteristics.
1832 @defmac FIRST_PSEUDO_REGISTER
1833 Number of hardware registers known to the compiler. They receive
1834 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1835 pseudo register's number really is assigned the number
1836 @code{FIRST_PSEUDO_REGISTER}.
1839 @defmac FIXED_REGISTERS
1840 @cindex fixed register
1841 An initializer that says which registers are used for fixed purposes
1842 all throughout the compiled code and are therefore not available for
1843 general allocation. These would include the stack pointer, the frame
1844 pointer (except on machines where that can be used as a general
1845 register when no frame pointer is needed), the program counter on
1846 machines where that is considered one of the addressable registers,
1847 and any other numbered register with a standard use.
1849 This information is expressed as a sequence of numbers, separated by
1850 commas and surrounded by braces. The @var{n}th number is 1 if
1851 register @var{n} is fixed, 0 otherwise.
1853 The table initialized from this macro, and the table initialized by
1854 the following one, may be overridden at run time either automatically,
1855 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1856 the user with the command options @option{-ffixed-@var{reg}},
1857 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1860 @defmac CALL_USED_REGISTERS
1861 @cindex call-used register
1862 @cindex call-clobbered register
1863 @cindex call-saved register
1864 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1865 clobbered (in general) by function calls as well as for fixed
1866 registers. This macro therefore identifies the registers that are not
1867 available for general allocation of values that must live across
1870 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1871 automatically saves it on function entry and restores it on function
1872 exit, if the register is used within the function.
1874 Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS}
1875 must be defined. Modern ports should define @code{CALL_REALLY_USED_REGISTERS}.
1878 @defmac CALL_REALLY_USED_REGISTERS
1879 @cindex call-used register
1880 @cindex call-clobbered register
1881 @cindex call-saved register
1882 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1883 that the entire set of @code{FIXED_REGISTERS} be included.
1884 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1886 Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS}
1887 must be defined. Modern ports should define @code{CALL_REALLY_USED_REGISTERS}.
1890 @cindex call-used register
1891 @cindex call-clobbered register
1892 @cindex call-saved register
1893 @deftypefn {Target Hook} {const predefined_function_abi &} TARGET_FNTYPE_ABI (const_tree @var{type})
1894 Return the ABI used by a function with type @var{type}; see the
1895 definition of @code{predefined_function_abi} for details of the ABI
1896 descriptor. Targets only need to define this hook if they support
1897 interoperability between several ABIs in the same translation unit.
1900 @deftypefn {Target Hook} {const predefined_function_abi &} TARGET_INSN_CALLEE_ABI (const rtx_insn *@var{insn})
1901 This hook returns a description of the ABI used by the target of
1902 call instruction @var{insn}; see the definition of
1903 @code{predefined_function_abi} for details of the ABI descriptor.
1904 Only the global function @code{insn_callee_abi} should call this hook
1907 Targets only need to define this hook if they support
1908 interoperability between several ABIs in the same translation unit.
1911 @cindex call-used register
1912 @cindex call-clobbered register
1913 @cindex call-saved register
1914 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_CALL_PART_CLOBBERED (unsigned int @var{abi_id}, unsigned int @var{regno}, machine_mode @var{mode})
1915 ABIs usually specify that calls must preserve the full contents
1916 of a particular register, or that calls can alter any part of a
1917 particular register. This information is captured by the target macro
1918 @code{CALL_REALLY_USED_REGISTERS}. However, some ABIs specify that calls
1919 must preserve certain bits of a particular register but can alter others.
1920 This hook should return true if this applies to at least one of the
1921 registers in @samp{(reg:@var{mode} @var{regno})}, and if as a result the
1922 call would alter part of the @var{mode} value. For example, if a call
1923 preserves the low 32 bits of a 64-bit hard register @var{regno} but can
1924 clobber the upper 32 bits, this hook should return true for a 64-bit mode
1925 but false for a 32-bit mode.
1927 The value of @var{abi_id} comes from the @code{predefined_function_abi}
1928 structure that describes the ABI of the call; see the definition of the
1929 structure for more details. If (as is usual) the target uses the same ABI
1930 for all functions in a translation unit, @var{abi_id} is always 0.
1932 The default implementation returns false, which is correct
1933 for targets that don't have partly call-clobbered registers.
1936 @deftypefn {Target Hook} {const char *} TARGET_GET_MULTILIB_ABI_NAME (void)
1937 This hook returns name of multilib ABI name.
1941 @findex call_used_regs
1944 @findex reg_class_contents
1945 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1946 This hook may conditionally modify five variables
1947 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1948 @code{reg_names}, and @code{reg_class_contents}, to take into account
1949 any dependence of these register sets on target flags. The first three
1950 of these are of type @code{char []} (interpreted as boolean vectors).
1951 @code{global_regs} is a @code{const char *[]}, and
1952 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1953 called, @code{fixed_regs}, @code{call_used_regs},
1954 @code{reg_class_contents}, and @code{reg_names} have been initialized
1955 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1956 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1957 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1958 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1959 command options have been applied.
1961 @cindex disabling certain registers
1962 @cindex controlling register usage
1963 If the usage of an entire class of registers depends on the target
1964 flags, you may indicate this to GCC by using this macro to modify
1965 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1966 registers in the classes which should not be used by GCC@. Also make
1967 @code{define_register_constraint}s return @code{NO_REGS} for constraints
1968 that shouldn't be used.
1970 (However, if this class is not included in @code{GENERAL_REGS} and all
1971 of the insn patterns whose constraints permit this class are
1972 controlled by target switches, then GCC will automatically avoid using
1973 these registers when the target switches are opposed to them.)
1976 @defmac INCOMING_REGNO (@var{out})
1977 Define this macro if the target machine has register windows. This C
1978 expression returns the register number as seen by the called function
1979 corresponding to the register number @var{out} as seen by the calling
1980 function. Return @var{out} if register number @var{out} is not an
1984 @defmac OUTGOING_REGNO (@var{in})
1985 Define this macro if the target machine has register windows. This C
1986 expression returns the register number as seen by the calling function
1987 corresponding to the register number @var{in} as seen by the called
1988 function. Return @var{in} if register number @var{in} is not an inbound
1992 @defmac LOCAL_REGNO (@var{regno})
1993 Define this macro if the target machine has register windows. This C
1994 expression returns true if the register is call-saved but is in the
1995 register window. Unlike most call-saved registers, such registers
1996 need not be explicitly restored on function exit or during non-local
2001 If the program counter has a register number, define this as that
2002 register number. Otherwise, do not define it.
2005 @node Allocation Order
2006 @subsection Order of Allocation of Registers
2007 @cindex order of register allocation
2008 @cindex register allocation order
2010 @c prevent bad page break with this line
2011 Registers are allocated in order.
2013 @defmac REG_ALLOC_ORDER
2014 If defined, an initializer for a vector of integers, containing the
2015 numbers of hard registers in the order in which GCC should prefer
2016 to use them (from most preferred to least).
2018 If this macro is not defined, registers are used lowest numbered first
2019 (all else being equal).
2021 One use of this macro is on machines where the highest numbered
2022 registers must always be saved and the save-multiple-registers
2023 instruction supports only sequences of consecutive registers. On such
2024 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2025 the highest numbered allocable register first.
2028 @defmac ADJUST_REG_ALLOC_ORDER
2029 A C statement (sans semicolon) to choose the order in which to allocate
2030 hard registers for pseudo-registers local to a basic block.
2032 Store the desired register order in the array @code{reg_alloc_order}.
2033 Element 0 should be the register to allocate first; element 1, the next
2034 register; and so on.
2036 The macro body should not assume anything about the contents of
2037 @code{reg_alloc_order} before execution of the macro.
2039 On most machines, it is not necessary to define this macro.
2042 @defmac HONOR_REG_ALLOC_ORDER
2043 Normally, IRA tries to estimate the costs for saving a register in the
2044 prologue and restoring it in the epilogue. This discourages it from
2045 using call-saved registers. If a machine wants to ensure that IRA
2046 allocates registers in the order given by REG_ALLOC_ORDER even if some
2047 call-saved registers appear earlier than call-used ones, then define this
2048 macro as a C expression to nonzero. Default is 0.
2051 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2052 In some case register allocation order is not enough for the
2053 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2054 If this macro is defined, it should return a floating point value
2055 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2056 be increased by approximately the pseudo's usage frequency times the
2057 value returned by this macro. Not defining this macro is equivalent
2058 to having it always return @code{0.0}.
2060 On most machines, it is not necessary to define this macro.
2063 @node Values in Registers
2064 @subsection How Values Fit in Registers
2066 This section discusses the macros that describe which kinds of values
2067 (specifically, which machine modes) each register can hold, and how many
2068 consecutive registers are needed for a given mode.
2070 @deftypefn {Target Hook} {unsigned int} TARGET_HARD_REGNO_NREGS (unsigned int @var{regno}, machine_mode @var{mode})
2071 This hook returns the number of consecutive hard registers, starting
2072 at register number @var{regno}, required to hold a value of mode
2073 @var{mode}. This hook must never return zero, even if a register
2074 cannot hold the requested mode - indicate that with
2075 @code{TARGET_HARD_REGNO_MODE_OK} and/or
2076 @code{TARGET_CAN_CHANGE_MODE_CLASS} instead.
2078 The default definition returns the number of words in @var{mode}.
2081 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2082 A C expression that is nonzero if a value of mode @var{mode}, stored
2083 in memory, ends with padding that causes it to take up more space than
2084 in registers starting at register number @var{regno} (as determined by
2085 multiplying GCC's notion of the size of the register when containing
2086 this mode by the number of registers returned by
2087 @code{TARGET_HARD_REGNO_NREGS}). By default this is zero.
2089 For example, if a floating-point value is stored in three 32-bit
2090 registers but takes up 128 bits in memory, then this would be
2093 This macros only needs to be defined if there are cases where
2094 @code{subreg_get_info}
2095 would otherwise wrongly determine that a @code{subreg} can be
2096 represented by an offset to the register number, when in fact such a
2097 @code{subreg} would contain some of the padding not stored in
2098 registers and so not be representable.
2101 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2102 For values of @var{regno} and @var{mode} for which
2103 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2104 returning the greater number of registers required to hold the value
2105 including any padding. In the example above, the value would be four.
2108 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2109 Define this macro if the natural size of registers that hold values
2110 of mode @var{mode} is not the word size. It is a C expression that
2111 should give the natural size in bytes for the specified mode. It is
2112 used by the register allocator to try to optimize its results. This
2113 happens for example on SPARC 64-bit where the natural size of
2114 floating-point registers is still 32-bit.
2117 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_MODE_OK (unsigned int @var{regno}, machine_mode @var{mode})
2118 This hook returns true if it is permissible to store a value
2119 of mode @var{mode} in hard register number @var{regno} (or in several
2120 registers starting with that one). The default definition returns true
2123 You need not include code to check for the numbers of fixed registers,
2124 because the allocation mechanism considers them to be always occupied.
2126 @cindex register pairs
2127 On some machines, double-precision values must be kept in even/odd
2128 register pairs. You can implement that by defining this hook to reject
2129 odd register numbers for such modes.
2131 The minimum requirement for a mode to be OK in a register is that the
2132 @samp{mov@var{mode}} instruction pattern support moves between the
2133 register and other hard register in the same class and that moving a
2134 value into the register and back out not alter it.
2136 Since the same instruction used to move @code{word_mode} will work for
2137 all narrower integer modes, it is not necessary on any machine for
2138 this hook to distinguish between these modes, provided you define
2139 patterns @samp{movhi}, etc., to take advantage of this. This is
2140 useful because of the interaction between @code{TARGET_HARD_REGNO_MODE_OK}
2141 and @code{TARGET_MODES_TIEABLE_P}; it is very desirable for all integer
2142 modes to be tieable.
2144 Many machines have special registers for floating point arithmetic.
2145 Often people assume that floating point machine modes are allowed only
2146 in floating point registers. This is not true. Any registers that
2147 can hold integers can safely @emph{hold} a floating point machine
2148 mode, whether or not floating arithmetic can be done on it in those
2149 registers. Integer move instructions can be used to move the values.
2151 On some machines, though, the converse is true: fixed-point machine
2152 modes may not go in floating registers. This is true if the floating
2153 registers normalize any value stored in them, because storing a
2154 non-floating value there would garble it. In this case,
2155 @code{TARGET_HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2156 floating registers. But if the floating registers do not automatically
2157 normalize, if you can store any bit pattern in one and retrieve it
2158 unchanged without a trap, then any machine mode may go in a floating
2159 register, so you can define this hook to say so.
2161 The primary significance of special floating registers is rather that
2162 they are the registers acceptable in floating point arithmetic
2163 instructions. However, this is of no concern to
2164 @code{TARGET_HARD_REGNO_MODE_OK}. You handle it by writing the proper
2165 constraints for those instructions.
2167 On some machines, the floating registers are especially slow to access,
2168 so that it is better to store a value in a stack frame than in such a
2169 register if floating point arithmetic is not being done. As long as the
2170 floating registers are not in class @code{GENERAL_REGS}, they will not
2171 be used unless some pattern's constraint asks for one.
2174 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2175 A C expression that is nonzero if it is OK to rename a hard register
2176 @var{from} to another hard register @var{to}.
2178 One common use of this macro is to prevent renaming of a register to
2179 another register that is not saved by a prologue in an interrupt
2182 The default is always nonzero.
2185 @deftypefn {Target Hook} bool TARGET_MODES_TIEABLE_P (machine_mode @var{mode1}, machine_mode @var{mode2})
2186 This hook returns true if a value of mode @var{mode1} is accessible
2187 in mode @var{mode2} without copying.
2189 If @code{TARGET_HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2190 @code{TARGET_HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always
2191 the same for any @var{r}, then
2192 @code{TARGET_MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2193 should be true. If they differ for any @var{r}, you should define
2194 this hook to return false unless some other mechanism ensures the
2195 accessibility of the value in a narrower mode.
2197 You should define this hook to return true in as many cases as
2198 possible since doing so will allow GCC to perform better register
2199 allocation. The default definition returns true unconditionally.
2202 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2203 This target hook should return @code{true} if it is OK to use a hard register
2204 @var{regno} as scratch reg in peephole2.
2206 One common use of this macro is to prevent using of a register that
2207 is not saved by a prologue in an interrupt handler.
2209 The default version of this hook always returns @code{true}.
2212 @defmac AVOID_CCMODE_COPIES
2213 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2214 registers. You should only define this macro if support for copying to/from
2215 @code{CCmode} is incomplete.
2218 @node Leaf Functions
2219 @subsection Handling Leaf Functions
2221 @cindex leaf functions
2222 @cindex functions, leaf
2223 On some machines, a leaf function (i.e., one which makes no calls) can run
2224 more efficiently if it does not make its own register window. Often this
2225 means it is required to receive its arguments in the registers where they
2226 are passed by the caller, instead of the registers where they would
2229 The special treatment for leaf functions generally applies only when
2230 other conditions are met; for example, often they may use only those
2231 registers for its own variables and temporaries. We use the term ``leaf
2232 function'' to mean a function that is suitable for this special
2233 handling, so that functions with no calls are not necessarily ``leaf
2236 GCC assigns register numbers before it knows whether the function is
2237 suitable for leaf function treatment. So it needs to renumber the
2238 registers in order to output a leaf function. The following macros
2241 @defmac LEAF_REGISTERS
2242 Name of a char vector, indexed by hard register number, which
2243 contains 1 for a register that is allowable in a candidate for leaf
2246 If leaf function treatment involves renumbering the registers, then the
2247 registers marked here should be the ones before renumbering---those that
2248 GCC would ordinarily allocate. The registers which will actually be
2249 used in the assembler code, after renumbering, should not be marked with 1
2252 Define this macro only if the target machine offers a way to optimize
2253 the treatment of leaf functions.
2256 @defmac LEAF_REG_REMAP (@var{regno})
2257 A C expression whose value is the register number to which @var{regno}
2258 should be renumbered, when a function is treated as a leaf function.
2260 If @var{regno} is a register number which should not appear in a leaf
2261 function before renumbering, then the expression should yield @minus{}1, which
2262 will cause the compiler to abort.
2264 Define this macro only if the target machine offers a way to optimize the
2265 treatment of leaf functions, and registers need to be renumbered to do
2269 @findex current_function_is_leaf
2270 @findex current_function_uses_only_leaf_regs
2271 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2272 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2273 specially. They can test the C variable @code{current_function_is_leaf}
2274 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2275 set prior to local register allocation and is valid for the remaining
2276 compiler passes. They can also test the C variable
2277 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2278 functions which only use leaf registers.
2279 @code{current_function_uses_only_leaf_regs} is valid after all passes
2280 that modify the instructions have been run and is only useful if
2281 @code{LEAF_REGISTERS} is defined.
2282 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2283 @c of the next paragraph?! --mew 2feb93
2285 @node Stack Registers
2286 @subsection Registers That Form a Stack
2288 There are special features to handle computers where some of the
2289 ``registers'' form a stack. Stack registers are normally written by
2290 pushing onto the stack, and are numbered relative to the top of the
2293 Currently, GCC can only handle one group of stack-like registers, and
2294 they must be consecutively numbered. Furthermore, the existing
2295 support for stack-like registers is specific to the 80387 floating
2296 point coprocessor. If you have a new architecture that uses
2297 stack-like registers, you will need to do substantial work on
2298 @file{reg-stack.c} and write your machine description to cooperate
2299 with it, as well as defining these macros.
2302 Define this if the machine has any stack-like registers.
2305 @defmac STACK_REG_COVER_CLASS
2306 This is a cover class containing the stack registers. Define this if
2307 the machine has any stack-like registers.
2310 @defmac FIRST_STACK_REG
2311 The number of the first stack-like register. This one is the top
2315 @defmac LAST_STACK_REG
2316 The number of the last stack-like register. This one is the bottom of
2320 @node Register Classes
2321 @section Register Classes
2322 @cindex register class definitions
2323 @cindex class definitions, register
2325 On many machines, the numbered registers are not all equivalent.
2326 For example, certain registers may not be allowed for indexed addressing;
2327 certain registers may not be allowed in some instructions. These machine
2328 restrictions are described to the compiler using @dfn{register classes}.
2330 You define a number of register classes, giving each one a name and saying
2331 which of the registers belong to it. Then you can specify register classes
2332 that are allowed as operands to particular instruction patterns.
2336 In general, each register will belong to several classes. In fact, one
2337 class must be named @code{ALL_REGS} and contain all the registers. Another
2338 class must be named @code{NO_REGS} and contain no registers. Often the
2339 union of two classes will be another class; however, this is not required.
2341 @findex GENERAL_REGS
2342 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2343 terribly special about the name, but the operand constraint letters
2344 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2345 the same as @code{ALL_REGS}, just define it as a macro which expands
2348 Order the classes so that if class @var{x} is contained in class @var{y}
2349 then @var{x} has a lower class number than @var{y}.
2351 The way classes other than @code{GENERAL_REGS} are specified in operand
2352 constraints is through machine-dependent operand constraint letters.
2353 You can define such letters to correspond to various classes, then use
2354 them in operand constraints.
2356 You must define the narrowest register classes for allocatable
2357 registers, so that each class either has no subclasses, or that for
2358 some mode, the move cost between registers within the class is
2359 cheaper than moving a register in the class to or from memory
2362 You should define a class for the union of two classes whenever some
2363 instruction allows both classes. For example, if an instruction allows
2364 either a floating point (coprocessor) register or a general register for a
2365 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2366 which includes both of them. Otherwise you will get suboptimal code,
2367 or even internal compiler errors when reload cannot find a register in the
2368 class computed via @code{reg_class_subunion}.
2370 You must also specify certain redundant information about the register
2371 classes: for each class, which classes contain it and which ones are
2372 contained in it; for each pair of classes, the largest class contained
2375 When a value occupying several consecutive registers is expected in a
2376 certain class, all the registers used must belong to that class.
2377 Therefore, register classes cannot be used to enforce a requirement for
2378 a register pair to start with an even-numbered register. The way to
2379 specify this requirement is with @code{TARGET_HARD_REGNO_MODE_OK}.
2381 Register classes used for input-operands of bitwise-and or shift
2382 instructions have a special requirement: each such class must have, for
2383 each fixed-point machine mode, a subclass whose registers can transfer that
2384 mode to or from memory. For example, on some machines, the operations for
2385 single-byte values (@code{QImode}) are limited to certain registers. When
2386 this is so, each register class that is used in a bitwise-and or shift
2387 instruction must have a subclass consisting of registers from which
2388 single-byte values can be loaded or stored. This is so that
2389 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2391 @deftp {Data type} {enum reg_class}
2392 An enumerated type that must be defined with all the register class names
2393 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2394 must be the last register class, followed by one more enumerated value,
2395 @code{LIM_REG_CLASSES}, which is not a register class but rather
2396 tells how many classes there are.
2398 Each register class has a number, which is the value of casting
2399 the class name to type @code{int}. The number serves as an index
2400 in many of the tables described below.
2403 @defmac N_REG_CLASSES
2404 The number of distinct register classes, defined as follows:
2407 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2411 @defmac REG_CLASS_NAMES
2412 An initializer containing the names of the register classes as C string
2413 constants. These names are used in writing some of the debugging dumps.
2416 @defmac REG_CLASS_CONTENTS
2417 An initializer containing the contents of the register classes, as integers
2418 which are bit masks. The @var{n}th integer specifies the contents of class
2419 @var{n}. The way the integer @var{mask} is interpreted is that
2420 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2422 When the machine has more than 32 registers, an integer does not suffice.
2423 Then the integers are replaced by sub-initializers, braced groupings containing
2424 several integers. Each sub-initializer must be suitable as an initializer
2425 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2426 In this situation, the first integer in each sub-initializer corresponds to
2427 registers 0 through 31, the second integer to registers 32 through 63, and
2431 @defmac REGNO_REG_CLASS (@var{regno})
2432 A C expression whose value is a register class containing hard register
2433 @var{regno}. In general there is more than one such class; choose a class
2434 which is @dfn{minimal}, meaning that no smaller class also contains the
2438 @defmac BASE_REG_CLASS
2439 A macro whose definition is the name of the class to which a valid
2440 base register must belong. A base register is one used in an address
2441 which is the register value plus a displacement.
2444 @defmac MODE_BASE_REG_CLASS (@var{mode})
2445 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2446 the selection of a base register in a mode dependent manner. If
2447 @var{mode} is VOIDmode then it should return the same value as
2448 @code{BASE_REG_CLASS}.
2451 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2452 A C expression whose value is the register class to which a valid
2453 base register must belong in order to be used in a base plus index
2454 register address. You should define this macro if base plus index
2455 addresses have different requirements than other base register uses.
2458 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2459 A C expression whose value is the register class to which a valid
2460 base register for a memory reference in mode @var{mode} to address
2461 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2462 define the context in which the base register occurs. @var{outer_code} is
2463 the code of the immediately enclosing expression (@code{MEM} for the top level
2464 of an address, @code{ADDRESS} for something that occurs in an
2465 @code{address_operand}). @var{index_code} is the code of the corresponding
2466 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2469 @defmac INDEX_REG_CLASS
2470 A macro whose definition is the name of the class to which a valid
2471 index register must belong. An index register is one used in an
2472 address where its value is either multiplied by a scale factor or
2473 added to another register (as well as added to a displacement).
2476 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2477 A C expression which is nonzero if register number @var{num} is
2478 suitable for use as a base register in operand addresses.
2481 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2482 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2483 that expression may examine the mode of the memory reference in
2484 @var{mode}. You should define this macro if the mode of the memory
2485 reference affects whether a register may be used as a base register. If
2486 you define this macro, the compiler will use it instead of
2487 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2488 addresses that appear outside a @code{MEM}, i.e., as an
2489 @code{address_operand}.
2492 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2493 A C expression which is nonzero if register number @var{num} is suitable for
2494 use as a base register in base plus index operand addresses, accessing
2495 memory in mode @var{mode}. It may be either a suitable hard register or a
2496 pseudo register that has been allocated such a hard register. You should
2497 define this macro if base plus index addresses have different requirements
2498 than other base register uses.
2500 Use of this macro is deprecated; please use the more general
2501 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2504 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2505 A C expression which is nonzero if register number @var{num} is
2506 suitable for use as a base register in operand addresses, accessing
2507 memory in mode @var{mode} in address space @var{address_space}.
2508 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2509 that that expression may examine the context in which the register
2510 appears in the memory reference. @var{outer_code} is the code of the
2511 immediately enclosing expression (@code{MEM} if at the top level of the
2512 address, @code{ADDRESS} for something that occurs in an
2513 @code{address_operand}). @var{index_code} is the code of the
2514 corresponding index expression if @var{outer_code} is @code{PLUS};
2515 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2516 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2519 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2520 A C expression which is nonzero if register number @var{num} is
2521 suitable for use as an index register in operand addresses. It may be
2522 either a suitable hard register or a pseudo register that has been
2523 allocated such a hard register.
2525 The difference between an index register and a base register is that
2526 the index register may be scaled. If an address involves the sum of
2527 two registers, neither one of them scaled, then either one may be
2528 labeled the ``base'' and the other the ``index''; but whichever
2529 labeling is used must fit the machine's constraints of which registers
2530 may serve in each capacity. The compiler will try both labelings,
2531 looking for one that is valid, and will reload one or both registers
2532 only if neither labeling works.
2535 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2536 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.
2539 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2540 A target hook that places additional restrictions on the register class
2541 to use when it is necessary to copy value @var{x} into a register in class
2542 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2543 another, smaller class.
2545 The default version of this hook always returns value of @code{rclass} argument.
2547 Sometimes returning a more restrictive class makes better code. For
2548 example, on the 68000, when @var{x} is an integer constant that is in range
2549 for a @samp{moveq} instruction, the value of this macro is always
2550 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2551 Requiring a data register guarantees that a @samp{moveq} will be used.
2553 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2554 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2555 loaded into some register class. By returning @code{NO_REGS} you can
2556 force @var{x} into a memory location. For example, rs6000 can load
2557 immediate values into general-purpose registers, but does not have an
2558 instruction for loading an immediate value into a floating-point
2559 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2560 @var{x} is a floating-point constant. If the constant can't be loaded
2561 into any kind of register, code generation will be better if
2562 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2563 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2565 If an insn has pseudos in it after register allocation, reload will go
2566 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2567 to find the best one. Returning @code{NO_REGS}, in this case, makes
2568 reload add a @code{!} in front of the constraint: the x86 back-end uses
2569 this feature to discourage usage of 387 registers when math is done in
2570 the SSE registers (and vice versa).
2573 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2574 A C expression that places additional restrictions on the register class
2575 to use when it is necessary to copy value @var{x} into a register in class
2576 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2577 another, smaller class. On many machines, the following definition is
2581 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2584 Sometimes returning a more restrictive class makes better code. For
2585 example, on the 68000, when @var{x} is an integer constant that is in range
2586 for a @samp{moveq} instruction, the value of this macro is always
2587 @code{DATA_REGS} as long as @var{class} includes the data registers.
2588 Requiring a data register guarantees that a @samp{moveq} will be used.
2590 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2591 @var{class} is if @var{x} is a legitimate constant which cannot be
2592 loaded into some register class. By returning @code{NO_REGS} you can
2593 force @var{x} into a memory location. For example, rs6000 can load
2594 immediate values into general-purpose registers, but does not have an
2595 instruction for loading an immediate value into a floating-point
2596 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2597 @var{x} is a floating-point constant. If the constant cannot be loaded
2598 into any kind of register, code generation will be better if
2599 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2600 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2602 If an insn has pseudos in it after register allocation, reload will go
2603 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2604 to find the best one. Returning @code{NO_REGS}, in this case, makes
2605 reload add a @code{!} in front of the constraint: the x86 back-end uses
2606 this feature to discourage usage of 387 registers when math is done in
2607 the SSE registers (and vice versa).
2610 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2611 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2614 The default version of this hook always returns value of @code{rclass}
2617 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2618 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2621 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2622 A C expression that places additional restrictions on the register class
2623 to use when it is necessary to be able to hold a value of mode
2624 @var{mode} in a reload register for which class @var{class} would
2627 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2628 there are certain modes that simply cannot go in certain reload classes.
2630 The value is a register class; perhaps @var{class}, or perhaps another,
2633 Don't define this macro unless the target machine has limitations which
2634 require the macro to do something nontrivial.
2637 @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})
2638 Many machines have some registers that cannot be copied directly to or
2639 from memory or even from other types of registers. An example is the
2640 @samp{MQ} register, which on most machines, can only be copied to or
2641 from general registers, but not memory. Below, we shall be using the
2642 term 'intermediate register' when a move operation cannot be performed
2643 directly, but has to be done by copying the source into the intermediate
2644 register first, and then copying the intermediate register to the
2645 destination. An intermediate register always has the same mode as
2646 source and destination. Since it holds the actual value being copied,
2647 reload might apply optimizations to re-use an intermediate register
2648 and eliding the copy from the source when it can determine that the
2649 intermediate register still holds the required value.
2651 Another kind of secondary reload is required on some machines which
2652 allow copying all registers to and from memory, but require a scratch
2653 register for stores to some memory locations (e.g., those with symbolic
2654 address on the RT, and those with certain symbolic address on the SPARC
2655 when compiling PIC)@. Scratch registers need not have the same mode
2656 as the value being copied, and usually hold a different value than
2657 that being copied. Special patterns in the md file are needed to
2658 describe how the copy is performed with the help of the scratch register;
2659 these patterns also describe the number, register class(es) and mode(s)
2660 of the scratch register(s).
2662 In some cases, both an intermediate and a scratch register are required.
2664 For input reloads, this target hook is called with nonzero @var{in_p},
2665 and @var{x} is an rtx that needs to be copied to a register of class
2666 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2667 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2668 needs to be copied to rtx @var{x} in @var{reload_mode}.
2670 If copying a register of @var{reload_class} from/to @var{x} requires
2671 an intermediate register, the hook @code{secondary_reload} should
2672 return the register class required for this intermediate register.
2673 If no intermediate register is required, it should return NO_REGS.
2674 If more than one intermediate register is required, describe the one
2675 that is closest in the copy chain to the reload register.
2677 If scratch registers are needed, you also have to describe how to
2678 perform the copy from/to the reload register to/from this
2679 closest intermediate register. Or if no intermediate register is
2680 required, but still a scratch register is needed, describe the
2681 copy from/to the reload register to/from the reload operand @var{x}.
2683 You do this by setting @code{sri->icode} to the instruction code of a pattern
2684 in the md file which performs the move. Operands 0 and 1 are the output
2685 and input of this copy, respectively. Operands from operand 2 onward are
2686 for scratch operands. These scratch operands must have a mode, and a
2687 single-register-class
2688 @c [later: or memory]
2691 When an intermediate register is used, the @code{secondary_reload}
2692 hook will be called again to determine how to copy the intermediate
2693 register to/from the reload operand @var{x}, so your hook must also
2694 have code to handle the register class of the intermediate operand.
2696 @c [For later: maybe we'll allow multi-alternative reload patterns -
2697 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2698 @c and match the constraints of input and output to determine the required
2699 @c alternative. A restriction would be that constraints used to match
2700 @c against reloads registers would have to be written as register class
2701 @c constraints, or we need a new target macro / hook that tells us if an
2702 @c arbitrary constraint can match an unknown register of a given class.
2703 @c Such a macro / hook would also be useful in other places.]
2706 @var{x} might be a pseudo-register or a @code{subreg} of a
2707 pseudo-register, which could either be in a hard register or in memory.
2708 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2709 in memory and the hard register number if it is in a register.
2711 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2712 currently not supported. For the time being, you will have to continue
2713 to use @code{TARGET_SECONDARY_MEMORY_NEEDED} for that purpose.
2715 @code{copy_cost} also uses this target hook to find out how values are
2716 copied. If you want it to include some extra cost for the need to allocate
2717 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2718 Or if two dependent moves are supposed to have a lower cost than the sum
2719 of the individual moves due to expected fortuitous scheduling and/or special
2720 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2723 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2724 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2725 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2726 These macros are obsolete, new ports should use the target hook
2727 @code{TARGET_SECONDARY_RELOAD} instead.
2729 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2730 target hook. Older ports still define these macros to indicate to the
2731 reload phase that it may
2732 need to allocate at least one register for a reload in addition to the
2733 register to contain the data. Specifically, if copying @var{x} to a
2734 register @var{class} in @var{mode} requires an intermediate register,
2735 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2736 largest register class all of whose registers can be used as
2737 intermediate registers or scratch registers.
2739 If copying a register @var{class} in @var{mode} to @var{x} requires an
2740 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2741 was supposed to be defined be defined to return the largest register
2742 class required. If the
2743 requirements for input and output reloads were the same, the macro
2744 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2747 The values returned by these macros are often @code{GENERAL_REGS}.
2748 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2749 can be directly copied to or from a register of @var{class} in
2750 @var{mode} without requiring a scratch register. Do not define this
2751 macro if it would always return @code{NO_REGS}.
2753 If a scratch register is required (either with or without an
2754 intermediate register), you were supposed to define patterns for
2755 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2756 (@pxref{Standard Names}. These patterns, which were normally
2757 implemented with a @code{define_expand}, should be similar to the
2758 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2761 These patterns need constraints for the reload register and scratch
2763 contain a single register class. If the original reload register (whose
2764 class is @var{class}) can meet the constraint given in the pattern, the
2765 value returned by these macros is used for the class of the scratch
2766 register. Otherwise, two additional reload registers are required.
2767 Their classes are obtained from the constraints in the insn pattern.
2769 @var{x} might be a pseudo-register or a @code{subreg} of a
2770 pseudo-register, which could either be in a hard register or in memory.
2771 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2772 in memory and the hard register number if it is in a register.
2774 These macros should not be used in the case where a particular class of
2775 registers can only be copied to memory and not to another class of
2776 registers. In that case, secondary reload registers are not needed and
2777 would not be helpful. Instead, a stack location must be used to perform
2778 the copy and the @code{mov@var{m}} pattern should use memory as an
2779 intermediate storage. This case often occurs between floating-point and
2783 @deftypefn {Target Hook} bool TARGET_SECONDARY_MEMORY_NEEDED (machine_mode @var{mode}, reg_class_t @var{class1}, reg_class_t @var{class2})
2784 Certain machines have the property that some registers cannot be copied
2785 to some other registers without using memory. Define this hook on
2786 those machines to return true if objects of mode @var{m} in registers
2787 of @var{class1} can only be copied to registers of class @var{class2} by
2788 storing a register of @var{class1} into memory and loading that memory
2789 location into a register of @var{class2}. The default definition returns
2790 false for all inputs.
2793 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2794 Normally when @code{TARGET_SECONDARY_MEMORY_NEEDED} is defined, the compiler
2795 allocates a stack slot for a memory location needed for register copies.
2796 If this macro is defined, the compiler instead uses the memory location
2797 defined by this macro.
2799 Do not define this macro if you do not define
2800 @code{TARGET_SECONDARY_MEMORY_NEEDED}.
2803 @deftypefn {Target Hook} machine_mode TARGET_SECONDARY_MEMORY_NEEDED_MODE (machine_mode @var{mode})
2804 If @code{TARGET_SECONDARY_MEMORY_NEEDED} tells the compiler to use memory
2805 when moving between two particular registers of mode @var{mode},
2806 this hook specifies the mode that the memory should have.
2808 The default depends on @code{TARGET_LRA_P}. Without LRA, the default
2809 is to use a word-sized mode for integral modes that are smaller than a
2810 a word. This is right thing to do on most machines because it ensures
2811 that all bits of the register are copied and prevents accesses to the
2812 registers in a narrower mode, which some machines prohibit for
2813 floating-point registers.
2815 However, this default behavior is not correct on some machines, such as
2816 the DEC Alpha, that store short integers in floating-point registers
2817 differently than in integer registers. On those machines, the default
2818 widening will not work correctly and you must define this hook to
2819 suppress that widening in some cases. See the file @file{alpha.c} for
2822 With LRA, the default is to use @var{mode} unmodified.
2825 @deftypefn {Target Hook} void TARGET_SELECT_EARLY_REMAT_MODES (sbitmap @var{modes})
2826 On some targets, certain modes cannot be held in registers around a
2827 standard ABI call and are relatively expensive to spill to the stack.
2828 The early rematerialization pass can help in such cases by aggressively
2829 recomputing values after calls, so that they don't need to be spilled.
2831 This hook returns the set of such modes by setting the associated bits
2832 in @var{modes}. The default implementation selects no modes, which has
2833 the effect of disabling the early rematerialization pass.
2836 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2837 A target hook which returns @code{true} if pseudos that have been assigned
2838 to registers of class @var{rclass} would likely be spilled because
2839 registers of @var{rclass} are needed for spill registers.
2841 The default version of this target hook returns @code{true} if @var{rclass}
2842 has exactly one register and @code{false} otherwise. On most machines, this
2843 default should be used. For generally register-starved machines, such as
2844 i386, or machines with right register constraints, such as SH, this hook
2845 can be used to avoid excessive spilling.
2847 This hook is also used by some of the global intra-procedural code
2848 transformations to throtle code motion, to avoid increasing register
2852 @deftypefn {Target Hook} {unsigned char} TARGET_CLASS_MAX_NREGS (reg_class_t @var{rclass}, machine_mode @var{mode})
2853 A target hook returns the maximum number of consecutive registers
2854 of class @var{rclass} needed to hold a value of mode @var{mode}.
2856 This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}.
2857 In fact, the value returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass},
2858 @var{mode})} target hook should be the maximum value of
2859 @code{TARGET_HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno}
2860 values in the class @var{rclass}.
2862 This target hook helps control the handling of multiple-word values
2865 The default version of this target hook returns the size of @var{mode}
2869 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2870 A C expression for the maximum number of consecutive registers
2871 of class @var{class} needed to hold a value of mode @var{mode}.
2873 This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}. In fact,
2874 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2875 should be the maximum value of @code{TARGET_HARD_REGNO_NREGS (@var{regno},
2876 @var{mode})} for all @var{regno} values in the class @var{class}.
2878 This macro helps control the handling of multiple-word values
2882 @deftypefn {Target Hook} bool TARGET_CAN_CHANGE_MODE_CLASS (machine_mode @var{from}, machine_mode @var{to}, reg_class_t @var{rclass})
2883 This hook returns true if it is possible to bitcast values held in
2884 registers of class @var{rclass} from mode @var{from} to mode @var{to}
2885 and if doing so preserves the low-order bits that are common to both modes.
2886 The result is only meaningful if @var{rclass} has registers that can hold
2887 both @code{from} and @code{to}. The default implementation returns true.
2889 As an example of when such bitcasting is invalid, loading 32-bit integer or
2890 floating-point objects into floating-point registers on Alpha extends them
2891 to 64 bits. Therefore loading a 64-bit object and then storing it as a
2892 32-bit object does not store the low-order 32 bits, as would be the case
2893 for a normal register. Therefore, @file{alpha.h} defines
2894 @code{TARGET_CAN_CHANGE_MODE_CLASS} to return:
2897 (GET_MODE_SIZE (from) == GET_MODE_SIZE (to)
2898 || !reg_classes_intersect_p (FLOAT_REGS, rclass))
2901 Even if storing from a register in mode @var{to} would be valid,
2902 if both @var{from} and @code{raw_reg_mode} for @var{rclass} are wider
2903 than @code{word_mode}, then we must prevent @var{to} narrowing the
2904 mode. This happens when the middle-end assumes that it can load
2905 or store pieces of an @var{N}-word pseudo, and that the pseudo will
2906 eventually be allocated to @var{N} @code{word_mode} hard registers.
2907 Failure to prevent this kind of mode change will result in the
2908 entire @code{raw_reg_mode} being modified instead of the partial
2909 value that the middle-end intended.
2912 @deftypefn {Target Hook} reg_class_t TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS (int, @var{reg_class_t}, @var{reg_class_t})
2913 A target hook which can change allocno class for given pseudo from
2914 allocno and best class calculated by IRA.
2916 The default version of this target hook always returns given class.
2919 @deftypefn {Target Hook} bool TARGET_LRA_P (void)
2920 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.
2923 @deftypefn {Target Hook} int TARGET_REGISTER_PRIORITY (int)
2924 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.
2927 @deftypefn {Target Hook} bool TARGET_REGISTER_USAGE_LEVELING_P (void)
2928 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.
2931 @deftypefn {Target Hook} bool TARGET_DIFFERENT_ADDR_DISPLACEMENT_P (void)
2932 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.
2935 @deftypefn {Target Hook} bool TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P (rtx @var{subst})
2936 A target hook which returns @code{true} if @var{subst} can't
2937 substitute safely pseudos with equivalent memory values during
2938 register allocation.
2939 The default version of this target hook returns @code{false}.
2940 On most machines, this default should be used. For generally
2941 machines with non orthogonal register usage for addressing, such
2942 as SH, this hook can be used to avoid excessive spilling.
2945 @deftypefn {Target Hook} bool TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT (rtx *@var{offset1}, rtx *@var{offset2}, poly_int64 @var{orig_offset}, machine_mode @var{mode})
2946 This hook tries to split address offset @var{orig_offset} into
2947 two parts: one that should be added to the base address to create
2948 a local anchor point, and an additional offset that can be applied
2949 to the anchor to address a value of mode @var{mode}. The idea is that
2950 the local anchor could be shared by other accesses to nearby locations.
2952 The hook returns true if it succeeds, storing the offset of the
2953 anchor from the base in @var{offset1} and the offset of the final address
2954 from the anchor in @var{offset2}. The default implementation returns false.
2957 @deftypefn {Target Hook} reg_class_t TARGET_SPILL_CLASS (reg_class_t, @var{machine_mode})
2958 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.
2961 @deftypefn {Target Hook} bool TARGET_ADDITIONAL_ALLOCNO_CLASS_P (reg_class_t)
2962 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.
2965 @deftypefn {Target Hook} scalar_int_mode TARGET_CSTORE_MODE (enum insn_code @var{icode})
2966 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.
2969 @deftypefn {Target Hook} int TARGET_COMPUTE_PRESSURE_CLASSES (enum reg_class *@var{pressure_classes})
2970 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}.
2973 @node Stack and Calling
2974 @section Stack Layout and Calling Conventions
2975 @cindex calling conventions
2977 @c prevent bad page break with this line
2978 This describes the stack layout and calling conventions.
2982 * Exception Handling::
2987 * Register Arguments::
2989 * Aggregate Return::
2994 * Shrink-wrapping separate components::
2995 * Stack Smashing Protection::
2996 * Miscellaneous Register Hooks::
3000 @subsection Basic Stack Layout
3001 @cindex stack frame layout
3002 @cindex frame layout
3004 @c prevent bad page break with this line
3005 Here is the basic stack layout.
3007 @defmac STACK_GROWS_DOWNWARD
3008 Define this macro to be true if pushing a word onto the stack moves the stack
3009 pointer to a smaller address, and false otherwise.
3012 @defmac STACK_PUSH_CODE
3013 This macro defines the operation used when something is pushed
3014 on the stack. In RTL, a push operation will be
3015 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3017 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3018 and @code{POST_INC}. Which of these is correct depends on
3019 the stack direction and on whether the stack pointer points
3020 to the last item on the stack or whether it points to the
3021 space for the next item on the stack.
3023 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3024 true, which is almost always right, and @code{PRE_INC} otherwise,
3025 which is often wrong.
3028 @defmac FRAME_GROWS_DOWNWARD
3029 Define this macro to nonzero value if the addresses of local variable slots
3030 are at negative offsets from the frame pointer.
3033 @defmac ARGS_GROW_DOWNWARD
3034 Define this macro if successive arguments to a function occupy decreasing
3035 addresses on the stack.
3038 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_STARTING_FRAME_OFFSET (void)
3039 This hook returns the offset from the frame pointer to the first local
3040 variable slot to be allocated. If @code{FRAME_GROWS_DOWNWARD}, it is the
3041 offset to @emph{end} of the first slot allocated, otherwise it is the
3042 offset to @emph{beginning} of the first slot allocated. The default
3043 implementation returns 0.
3046 @defmac STACK_ALIGNMENT_NEEDED
3047 Define to zero to disable final alignment of the stack during reload.
3048 The nonzero default for this macro is suitable for most ports.
3050 On ports where @code{TARGET_STARTING_FRAME_OFFSET} is nonzero or where there
3051 is a register save block following the local block that doesn't require
3052 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3053 stack alignment and do it in the backend.
3056 @defmac STACK_POINTER_OFFSET
3057 Offset from the stack pointer register to the first location at which
3058 outgoing arguments are placed. If not specified, the default value of
3059 zero is used. This is the proper value for most machines.
3061 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3062 the first location at which outgoing arguments are placed.
3065 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3066 Offset from the argument pointer register to the first argument's
3067 address. On some machines it may depend on the data type of the
3070 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3071 the first argument's address.
3074 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3075 Offset from the stack pointer register to an item dynamically allocated
3076 on the stack, e.g., by @code{alloca}.
3078 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3079 length of the outgoing arguments. The default is correct for most
3080 machines. See @file{function.c} for details.
3083 @defmac INITIAL_FRAME_ADDRESS_RTX
3084 A C expression whose value is RTL representing the address of the initial
3085 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3086 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3087 default value will be used. Define this macro in order to make frame pointer
3088 elimination work in the presence of @code{__builtin_frame_address (count)} and
3089 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3092 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3093 A C expression whose value is RTL representing the address in a stack
3094 frame where the pointer to the caller's frame is stored. Assume that
3095 @var{frameaddr} is an RTL expression for the address of the stack frame
3098 If you don't define this macro, the default is to return the value
3099 of @var{frameaddr}---that is, the stack frame address is also the
3100 address of the stack word that points to the previous frame.
3103 @defmac SETUP_FRAME_ADDRESSES
3104 A C expression that produces the machine-specific code to
3105 setup the stack so that arbitrary frames can be accessed. For example,
3106 on the SPARC, we must flush all of the register windows to the stack
3107 before we can access arbitrary stack frames. You will seldom need to
3108 define this macro. The default is to do nothing.
3111 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3112 This target hook should return an rtx that is used to store
3113 the address of the current frame into the built in @code{setjmp} buffer.
3114 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3115 machines. One reason you may need to define this target hook is if
3116 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3119 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3120 A C expression whose value is RTL representing the value of the frame
3121 address for the current frame. @var{frameaddr} is the frame pointer
3122 of the current frame. This is used for __builtin_frame_address.
3123 You need only define this macro if the frame address is not the same
3124 as the frame pointer. Most machines do not need to define it.
3127 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3128 A C expression whose value is RTL representing the value of the return
3129 address for the frame @var{count} steps up from the current frame, after
3130 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3131 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3132 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero.
3134 The value of the expression must always be the correct address when
3135 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3136 determine the return address of other frames.
3139 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3140 Define this macro to nonzero value if the return address of a particular
3141 stack frame is accessed from the frame pointer of the previous stack
3142 frame. The zero default for this macro is suitable for most ports.
3145 @defmac INCOMING_RETURN_ADDR_RTX
3146 A C expression whose value is RTL representing the location of the
3147 incoming return address at the beginning of any function, before the
3148 prologue. This RTL is either a @code{REG}, indicating that the return
3149 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3152 You only need to define this macro if you want to support call frame
3153 debugging information like that provided by DWARF 2.
3155 If this RTL is a @code{REG}, you should also define
3156 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3159 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3160 A C expression whose value is an integer giving a DWARF 2 column
3161 number that may be used as an alternative return column. The column
3162 must not correspond to any gcc hard register (that is, it must not
3163 be in the range of @code{DWARF_FRAME_REGNUM}).
3165 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3166 general register, but an alternative column needs to be used for signal
3167 frames. Some targets have also used different frame return columns
3171 @defmac DWARF_ZERO_REG
3172 A C expression whose value is an integer giving a DWARF 2 register
3173 number that is considered to always have the value zero. This should
3174 only be defined if the target has an architected zero register, and
3175 someone decided it was a good idea to use that register number to
3176 terminate the stack backtrace. New ports should avoid this.
3179 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3180 This target hook allows the backend to emit frame-related insns that
3181 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3182 info engine will invoke it on insns of the form
3184 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3188 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3190 to let the backend emit the call frame instructions. @var{label} is
3191 the CFI label attached to the insn, @var{pattern} is the pattern of
3192 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3195 @deftypefn {Target Hook} {unsigned int} TARGET_DWARF_POLY_INDETERMINATE_VALUE (unsigned int @var{i}, unsigned int *@var{factor}, int *@var{offset})
3196 Express the value of @code{poly_int} indeterminate @var{i} as a DWARF
3197 expression, with @var{i} counting from 1. Return the number of a DWARF
3198 register @var{R} and set @samp{*@var{factor}} and @samp{*@var{offset}} such
3199 that the value of the indeterminate is:
3201 value_of(@var{R}) / @var{factor} - @var{offset}
3204 A target only needs to define this hook if it sets
3205 @samp{NUM_POLY_INT_COEFFS} to a value greater than 1.
3208 @defmac INCOMING_FRAME_SP_OFFSET
3209 A C expression whose value is an integer giving the offset, in bytes,
3210 from the value of the stack pointer register to the top of the stack
3211 frame at the beginning of any function, before the prologue. The top of
3212 the frame is defined to be the value of the stack pointer in the
3213 previous frame, just before the call instruction.
3215 You only need to define this macro if you want to support call frame
3216 debugging information like that provided by DWARF 2.
3219 @defmac DEFAULT_INCOMING_FRAME_SP_OFFSET
3220 Like @code{INCOMING_FRAME_SP_OFFSET}, but must be the same for all
3221 functions of the same ABI, and when using GAS @code{.cfi_*} directives
3222 must also agree with the default CFI GAS emits. Define this macro
3223 only if @code{INCOMING_FRAME_SP_OFFSET} can have different values
3224 between different functions of the same ABI or when
3225 @code{INCOMING_FRAME_SP_OFFSET} does not agree with GAS default CFI.
3228 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3229 A C expression whose value is an integer giving the offset, in bytes,
3230 from the argument pointer to the canonical frame address (cfa). The
3231 final value should coincide with that calculated by
3232 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3233 during virtual register instantiation.
3235 The default value for this macro is
3236 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3237 which is correct for most machines; in general, the arguments are found
3238 immediately before the stack frame. Note that this is not the case on
3239 some targets that save registers into the caller's frame, such as SPARC
3240 and rs6000, and so such targets need to define this macro.
3242 You only need to define this macro if the default is incorrect, and you
3243 want to support call frame debugging information like that provided by
3247 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3248 If defined, a C expression whose value is an integer giving the offset
3249 in bytes from the frame pointer to the canonical frame address (cfa).
3250 The final value should coincide with that calculated by
3251 @code{INCOMING_FRAME_SP_OFFSET}.
3253 Normally the CFA is calculated as an offset from the argument pointer,
3254 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3255 variable due to the ABI, this may not be possible. If this macro is
3256 defined, it implies that the virtual register instantiation should be
3257 based on the frame pointer instead of the argument pointer. Only one
3258 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3262 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3263 If defined, a C expression whose value is an integer giving the offset
3264 in bytes from the canonical frame address (cfa) to the frame base used
3265 in DWARF 2 debug information. The default is zero. A different value
3266 may reduce the size of debug information on some ports.
3269 @node Exception Handling
3270 @subsection Exception Handling Support
3271 @cindex exception handling
3273 @defmac EH_RETURN_DATA_REGNO (@var{N})
3274 A C expression whose value is the @var{N}th register number used for
3275 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3276 @var{N} registers are usable.
3278 The exception handling library routines communicate with the exception
3279 handlers via a set of agreed upon registers. Ideally these registers
3280 should be call-clobbered; it is possible to use call-saved registers,
3281 but may negatively impact code size. The target must support at least
3282 2 data registers, but should define 4 if there are enough free registers.
3284 You must define this macro if you want to support call frame exception
3285 handling like that provided by DWARF 2.
3288 @defmac EH_RETURN_STACKADJ_RTX
3289 A C expression whose value is RTL representing a location in which
3290 to store a stack adjustment to be applied before function return.
3291 This is used to unwind the stack to an exception handler's call frame.
3292 It will be assigned zero on code paths that return normally.
3294 Typically this is a call-clobbered hard register that is otherwise
3295 untouched by the epilogue, but could also be a stack slot.
3297 Do not define this macro if the stack pointer is saved and restored
3298 by the regular prolog and epilog code in the call frame itself; in
3299 this case, the exception handling library routines will update the
3300 stack location to be restored in place. Otherwise, you must define
3301 this macro if you want to support call frame exception handling like
3302 that provided by DWARF 2.
3305 @defmac EH_RETURN_HANDLER_RTX
3306 A C expression whose value is RTL representing a location in which
3307 to store the address of an exception handler to which we should
3308 return. It will not be assigned on code paths that return normally.
3310 Typically this is the location in the call frame at which the normal
3311 return address is stored. For targets that return by popping an
3312 address off the stack, this might be a memory address just below
3313 the @emph{target} call frame rather than inside the current call
3314 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3315 been assigned, so it may be used to calculate the location of the
3318 Some targets have more complex requirements than storing to an
3319 address calculable during initial code generation. In that case
3320 the @code{eh_return} instruction pattern should be used instead.
3322 If you want to support call frame exception handling, you must
3323 define either this macro or the @code{eh_return} instruction pattern.
3326 @defmac RETURN_ADDR_OFFSET
3327 If defined, an integer-valued C expression for which rtl will be generated
3328 to add it to the exception handler address before it is searched in the
3329 exception handling tables, and to subtract it again from the address before
3330 using it to return to the exception handler.
3333 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3334 This macro chooses the encoding of pointers embedded in the exception
3335 handling sections. If at all possible, this should be defined such
3336 that the exception handling section will not require dynamic relocations,
3337 and so may be read-only.
3339 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3340 @var{global} is true if the symbol may be affected by dynamic relocations.
3341 The macro should return a combination of the @code{DW_EH_PE_*} defines
3342 as found in @file{dwarf2.h}.
3344 If this macro is not defined, pointers will not be encoded but
3345 represented directly.
3348 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3349 This macro allows the target to emit whatever special magic is required
3350 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3351 Generic code takes care of pc-relative and indirect encodings; this must
3352 be defined if the target uses text-relative or data-relative encodings.
3354 This is a C statement that branches to @var{done} if the format was
3355 handled. @var{encoding} is the format chosen, @var{size} is the number
3356 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3360 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3361 This macro allows the target to add CPU and operating system specific
3362 code to the call-frame unwinder for use when there is no unwind data
3363 available. The most common reason to implement this macro is to unwind
3364 through signal frames.
3366 This macro is called from @code{uw_frame_state_for} in
3367 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3368 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3369 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3370 for the address of the code being executed and @code{context->cfa} for
3371 the stack pointer value. If the frame can be decoded, the register
3372 save addresses should be updated in @var{fs} and the macro should
3373 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3374 the macro should evaluate to @code{_URC_END_OF_STACK}.
3376 For proper signal handling in Java this macro is accompanied by
3377 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3380 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3381 This macro allows the target to add operating system specific code to the
3382 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3383 usually used for signal or interrupt frames.
3385 This macro is called from @code{uw_update_context} in libgcc's
3386 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3387 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3388 for the abi and context in the @code{.unwabi} directive. If the
3389 @code{.unwabi} directive can be handled, the register save addresses should
3390 be updated in @var{fs}.
3393 @defmac TARGET_USES_WEAK_UNWIND_INFO
3394 A C expression that evaluates to true if the target requires unwind
3395 info to be given comdat linkage. Define it to be @code{1} if comdat
3396 linkage is necessary. The default is @code{0}.
3399 @node Stack Checking
3400 @subsection Specifying How Stack Checking is Done
3402 GCC will check that stack references are within the boundaries of the
3403 stack, if the option @option{-fstack-check} is specified, in one of
3408 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3409 will assume that you have arranged for full stack checking to be done
3410 at appropriate places in the configuration files. GCC will not do
3411 other special processing.
3414 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3415 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3416 that you have arranged for static stack checking (checking of the
3417 static stack frame of functions) to be done at appropriate places
3418 in the configuration files. GCC will only emit code to do dynamic
3419 stack checking (checking on dynamic stack allocations) using the third
3423 If neither of the above are true, GCC will generate code to periodically
3424 ``probe'' the stack pointer using the values of the macros defined below.
3427 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3428 GCC will change its allocation strategy for large objects if the option
3429 @option{-fstack-check} is specified: they will always be allocated
3430 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3432 @defmac STACK_CHECK_BUILTIN
3433 A nonzero value if stack checking is done by the configuration files in a
3434 machine-dependent manner. You should define this macro if stack checking
3435 is required by the ABI of your machine or if you would like to do stack
3436 checking in some more efficient way than the generic approach. The default
3437 value of this macro is zero.
3440 @defmac STACK_CHECK_STATIC_BUILTIN
3441 A nonzero value if static stack checking is done by the configuration files
3442 in a machine-dependent manner. You should define this macro if you would
3443 like to do static stack checking in some more efficient way than the generic
3444 approach. The default value of this macro is zero.
3447 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3448 An integer specifying the interval at which GCC must generate stack probe
3449 instructions, defined as 2 raised to this integer. You will normally
3450 define this macro so that the interval be no larger than the size of
3451 the ``guard pages'' at the end of a stack area. The default value
3452 of 12 (4096-byte interval) is suitable for most systems.
3455 @defmac STACK_CHECK_MOVING_SP
3456 An integer which is nonzero if GCC should move the stack pointer page by page
3457 when doing probes. This can be necessary on systems where the stack pointer
3458 contains the bottom address of the memory area accessible to the executing
3459 thread at any point in time. In this situation an alternate signal stack
3460 is required in order to be able to recover from a stack overflow. The
3461 default value of this macro is zero.
3464 @defmac STACK_CHECK_PROTECT
3465 The number of bytes of stack needed to recover from a stack overflow, for
3466 languages where such a recovery is supported. The default value of 4KB/8KB
3467 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3468 8KB/12KB with other exception handling mechanisms should be adequate for most
3469 architectures and operating systems.
3472 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3473 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3474 in the opposite case.
3476 @defmac STACK_CHECK_MAX_FRAME_SIZE
3477 The maximum size of a stack frame, in bytes. GCC will generate probe
3478 instructions in non-leaf functions to ensure at least this many bytes of
3479 stack are available. If a stack frame is larger than this size, stack
3480 checking will not be reliable and GCC will issue a warning. The
3481 default is chosen so that GCC only generates one instruction on most
3482 systems. You should normally not change the default value of this macro.
3485 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3486 GCC uses this value to generate the above warning message. It
3487 represents the amount of fixed frame used by a function, not including
3488 space for any callee-saved registers, temporaries and user variables.
3489 You need only specify an upper bound for this amount and will normally
3490 use the default of four words.
3493 @defmac STACK_CHECK_MAX_VAR_SIZE
3494 The maximum size, in bytes, of an object that GCC will place in the
3495 fixed area of the stack frame when the user specifies
3496 @option{-fstack-check}.
3497 GCC computed the default from the values of the above macros and you will
3498 normally not need to override that default.
3501 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE (void)
3502 Some targets have an ABI defined interval for which no probing needs to be done.
3503 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.
3504 On such targets this value can be set to override the default probing up interval.
3505 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.
3506 You need not define this macro if it would always have the value zero.
3510 @node Frame Registers
3511 @subsection Registers That Address the Stack Frame
3513 @c prevent bad page break with this line
3514 This discusses registers that address the stack frame.
3516 @defmac STACK_POINTER_REGNUM
3517 The register number of the stack pointer register, which must also be a
3518 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3519 the hardware determines which register this is.
3522 @defmac FRAME_POINTER_REGNUM
3523 The register number of the frame pointer register, which is used to
3524 access automatic variables in the stack frame. On some machines, the
3525 hardware determines which register this is. On other machines, you can
3526 choose any register you wish for this purpose.
3529 @defmac HARD_FRAME_POINTER_REGNUM
3530 On some machines the offset between the frame pointer and starting
3531 offset of the automatic variables is not known until after register
3532 allocation has been done (for example, because the saved registers are
3533 between these two locations). On those machines, define
3534 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3535 be used internally until the offset is known, and define
3536 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3537 used for the frame pointer.
3539 You should define this macro only in the very rare circumstances when it
3540 is not possible to calculate the offset between the frame pointer and
3541 the automatic variables until after register allocation has been
3542 completed. When this macro is defined, you must also indicate in your
3543 definition of @code{ELIMINABLE_REGS} how to eliminate
3544 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3545 or @code{STACK_POINTER_REGNUM}.
3547 Do not define this macro if it would be the same as
3548 @code{FRAME_POINTER_REGNUM}.
3551 @defmac ARG_POINTER_REGNUM
3552 The register number of the arg pointer register, which is used to access
3553 the function's argument list. On some machines, this is the same as the
3554 frame pointer register. On some machines, the hardware determines which
3555 register this is. On other machines, you can choose any register you
3556 wish for this purpose. If this is not the same register as the frame
3557 pointer register, then you must mark it as a fixed register according to
3558 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3559 (@pxref{Elimination}).
3562 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3563 Define this to a preprocessor constant that is nonzero if
3564 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3565 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3566 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3567 definition is not suitable for use in preprocessor conditionals.
3570 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3571 Define this to a preprocessor constant that is nonzero if
3572 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3573 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3574 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3575 definition is not suitable for use in preprocessor conditionals.
3578 @defmac RETURN_ADDRESS_POINTER_REGNUM
3579 The register number of the return address pointer register, which is used to
3580 access the current function's return address from the stack. On some
3581 machines, the return address is not at a fixed offset from the frame
3582 pointer or stack pointer or argument pointer. This register can be defined
3583 to point to the return address on the stack, and then be converted by
3584 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3586 Do not define this macro unless there is no other way to get the return
3587 address from the stack.
3590 @defmac STATIC_CHAIN_REGNUM
3591 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3592 Register numbers used for passing a function's static chain pointer. If
3593 register windows are used, the register number as seen by the called
3594 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3595 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3596 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3599 The static chain register need not be a fixed register.
3601 If the static chain is passed in memory, these macros should not be
3602 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3605 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl_or_type}, bool @var{incoming_p})
3606 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3607 targets that may use different static chain locations for different
3608 nested functions. This may be required if the target has function
3609 attributes that affect the calling conventions of the function and
3610 those calling conventions use different static chain locations.
3612 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3614 If the static chain is passed in memory, this hook should be used to
3615 provide rtx giving @code{mem} expressions that denote where they are stored.
3616 Often the @code{mem} expression as seen by the caller will be at an offset
3617 from the stack pointer and the @code{mem} expression as seen by the callee
3618 will be at an offset from the frame pointer.
3619 @findex stack_pointer_rtx
3620 @findex frame_pointer_rtx
3621 @findex arg_pointer_rtx
3622 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3623 @code{arg_pointer_rtx} will have been initialized and should be used
3624 to refer to those items.
3627 @defmac DWARF_FRAME_REGISTERS
3628 This macro specifies the maximum number of hard registers that can be
3629 saved in a call frame. This is used to size data structures used in
3630 DWARF2 exception handling.
3632 Prior to GCC 3.0, this macro was needed in order to establish a stable
3633 exception handling ABI in the face of adding new hard registers for ISA
3634 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3635 in the number of hard registers. Nevertheless, this macro can still be
3636 used to reduce the runtime memory requirements of the exception handling
3637 routines, which can be substantial if the ISA contains a lot of
3638 registers that are not call-saved.
3640 If this macro is not defined, it defaults to
3641 @code{FIRST_PSEUDO_REGISTER}.
3644 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3646 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3647 for backward compatibility in pre GCC 3.0 compiled code.
3649 If this macro is not defined, it defaults to
3650 @code{DWARF_FRAME_REGISTERS}.
3653 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3655 Define this macro if the target's representation for dwarf registers
3656 is different than the internal representation for unwind column.
3657 Given a dwarf register, this macro should return the internal unwind
3658 column number to use instead.
3661 @defmac DWARF_FRAME_REGNUM (@var{regno})
3663 Define this macro if the target's representation for dwarf registers
3664 used in .eh_frame or .debug_frame is different from that used in other
3665 debug info sections. Given a GCC hard register number, this macro
3666 should return the .eh_frame register number. The default is
3667 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3671 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3673 Define this macro to map register numbers held in the call frame info
3674 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3675 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3676 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3677 return @code{@var{regno}}.
3681 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3683 Define this macro if the target stores register values as
3684 @code{_Unwind_Word} type in unwind context. It should be defined if
3685 target register size is larger than the size of @code{void *}. The
3686 default is to store register values as @code{void *} type.
3690 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3692 Define this macro to be 1 if the target always uses extended unwind
3693 context with version, args_size and by_value fields. If it is undefined,
3694 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3695 defined and 0 otherwise.
3699 @defmac DWARF_LAZY_REGISTER_VALUE (@var{regno}, @var{value})
3700 Define this macro if the target has pseudo DWARF registers whose
3701 values need to be computed lazily on demand by the unwinder (such as when
3702 referenced in a CFA expression). The macro returns true if @var{regno}
3703 is such a register and stores its value in @samp{*@var{value}} if so.
3707 @subsection Eliminating Frame Pointer and Arg Pointer
3709 @c prevent bad page break with this line
3710 This is about eliminating the frame pointer and arg pointer.
3712 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3713 This target hook should return @code{true} if a function must have and use
3714 a frame pointer. This target hook is called in the reload pass. If its return
3715 value is @code{true} the function will have a frame pointer.
3717 This target hook can in principle examine the current function and decide
3718 according to the facts, but on most machines the constant @code{false} or the
3719 constant @code{true} suffices. Use @code{false} when the machine allows code
3720 to be generated with no frame pointer, and doing so saves some time or space.
3721 Use @code{true} when there is no possible advantage to avoiding a frame
3724 In certain cases, the compiler does not know how to produce valid code
3725 without a frame pointer. The compiler recognizes those cases and
3726 automatically gives the function a frame pointer regardless of what
3727 @code{targetm.frame_pointer_required} returns. You don't need to worry about
3730 In a function that does not require a frame pointer, the frame pointer
3731 register can be allocated for ordinary usage, unless you mark it as a
3732 fixed register. See @code{FIXED_REGISTERS} for more information.
3734 Default return value is @code{false}.
3737 @defmac ELIMINABLE_REGS
3738 This macro specifies a table of register pairs used to eliminate
3739 unneeded registers that point into the stack frame.
3741 The definition of this macro is a list of structure initializations, each
3742 of which specifies an original and replacement register.
3744 On some machines, the position of the argument pointer is not known until
3745 the compilation is completed. In such a case, a separate hard register
3746 must be used for the argument pointer. This register can be eliminated by
3747 replacing it with either the frame pointer or the argument pointer,
3748 depending on whether or not the frame pointer has been eliminated.
3750 In this case, you might specify:
3752 #define ELIMINABLE_REGS \
3753 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3754 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3755 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3758 Note that the elimination of the argument pointer with the stack pointer is
3759 specified first since that is the preferred elimination.
3762 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3763 This target hook should return @code{true} if the compiler is allowed to
3764 try to replace register number @var{from_reg} with register number
3765 @var{to_reg}. This target hook will usually be @code{true}, since most of the
3766 cases preventing register elimination are things that the compiler already
3769 Default return value is @code{true}.
3772 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3773 This macro returns the initial difference between the specified pair
3774 of registers. The value would be computed from information
3775 such as the result of @code{get_frame_size ()} and the tables of
3776 registers @code{df_regs_ever_live_p} and @code{call_used_regs}.
3779 @deftypefn {Target Hook} void TARGET_COMPUTE_FRAME_LAYOUT (void)
3780 This target hook is called once each time the frame layout needs to be
3781 recalculated. The calculations can be cached by the target and can then
3782 be used by @code{INITIAL_ELIMINATION_OFFSET} instead of re-computing the
3783 layout on every invocation of that hook. This is particularly useful
3784 for targets that have an expensive frame layout function. Implementing
3785 this callback is optional.
3788 @node Stack Arguments
3789 @subsection Passing Function Arguments on the Stack
3790 @cindex arguments on stack
3791 @cindex stack arguments
3793 The macros in this section control how arguments are passed
3794 on the stack. See the following section for other macros that
3795 control passing certain arguments in registers.
3797 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3798 This target hook returns @code{true} if an argument declared in a
3799 prototype as an integral type smaller than @code{int} should actually be
3800 passed as an @code{int}. In addition to avoiding errors in certain
3801 cases of mismatch, it also makes for better code on certain machines.
3802 The default is to not promote prototypes.
3806 A C expression. If nonzero, push insns will be used to pass
3808 If the target machine does not have a push instruction, set it to zero.
3809 That directs GCC to use an alternate strategy: to
3810 allocate the entire argument block and then store the arguments into
3811 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3814 @defmac PUSH_ARGS_REVERSED
3815 A C expression. If nonzero, function arguments will be evaluated from
3816 last to first, rather than from first to last. If this macro is not
3817 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3818 and args grow in opposite directions, and 0 otherwise.
3821 @defmac PUSH_ROUNDING (@var{npushed})
3822 A C expression that is the number of bytes actually pushed onto the
3823 stack when an instruction attempts to push @var{npushed} bytes.
3825 On some machines, the definition
3828 #define PUSH_ROUNDING(BYTES) (BYTES)
3832 will suffice. But on other machines, instructions that appear
3833 to push one byte actually push two bytes in an attempt to maintain
3834 alignment. Then the definition should be
3837 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3840 If the value of this macro has a type, it should be an unsigned type.
3843 @findex outgoing_args_size
3844 @findex crtl->outgoing_args_size
3845 @defmac ACCUMULATE_OUTGOING_ARGS
3846 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3847 will be computed and placed into
3848 @code{crtl->outgoing_args_size}. No space will be pushed
3849 onto the stack for each call; instead, the function prologue should
3850 increase the stack frame size by this amount.
3852 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3856 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3857 Define this macro if functions should assume that stack space has been
3858 allocated for arguments even when their values are passed in
3861 The value of this macro is the size, in bytes, of the area reserved for
3862 arguments passed in registers for the function represented by @var{fndecl},
3863 which can be zero if GCC is calling a library function.
3864 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3867 This space can be allocated by the caller, or be a part of the
3868 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3871 @c above is overfull. not sure what to do. --mew 5feb93 did
3872 @c something, not sure if it looks good. --mew 10feb93
3874 @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3875 Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3876 Define this macro if space guaranteed when compiling a function body
3877 is different to space required when making a call, a situation that
3878 can arise with K&R style function definitions.
3881 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3882 Define this to a nonzero value if it is the responsibility of the
3883 caller to allocate the area reserved for arguments passed in registers
3884 when calling a function of @var{fntype}. @var{fntype} may be NULL
3885 if the function called is a library function.
3887 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3888 whether the space for these arguments counts in the value of
3889 @code{crtl->outgoing_args_size}.
3892 @defmac STACK_PARMS_IN_REG_PARM_AREA
3893 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3894 stack parameters don't skip the area specified by it.
3895 @c i changed this, makes more sens and it should have taken care of the
3896 @c overfull.. not as specific, tho. --mew 5feb93
3898 Normally, when a parameter is not passed in registers, it is placed on the
3899 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3900 suppresses this behavior and causes the parameter to be passed on the
3901 stack in its natural location.
3904 @deftypefn {Target Hook} poly_int64 TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, poly_int64 @var{size})
3905 This target hook returns the number of bytes of its own arguments that
3906 a function pops on returning, or 0 if the function pops no arguments
3907 and the caller must therefore pop them all after the function returns.
3909 @var{fundecl} is a C variable whose value is a tree node that describes
3910 the function in question. Normally it is a node of type
3911 @code{FUNCTION_DECL} that describes the declaration of the function.
3912 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3914 @var{funtype} is a C variable whose value is a tree node that
3915 describes the function in question. Normally it is a node of type
3916 @code{FUNCTION_TYPE} that describes the data type of the function.
3917 From this it is possible to obtain the data types of the value and
3918 arguments (if known).
3920 When a call to a library function is being considered, @var{fundecl}
3921 will contain an identifier node for the library function. Thus, if
3922 you need to distinguish among various library functions, you can do so
3923 by their names. Note that ``library function'' in this context means
3924 a function used to perform arithmetic, whose name is known specially
3925 in the compiler and was not mentioned in the C code being compiled.
3927 @var{size} is the number of bytes of arguments passed on the
3928 stack. If a variable number of bytes is passed, it is zero, and
3929 argument popping will always be the responsibility of the calling function.
3931 On the VAX, all functions always pop their arguments, so the definition
3932 of this macro is @var{size}. On the 68000, using the standard
3933 calling convention, no functions pop their arguments, so the value of
3934 the macro is always 0 in this case. But an alternative calling
3935 convention is available in which functions that take a fixed number of
3936 arguments pop them but other functions (such as @code{printf}) pop
3937 nothing (the caller pops all). When this convention is in use,
3938 @var{funtype} is examined to determine whether a function takes a fixed
3939 number of arguments.
3942 @defmac CALL_POPS_ARGS (@var{cum})
3943 A C expression that should indicate the number of bytes a call sequence
3944 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3945 when compiling a function call.
3947 @var{cum} is the variable in which all arguments to the called function
3948 have been accumulated.
3950 On certain architectures, such as the SH5, a call trampoline is used
3951 that pops certain registers off the stack, depending on the arguments
3952 that have been passed to the function. Since this is a property of the
3953 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3957 @node Register Arguments
3958 @subsection Passing Arguments in Registers
3959 @cindex arguments in registers
3960 @cindex registers arguments
3962 This section describes the macros which let you control how various
3963 types of arguments are passed in registers or how they are arranged in
3966 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, const function_arg_info @var{&arg})
3967 Return an RTX indicating whether function argument @var{arg} is passed
3968 in a register and if so, which register. Argument @var{ca} summarizes all
3969 the previous arguments.
3971 The return value is usually either a @code{reg} RTX for the hard
3972 register in which to pass the argument, or zero to pass the argument
3975 The return value can be a @code{const_int} which means argument is
3976 passed in a target specific slot with specified number. Target hooks
3977 should be used to store or load argument in such case. See
3978 @code{TARGET_STORE_BOUNDS_FOR_ARG} and @code{TARGET_LOAD_BOUNDS_FOR_ARG}
3979 for more information.
3981 The value of the expression can also be a @code{parallel} RTX@. This is
3982 used when an argument is passed in multiple locations. The mode of the
3983 @code{parallel} should be the mode of the entire argument. The
3984 @code{parallel} holds any number of @code{expr_list} pairs; each one
3985 describes where part of the argument is passed. In each
3986 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3987 register in which to pass this part of the argument, and the mode of the
3988 register RTX indicates how large this part of the argument is. The
3989 second operand of the @code{expr_list} is a @code{const_int} which gives
3990 the offset in bytes into the entire argument of where this part starts.
3991 As a special exception the first @code{expr_list} in the @code{parallel}
3992 RTX may have a first operand of zero. This indicates that the entire
3993 argument is also stored on the stack.
3995 The last time this hook is called, it is called with @code{MODE ==
3996 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3997 pattern as operands 2 and 3 respectively.
3999 @cindex @file{stdarg.h} and register arguments
4000 The usual way to make the ISO library @file{stdarg.h} work on a
4001 machine where some arguments are usually passed in registers, is to
4002 cause nameless arguments to be passed on the stack instead. This is
4003 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
4004 @var{named} is @code{false}.
4006 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
4007 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
4008 You may use the hook @code{targetm.calls.must_pass_in_stack}
4009 in the definition of this macro to determine if this argument is of a
4010 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4011 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
4012 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4013 defined, the argument will be computed in the stack and then loaded into
4017 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (const function_arg_info @var{&arg})
4018 This target hook should return @code{true} if we should not pass @var{arg}
4019 solely in registers. The file @file{expr.h} defines a
4020 definition that is usually appropriate, refer to @file{expr.h} for additional
4024 @deftypefn {Target Hook} rtx TARGET_FUNCTION_INCOMING_ARG (cumulative_args_t @var{ca}, const function_arg_info @var{&arg})
4025 Define this hook if the caller and callee on the target have different
4026 views of where arguments are passed. Also define this hook if there are
4027 functions that are never directly called, but are invoked by the hardware
4028 and which have nonstandard calling conventions.
4030 In this case @code{TARGET_FUNCTION_ARG} computes the register in
4031 which the caller passes the value, and
4032 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
4033 fashion to tell the function being called where the arguments will
4036 @code{TARGET_FUNCTION_INCOMING_ARG} can also return arbitrary address
4037 computation using hard register, which can be forced into a register,
4038 so that it can be used to pass special arguments.
4040 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
4041 @code{TARGET_FUNCTION_ARG} serves both purposes.
4044 @deftypefn {Target Hook} bool TARGET_USE_PSEUDO_PIC_REG (void)
4045 This hook should return 1 in case pseudo register should be created
4046 for pic_offset_table_rtx during function expand.
4049 @deftypefn {Target Hook} void TARGET_INIT_PIC_REG (void)
4050 Perform a target dependent initialization of pic_offset_table_rtx.
4051 This hook is called at the start of register allocation.
4054 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, const function_arg_info @var{&arg})
4055 This target hook returns the number of bytes at the beginning of an
4056 argument that must be put in registers. The value must be zero for
4057 arguments that are passed entirely in registers or that are entirely
4058 pushed on the stack.
4060 On some machines, certain arguments must be passed partially in
4061 registers and partially in memory. On these machines, typically the
4062 first few words of arguments are passed in registers, and the rest
4063 on the stack. If a multi-word argument (a @code{double} or a
4064 structure) crosses that boundary, its first few words must be passed
4065 in registers and the rest must be pushed. This macro tells the
4066 compiler when this occurs, and how many bytes should go in registers.
4068 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
4069 register to be used by the caller for this argument; likewise
4070 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
4073 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (cumulative_args_t @var{cum}, const function_arg_info @var{&arg})
4074 This target hook should return @code{true} if argument @var{arg} at the
4075 position indicated by @var{cum} should be passed by reference. This
4076 predicate is queried after target independent reasons for being
4077 passed by reference, such as @code{TREE_ADDRESSABLE (@var{arg}.type)}.
4079 If the hook returns true, a copy of that argument is made in memory and a
4080 pointer to the argument is passed instead of the argument itself.
4081 The pointer is passed in whatever way is appropriate for passing a pointer
4085 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, const function_arg_info @var{&arg})
4086 The function argument described by the parameters to this hook is
4087 known to be passed by reference. The hook should return true if the
4088 function argument should be copied by the callee instead of copied
4091 For any argument for which the hook returns true, if it can be
4092 determined that the argument is not modified, then a copy need
4095 The default version of this hook always returns false.
4098 @defmac CUMULATIVE_ARGS
4099 A C type for declaring a variable that is used as the first argument
4100 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4101 target machines, the type @code{int} suffices and can hold the number
4102 of bytes of argument so far.
4104 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4105 arguments that have been passed on the stack. The compiler has other
4106 variables to keep track of that. For target machines on which all
4107 arguments are passed on the stack, there is no need to store anything in
4108 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4109 should not be empty, so use @code{int}.
4112 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4113 If defined, this macro is called before generating any code for a
4114 function, but after the @var{cfun} descriptor for the function has been
4115 created. The back end may use this macro to update @var{cfun} to
4116 reflect an ABI other than that which would normally be used by default.
4117 If the compiler is generating code for a compiler-generated function,
4118 @var{fndecl} may be @code{NULL}.
4121 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4122 A C statement (sans semicolon) for initializing the variable
4123 @var{cum} for the state at the beginning of the argument list. The
4124 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4125 is the tree node for the data type of the function which will receive
4126 the args, or 0 if the args are to a compiler support library function.
4127 For direct calls that are not libcalls, @var{fndecl} contain the
4128 declaration node of the function. @var{fndecl} is also set when
4129 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4130 being compiled. @var{n_named_args} is set to the number of named
4131 arguments, including a structure return address if it is passed as a
4132 parameter, when making a call. When processing incoming arguments,
4133 @var{n_named_args} is set to @minus{}1.
4135 When processing a call to a compiler support library function,
4136 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4137 contains the name of the function, as a string. @var{libname} is 0 when
4138 an ordinary C function call is being processed. Thus, each time this
4139 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4140 never both of them at once.
4143 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4144 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4145 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4146 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4147 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4148 0)} is used instead.
4151 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4152 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4153 finding the arguments for the function being compiled. If this macro is
4154 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4156 The value passed for @var{libname} is always 0, since library routines
4157 with special calling conventions are never compiled with GCC@. The
4158 argument @var{libname} exists for symmetry with
4159 @code{INIT_CUMULATIVE_ARGS}.
4160 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4161 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4164 @deftypefn {Target Hook} void TARGET_FUNCTION_ARG_ADVANCE (cumulative_args_t @var{ca}, const function_arg_info @var{&arg})
4165 This hook updates the summarizer variable pointed to by @var{ca} to
4166 advance past argument @var{arg} in the argument list. Once this is done,
4167 the variable @var{cum} is suitable for analyzing the @emph{following}
4168 argument with @code{TARGET_FUNCTION_ARG}, etc.
4170 This hook need not do anything if the argument in question was passed
4171 on the stack. The compiler knows how to track the amount of stack space
4172 used for arguments without any special help.
4175 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_FUNCTION_ARG_OFFSET (machine_mode @var{mode}, const_tree @var{type})
4176 This hook returns the number of bytes to add to the offset of an
4177 argument of type @var{type} and mode @var{mode} when passed in memory.
4178 This is needed for the SPU, which passes @code{char} and @code{short}
4179 arguments in the preferred slot that is in the middle of the quad word
4180 instead of starting at the top. The default implementation returns 0.
4183 @deftypefn {Target Hook} pad_direction TARGET_FUNCTION_ARG_PADDING (machine_mode @var{mode}, const_tree @var{type})
4184 This hook determines whether, and in which direction, to pad out
4185 an argument of mode @var{mode} and type @var{type}. It returns
4186 @code{PAD_UPWARD} to insert padding above the argument, @code{PAD_DOWNWARD}
4187 to insert padding below the argument, or @code{PAD_NONE} to inhibit padding.
4189 The @emph{amount} of padding is not controlled by this hook, but by
4190 @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is always just enough
4191 to reach the next multiple of that boundary.
4193 This hook has a default definition that is right for most systems.
4194 For little-endian machines, the default is to pad upward. For
4195 big-endian machines, the default is to pad downward for an argument of
4196 constant size shorter than an @code{int}, and upward otherwise.
4199 @defmac PAD_VARARGS_DOWN
4200 If defined, a C expression which determines whether the default
4201 implementation of va_arg will attempt to pad down before reading the
4202 next argument, if that argument is smaller than its aligned space as
4203 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4204 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4207 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4208 Specify padding for the last element of a block move between registers and
4209 memory. @var{first} is nonzero if this is the only element. Defining this
4210 macro allows better control of register function parameters on big-endian
4211 machines, without using @code{PARALLEL} rtl. In particular,
4212 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4213 registers, as there is no longer a "wrong" part of a register; For example,
4214 a three byte aggregate may be passed in the high part of a register if so
4218 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (machine_mode @var{mode}, const_tree @var{type})
4219 This hook returns the alignment boundary, in bits, of an argument
4220 with the specified mode and type. The default hook returns
4221 @code{PARM_BOUNDARY} for all arguments.
4224 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_ROUND_BOUNDARY (machine_mode @var{mode}, const_tree @var{type})
4225 Normally, the size of an argument is rounded up to @code{PARM_BOUNDARY},
4226 which is the default value for this hook. You can define this hook to
4227 return a different value if an argument size must be rounded to a larger
4231 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4232 A C expression that is nonzero if @var{regno} is the number of a hard
4233 register in which function arguments are sometimes passed. This does
4234 @emph{not} include implicit arguments such as the static chain and
4235 the structure-value address. On many machines, no registers can be
4236 used for this purpose since all function arguments are pushed on the
4240 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4241 This hook should return true if parameter of type @var{type} are passed
4242 as two scalar parameters. By default, GCC will attempt to pack complex
4243 arguments into the target's word size. Some ABIs require complex arguments
4244 to be split and treated as their individual components. For example, on
4245 AIX64, complex floats should be passed in a pair of floating point
4246 registers, even though a complex float would fit in one 64-bit floating
4249 The default value of this hook is @code{NULL}, which is treated as always
4253 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4254 This hook returns a type node for @code{va_list} for the target.
4255 The default version of the hook returns @code{void*}.
4258 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4259 This target hook is used in function @code{c_common_nodes_and_builtins}
4260 to iterate through the target specific builtin types for va_list. The
4261 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4262 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4264 The arguments @var{pname} and @var{ptree} are used to store the result of
4265 this macro and are set to the name of the va_list builtin type and its
4267 If the return value of this macro is zero, then there is no more element.
4268 Otherwise the @var{IDX} should be increased for the next call of this
4269 macro to iterate through all types.
4272 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4273 This hook returns the va_list type of the calling convention specified by
4275 The default version of this hook returns @code{va_list_type_node}.
4278 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4279 This hook returns the va_list type of the calling convention specified by the
4280 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4284 @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})
4285 This hook performs target-specific gimplification of
4286 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4287 arguments to @code{va_arg}; the latter two are as in
4288 @code{gimplify.c:gimplify_expr}.
4291 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (scalar_int_mode @var{mode})
4292 Define this to return nonzero if the port can handle pointers
4293 with machine mode @var{mode}. The default version of this
4294 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4297 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (ao_ref *@var{ref})
4298 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.
4301 @deftypefn {Target Hook} machine_mode TARGET_TRANSLATE_MODE_ATTRIBUTE (machine_mode @var{mode})
4302 Define this hook if during mode attribute processing, the port should
4303 translate machine_mode @var{mode} to another mode. For example, rs6000's
4304 @code{KFmode}, when it is the same as @code{TFmode}.
4306 The default version of the hook returns that mode that was passed in.
4309 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (scalar_mode @var{mode})
4310 Define this to return nonzero if the port is prepared to handle
4311 insns involving scalar mode @var{mode}. For a scalar mode to be
4312 considered supported, all the basic arithmetic and comparisons
4315 The default version of this hook returns true for any mode
4316 required to handle the basic C types (as defined by the port).
4317 Included here are the double-word arithmetic supported by the
4318 code in @file{optabs.c}.
4321 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (machine_mode @var{mode})
4322 Define this to return nonzero if the port is prepared to handle
4323 insns involving vector mode @var{mode}. At the very least, it
4324 must have move patterns for this mode.
4327 @deftypefn {Target Hook} opt_machine_mode TARGET_ARRAY_MODE (machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4328 Return the mode that GCC should use for an array that has
4329 @var{nelems} elements, with each element having mode @var{mode}.
4330 Return no mode if the target has no special requirements. In the
4331 latter case, GCC looks for an integer mode of the appropriate size
4332 if available and uses BLKmode otherwise. Usually the search for the
4333 integer mode is limited to @code{MAX_FIXED_MODE_SIZE}, but the
4334 @code{TARGET_ARRAY_MODE_SUPPORTED_P} hook allows a larger mode to be
4335 used in specific cases.
4337 The main use of this hook is to specify that an array of vectors should
4338 also have a vector mode. The default implementation returns no mode.
4341 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4342 Return true if GCC should try to use a scalar mode to store an array
4343 of @var{nelems} elements, given that each element has mode @var{mode}.
4344 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4345 and allows GCC to use any defined integer mode.
4347 One use of this hook is to support vector load and store operations
4348 that operate on several homogeneous vectors. For example, ARM NEON
4349 has operations like:
4352 int8x8x3_t vld3_s8 (const int8_t *)
4355 where the return type is defined as:
4358 typedef struct int8x8x3_t
4364 If this hook allows @code{val} to have a scalar mode, then
4365 @code{int8x8x3_t} can have the same mode. GCC can then store
4366 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4369 @deftypefn {Target Hook} bool TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P (scalar_float_mode @var{mode})
4370 Define this to return nonzero if libgcc provides support for the
4371 floating-point mode @var{mode}, which is known to pass
4372 @code{TARGET_SCALAR_MODE_SUPPORTED_P}. The default version of this
4373 hook returns true for all of @code{SFmode}, @code{DFmode},
4374 @code{XFmode} and @code{TFmode}, if such modes exist.
4377 @deftypefn {Target Hook} opt_scalar_float_mode TARGET_FLOATN_MODE (int @var{n}, bool @var{extended})
4378 Define this to return the machine mode to use for the type
4379 @code{_Float@var{n}}, if @var{extended} is false, or the type
4380 @code{_Float@var{n}x}, if @var{extended} is true. If such a type is not
4381 supported, return @code{opt_scalar_float_mode ()}. The default version of
4382 this hook returns @code{SFmode} for @code{_Float32}, @code{DFmode} for
4383 @code{_Float64} and @code{_Float32x} and @code{TFmode} for
4384 @code{_Float128}, if those modes exist and satisfy the requirements for
4385 those types and pass @code{TARGET_SCALAR_MODE_SUPPORTED_P} and
4386 @code{TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P}; for @code{_Float64x}, it
4387 returns the first of @code{XFmode} and @code{TFmode} that exists and
4388 satisfies the same requirements; for other types, it returns
4389 @code{opt_scalar_float_mode ()}. The hook is only called for values
4390 of @var{n} and @var{extended} that are valid according to
4391 ISO/IEC TS 18661-3:2015; that is, @var{n} is one of 32, 64, 128, or,
4392 if @var{extended} is false, 16 or greater than 128 and a multiple of 32.
4395 @deftypefn {Target Hook} bool TARGET_FLOATN_BUILTIN_P (int @var{func})
4396 Define this to return true if the @code{_Float@var{n}} and
4397 @code{_Float@var{n}x} built-in functions should implicitly enable the
4398 built-in function without the @code{__builtin_} prefix in addition to the
4399 normal built-in function with the @code{__builtin_} prefix. The default is
4400 to only enable built-in functions without the @code{__builtin_} prefix for
4401 the GNU C langauge. In strict ANSI/ISO mode, the built-in function without
4402 the @code{__builtin_} prefix is not enabled. The argument @code{FUNC} is the
4403 @code{enum built_in_function} id of the function to be enabled.
4406 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (machine_mode @var{mode})
4407 Define this to return nonzero for machine modes for which the port has
4408 small register classes. If this target hook returns nonzero for a given
4409 @var{mode}, the compiler will try to minimize the lifetime of registers
4410 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4411 In this case, the hook is expected to return nonzero if it returns nonzero
4414 On some machines, it is risky to let hard registers live across arbitrary
4415 insns. Typically, these machines have instructions that require values
4416 to be in specific registers (like an accumulator), and reload will fail
4417 if the required hard register is used for another purpose across such an
4420 Passes before reload do not know which hard registers will be used
4421 in an instruction, but the machine modes of the registers set or used in
4422 the instruction are already known. And for some machines, register
4423 classes are small for, say, integer registers but not for floating point
4424 registers. For example, the AMD x86-64 architecture requires specific
4425 registers for the legacy x86 integer instructions, but there are many
4426 SSE registers for floating point operations. On such targets, a good
4427 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4428 machine modes but zero for the SSE register classes.
4430 The default version of this hook returns false for any mode. It is always
4431 safe to redefine this hook to return with a nonzero value. But if you
4432 unnecessarily define it, you will reduce the amount of optimizations
4433 that can be performed in some cases. If you do not define this hook
4434 to return a nonzero value when it is required, the compiler will run out
4435 of spill registers and print a fatal error message.
4439 @subsection How Scalar Function Values Are Returned
4440 @cindex return values in registers
4441 @cindex values, returned by functions
4442 @cindex scalars, returned as values
4444 This section discusses the macros that control returning scalars as
4445 values---values that can fit in registers.
4447 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4449 Define this to return an RTX representing the place where a function
4450 returns or receives a value of data type @var{ret_type}, a tree node
4451 representing a data type. @var{fn_decl_or_type} is a tree node
4452 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4453 function being called. If @var{outgoing} is false, the hook should
4454 compute the register in which the caller will see the return value.
4455 Otherwise, the hook should return an RTX representing the place where
4456 a function returns a value.
4458 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4459 (Actually, on most machines, scalar values are returned in the same
4460 place regardless of mode.) The value of the expression is usually a
4461 @code{reg} RTX for the hard register where the return value is stored.
4462 The value can also be a @code{parallel} RTX, if the return value is in
4463 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4464 @code{parallel} form. Note that the callee will populate every
4465 location specified in the @code{parallel}, but if the first element of
4466 the @code{parallel} contains the whole return value, callers will use
4467 that element as the canonical location and ignore the others. The m68k
4468 port uses this type of @code{parallel} to return pointers in both
4469 @samp{%a0} (the canonical location) and @samp{%d0}.
4471 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4472 the same promotion rules specified in @code{PROMOTE_MODE} if
4473 @var{valtype} is a scalar type.
4475 If the precise function being called is known, @var{func} is a tree
4476 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4477 pointer. This makes it possible to use a different value-returning
4478 convention for specific functions when all their calls are
4481 Some target machines have ``register windows'' so that the register in
4482 which a function returns its value is not the same as the one in which
4483 the caller sees the value. For such machines, you should return
4484 different RTX depending on @var{outgoing}.
4486 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4487 aggregate data types, because these are returned in another way. See
4488 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4491 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4492 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4493 a new target instead.
4496 @defmac LIBCALL_VALUE (@var{mode})
4497 A C expression to create an RTX representing the place where a library
4498 function returns a value of mode @var{mode}.
4500 Note that ``library function'' in this context means a compiler
4501 support routine, used to perform arithmetic, whose name is known
4502 specially by the compiler and was not mentioned in the C code being
4506 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (machine_mode @var{mode}, const_rtx @var{fun})
4507 Define this hook if the back-end needs to know the name of the libcall
4508 function in order to determine where the result should be returned.
4510 The mode of the result is given by @var{mode} and the name of the called
4511 library function is given by @var{fun}. The hook should return an RTX
4512 representing the place where the library function result will be returned.
4514 If this hook is not defined, then LIBCALL_VALUE will be used.
4517 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4518 A C expression that is nonzero if @var{regno} is the number of a hard
4519 register in which the values of called function may come back.
4521 A register whose use for returning values is limited to serving as the
4522 second of a pair (for a value of type @code{double}, say) need not be
4523 recognized by this macro. So for most machines, this definition
4527 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4530 If the machine has register windows, so that the caller and the called
4531 function use different registers for the return value, this macro
4532 should recognize only the caller's register numbers.
4534 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4535 for a new target instead.
4538 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4539 A target hook that return @code{true} if @var{regno} is the number of a hard
4540 register in which the values of called function may come back.
4542 A register whose use for returning values is limited to serving as the
4543 second of a pair (for a value of type @code{double}, say) need not be
4544 recognized by this target hook.
4546 If the machine has register windows, so that the caller and the called
4547 function use different registers for the return value, this target hook
4548 should recognize only the caller's register numbers.
4550 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4553 @defmac APPLY_RESULT_SIZE
4554 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4555 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4556 saving and restoring an arbitrary return value.
4559 @deftypevr {Target Hook} bool TARGET_OMIT_STRUCT_RETURN_REG
4560 Normally, when a function returns a structure by memory, the address
4561 is passed as an invisible pointer argument, but the compiler also
4562 arranges to return the address from the function like it would a normal
4563 pointer return value. Define this to true if that behavior is
4564 undesirable on your target.
4567 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4568 This hook should return true if values of type @var{type} are returned
4569 at the most significant end of a register (in other words, if they are
4570 padded at the least significant end). You can assume that @var{type}
4571 is returned in a register; the caller is required to check this.
4573 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4574 be able to hold the complete return value. For example, if a 1-, 2-
4575 or 3-byte structure is returned at the most significant end of a
4576 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4580 @node Aggregate Return
4581 @subsection How Large Values Are Returned
4582 @cindex aggregates as return values
4583 @cindex large return values
4584 @cindex returning aggregate values
4585 @cindex structure value address
4587 When a function value's mode is @code{BLKmode} (and in some other
4588 cases), the value is not returned according to
4589 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4590 caller passes the address of a block of memory in which the value
4591 should be stored. This address is called the @dfn{structure value
4594 This section describes how to control returning structure values in
4597 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4598 This target hook should return a nonzero value to say to return the
4599 function value in memory, just as large structures are always returned.
4600 Here @var{type} will be the data type of the value, and @var{fntype}
4601 will be the type of the function doing the returning, or @code{NULL} for
4604 Note that values of mode @code{BLKmode} must be explicitly handled
4605 by this function. Also, the option @option{-fpcc-struct-return}
4606 takes effect regardless of this macro. On most systems, it is
4607 possible to leave the hook undefined; this causes a default
4608 definition to be used, whose value is the constant 1 for @code{BLKmode}
4609 values, and 0 otherwise.
4611 Do not use this hook to indicate that structures and unions should always
4612 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4616 @defmac DEFAULT_PCC_STRUCT_RETURN
4617 Define this macro to be 1 if all structure and union return values must be
4618 in memory. Since this results in slower code, this should be defined
4619 only if needed for compatibility with other compilers or with an ABI@.
4620 If you define this macro to be 0, then the conventions used for structure
4621 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4624 If not defined, this defaults to the value 1.
4627 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4628 This target hook should return the location of the structure value
4629 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4630 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4631 be @code{NULL}, for libcalls. You do not need to define this target
4632 hook if the address is always passed as an ``invisible'' first
4635 On some architectures the place where the structure value address
4636 is found by the called function is not the same place that the
4637 caller put it. This can be due to register windows, or it could
4638 be because the function prologue moves it to a different place.
4639 @var{incoming} is @code{1} or @code{2} when the location is needed in
4640 the context of the called function, and @code{0} in the context of
4643 If @var{incoming} is nonzero and the address is to be found on the
4644 stack, return a @code{mem} which refers to the frame pointer. If
4645 @var{incoming} is @code{2}, the result is being used to fetch the
4646 structure value address at the beginning of a function. If you need
4647 to emit adjusting code, you should do it at this point.
4650 @defmac PCC_STATIC_STRUCT_RETURN
4651 Define this macro if the usual system convention on the target machine
4652 for returning structures and unions is for the called function to return
4653 the address of a static variable containing the value.
4655 Do not define this if the usual system convention is for the caller to
4656 pass an address to the subroutine.
4658 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4659 nothing when you use @option{-freg-struct-return} mode.
4662 @deftypefn {Target Hook} fixed_size_mode TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4663 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.
4666 @deftypefn {Target Hook} fixed_size_mode TARGET_GET_RAW_ARG_MODE (int @var{regno})
4667 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.
4670 @deftypefn {Target Hook} bool TARGET_EMPTY_RECORD_P (const_tree @var{type})
4671 This target hook returns true if the type is an empty record. The default
4672 is to return @code{false}.
4675 @deftypefn {Target Hook} void TARGET_WARN_PARAMETER_PASSING_ABI (cumulative_args_t @var{ca}, tree @var{type})
4676 This target hook warns about the change in empty class parameter passing
4681 @subsection Caller-Saves Register Allocation
4683 If you enable it, GCC can save registers around function calls. This
4684 makes it possible to use call-clobbered registers to hold variables that
4685 must live across calls.
4687 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4688 A C expression specifying which mode is required for saving @var{nregs}
4689 of a pseudo-register in call-clobbered hard register @var{regno}. If
4690 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4691 returned. For most machines this macro need not be defined since GCC
4692 will select the smallest suitable mode.
4695 @node Function Entry
4696 @subsection Function Entry and Exit
4697 @cindex function entry and exit
4701 This section describes the macros that output function entry
4702 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4704 @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})
4705 Generate a patchable area at the function start, consisting of
4706 @var{patch_area_size} NOP instructions. If the target supports named
4707 sections and if @var{record_p} is true, insert a pointer to the current
4708 location in the table of patchable functions. The default implementation
4709 of the hook places the table of pointers in the special section named
4710 @code{__patchable_function_entries}.
4713 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file})
4714 If defined, a function that outputs the assembler code for entry to a
4715 function. The prologue is responsible for setting up the stack frame,
4716 initializing the frame pointer register, saving registers that must be
4717 saved, and allocating @var{size} additional bytes of storage for the
4718 local variables. @var{file} is a stdio stream to which the assembler
4719 code should be output.
4721 The label for the beginning of the function need not be output by this
4722 macro. That has already been done when the macro is run.
4724 @findex regs_ever_live
4725 To determine which registers to save, the macro can refer to the array
4726 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4727 @var{r} is used anywhere within the function. This implies the function
4728 prologue should save register @var{r}, provided it is not one of the
4729 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4730 @code{regs_ever_live}.)
4732 On machines that have ``register windows'', the function entry code does
4733 not save on the stack the registers that are in the windows, even if
4734 they are supposed to be preserved by function calls; instead it takes
4735 appropriate steps to ``push'' the register stack, if any non-call-used
4736 registers are used in the function.
4738 @findex frame_pointer_needed
4739 On machines where functions may or may not have frame-pointers, the
4740 function entry code must vary accordingly; it must set up the frame
4741 pointer if one is wanted, and not otherwise. To determine whether a
4742 frame pointer is in wanted, the macro can refer to the variable
4743 @code{frame_pointer_needed}. The variable's value will be 1 at run
4744 time in a function that needs a frame pointer. @xref{Elimination}.
4746 The function entry code is responsible for allocating any stack space
4747 required for the function. This stack space consists of the regions
4748 listed below. In most cases, these regions are allocated in the
4749 order listed, with the last listed region closest to the top of the
4750 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4751 the highest address if it is not defined). You can use a different order
4752 for a machine if doing so is more convenient or required for
4753 compatibility reasons. Except in cases where required by standard
4754 or by a debugger, there is no reason why the stack layout used by GCC
4755 need agree with that used by other compilers for a machine.
4758 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4759 If defined, a function that outputs assembler code at the end of a
4760 prologue. This should be used when the function prologue is being
4761 emitted as RTL, and you have some extra assembler that needs to be
4762 emitted. @xref{prologue instruction pattern}.
4765 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4766 If defined, a function that outputs assembler code at the start of an
4767 epilogue. This should be used when the function epilogue is being
4768 emitted as RTL, and you have some extra assembler that needs to be
4769 emitted. @xref{epilogue instruction pattern}.
4772 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file})
4773 If defined, a function that outputs the assembler code for exit from a
4774 function. The epilogue is responsible for restoring the saved
4775 registers and stack pointer to their values when the function was
4776 called, and returning control to the caller. This macro takes the
4777 same argument as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4778 registers to restore are determined from @code{regs_ever_live} and
4779 @code{CALL_USED_REGISTERS} in the same way.
4781 On some machines, there is a single instruction that does all the work
4782 of returning from the function. On these machines, give that
4783 instruction the name @samp{return} and do not define the macro
4784 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4786 Do not define a pattern named @samp{return} if you want the
4787 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4788 switches to control whether return instructions or epilogues are used,
4789 define a @samp{return} pattern with a validity condition that tests the
4790 target switches appropriately. If the @samp{return} pattern's validity
4791 condition is false, epilogues will be used.
4793 On machines where functions may or may not have frame-pointers, the
4794 function exit code must vary accordingly. Sometimes the code for these
4795 two cases is completely different. To determine whether a frame pointer
4796 is wanted, the macro can refer to the variable
4797 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4798 a function that needs a frame pointer.
4800 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4801 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4802 The C variable @code{current_function_is_leaf} is nonzero for such a
4803 function. @xref{Leaf Functions}.
4805 On some machines, some functions pop their arguments on exit while
4806 others leave that for the caller to do. For example, the 68020 when
4807 given @option{-mrtd} pops arguments in functions that take a fixed
4808 number of arguments.
4811 @findex crtl->args.pops_args
4812 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4813 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4814 needs to know what was decided. The number of bytes of the current
4815 function's arguments that this function should pop is available in
4816 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4821 @findex pretend_args_size
4822 @findex crtl->args.pretend_args_size
4823 A region of @code{crtl->args.pretend_args_size} bytes of
4824 uninitialized space just underneath the first argument arriving on the
4825 stack. (This may not be at the very start of the allocated stack region
4826 if the calling sequence has pushed anything else since pushing the stack
4827 arguments. But usually, on such machines, nothing else has been pushed
4828 yet, because the function prologue itself does all the pushing.) This
4829 region is used on machines where an argument may be passed partly in
4830 registers and partly in memory, and, in some cases to support the
4831 features in @code{<stdarg.h>}.
4834 An area of memory used to save certain registers used by the function.
4835 The size of this area, which may also include space for such things as
4836 the return address and pointers to previous stack frames, is
4837 machine-specific and usually depends on which registers have been used
4838 in the function. Machines with register windows often do not require
4842 A region of at least @var{size} bytes, possibly rounded up to an allocation
4843 boundary, to contain the local variables of the function. On some machines,
4844 this region and the save area may occur in the opposite order, with the
4845 save area closer to the top of the stack.
4848 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4849 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4850 @code{crtl->outgoing_args_size} bytes to be used for outgoing
4851 argument lists of the function. @xref{Stack Arguments}.
4854 @defmac EXIT_IGNORE_STACK
4855 Define this macro as a C expression that is nonzero if the return
4856 instruction or the function epilogue ignores the value of the stack
4857 pointer; in other words, if it is safe to delete an instruction to
4858 adjust the stack pointer before a return from the function. The
4861 Note that this macro's value is relevant only for functions for which
4862 frame pointers are maintained. It is never safe to delete a final
4863 stack adjustment in a function that has no frame pointer, and the
4864 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4867 @defmac EPILOGUE_USES (@var{regno})
4868 Define this macro as a C expression that is nonzero for registers that are
4869 used by the epilogue or the @samp{return} pattern. The stack and frame
4870 pointer registers are already assumed to be used as needed.
4873 @defmac EH_USES (@var{regno})
4874 Define this macro as a C expression that is nonzero for registers that are
4875 used by the exception handling mechanism, and so should be considered live
4876 on entry to an exception edge.
4879 @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})
4880 A function that outputs the assembler code for a thunk
4881 function, used to implement C++ virtual function calls with multiple
4882 inheritance. The thunk acts as a wrapper around a virtual function,
4883 adjusting the implicit object parameter before handing control off to
4886 First, emit code to add the integer @var{delta} to the location that
4887 contains the incoming first argument. Assume that this argument
4888 contains a pointer, and is the one used to pass the @code{this} pointer
4889 in C++. This is the incoming argument @emph{before} the function prologue,
4890 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4891 all other incoming arguments.
4893 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4894 made after adding @code{delta}. In particular, if @var{p} is the
4895 adjusted pointer, the following adjustment should be made:
4898 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4901 After the additions, emit code to jump to @var{function}, which is a
4902 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4903 not touch the return address. Hence returning from @var{FUNCTION} will
4904 return to whoever called the current @samp{thunk}.
4906 The effect must be as if @var{function} had been called directly with
4907 the adjusted first argument. This macro is responsible for emitting all
4908 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4909 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4911 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4912 have already been extracted from it.) It might possibly be useful on
4913 some targets, but probably not.
4915 If you do not define this macro, the target-independent code in the C++
4916 front end will generate a less efficient heavyweight thunk that calls
4917 @var{function} instead of jumping to it. The generic approach does
4918 not support varargs.
4921 @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})
4922 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4923 to output the assembler code for the thunk function specified by the
4924 arguments it is passed, and false otherwise. In the latter case, the
4925 generic approach will be used by the C++ front end, with the limitations
4930 @subsection Generating Code for Profiling
4931 @cindex profiling, code generation
4933 These macros will help you generate code for profiling.
4935 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4936 A C statement or compound statement to output to @var{file} some
4937 assembler code to call the profiling subroutine @code{mcount}.
4940 The details of how @code{mcount} expects to be called are determined by
4941 your operating system environment, not by GCC@. To figure them out,
4942 compile a small program for profiling using the system's installed C
4943 compiler and look at the assembler code that results.
4945 Older implementations of @code{mcount} expect the address of a counter
4946 variable to be loaded into some register. The name of this variable is
4947 @samp{LP} followed by the number @var{labelno}, so you would generate
4948 the name using @samp{LP%d} in a @code{fprintf}.
4951 @defmac PROFILE_HOOK
4952 A C statement or compound statement to output to @var{file} some assembly
4953 code to call the profiling subroutine @code{mcount} even the target does
4954 not support profiling.
4957 @defmac NO_PROFILE_COUNTERS
4958 Define this macro to be an expression with a nonzero value if the
4959 @code{mcount} subroutine on your system does not need a counter variable
4960 allocated for each function. This is true for almost all modern
4961 implementations. If you define this macro, you must not use the
4962 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4965 @defmac PROFILE_BEFORE_PROLOGUE
4966 Define this macro if the code for function profiling should come before
4967 the function prologue. Normally, the profiling code comes after.
4970 @deftypefn {Target Hook} bool TARGET_KEEP_LEAF_WHEN_PROFILED (void)
4971 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.
4975 @subsection Permitting tail calls
4978 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4979 True if it is OK to do sibling call optimization for the specified
4980 call expression @var{exp}. @var{decl} will be the called function,
4981 or @code{NULL} if this is an indirect call.
4983 It is not uncommon for limitations of calling conventions to prevent
4984 tail calls to functions outside the current unit of translation, or
4985 during PIC compilation. The hook is used to enforce these restrictions,
4986 as the @code{sibcall} md pattern cannot fail, or fall over to a
4987 ``normal'' call. The criteria for successful sibling call optimization
4988 may vary greatly between different architectures.
4991 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4992 Add any hard registers to @var{regs} that are live on entry to the
4993 function. This hook only needs to be defined to provide registers that
4994 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4995 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4996 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4997 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
5000 @deftypefn {Target Hook} void TARGET_SET_UP_BY_PROLOGUE (struct hard_reg_set_container *@var{})
5001 This hook should add additional registers that are computed by the prologue to the hard regset for shrink-wrapping optimization purposes.
5004 @deftypefn {Target Hook} bool TARGET_WARN_FUNC_RETURN (tree)
5005 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.
5008 @node Shrink-wrapping separate components
5009 @subsection Shrink-wrapping separate components
5010 @cindex shrink-wrapping separate components
5012 The prologue may perform a variety of target dependent tasks such as
5013 saving callee-saved registers, saving the return address, aligning the
5014 stack, creating a stack frame, initializing the PIC register, setting
5015 up the static chain, etc.
5017 On some targets some of these tasks may be independent of others and
5018 thus may be shrink-wrapped separately. These independent tasks are
5019 referred to as components and are handled generically by the target
5020 independent parts of GCC.
5022 Using the following hooks those prologue or epilogue components can be
5023 shrink-wrapped separately, so that the initialization (and possibly
5024 teardown) those components do is not done as frequently on execution
5025 paths where this would unnecessary.
5027 What exactly those components are is up to the target code; the generic
5028 code treats them abstractly, as a bit in an @code{sbitmap}. These
5029 @code{sbitmap}s are allocated by the @code{shrink_wrap.get_separate_components}
5030 and @code{shrink_wrap.components_for_bb} hooks, and deallocated by the
5033 @deftypefn {Target Hook} sbitmap TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS (void)
5034 This hook should return an @code{sbitmap} with the bits set for those
5035 components that can be separately shrink-wrapped in the current function.
5036 Return @code{NULL} if the current function should not get any separate
5038 Don't define this hook if it would always return @code{NULL}.
5039 If it is defined, the other hooks in this group have to be defined as well.
5042 @deftypefn {Target Hook} sbitmap TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB (basic_block)
5043 This hook should return an @code{sbitmap} with the bits set for those
5044 components where either the prologue component has to be executed before
5045 the @code{basic_block}, or the epilogue component after it, or both.
5048 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS (sbitmap @var{components}, edge @var{e}, sbitmap @var{edge_components}, bool @var{is_prologue})
5049 This hook should clear the bits in the @var{components} bitmap for those
5050 components in @var{edge_components} that the target cannot handle on edge
5051 @var{e}, where @var{is_prologue} says if this is for a prologue or an
5055 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS (sbitmap)
5056 Emit prologue insns for the components indicated by the parameter.
5059 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS (sbitmap)
5060 Emit epilogue insns for the components indicated by the parameter.
5063 @deftypefn {Target Hook} void TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS (sbitmap)
5064 Mark the components in the parameter as handled, so that the
5065 @code{prologue} and @code{epilogue} named patterns know to ignore those
5066 components. The target code should not hang on to the @code{sbitmap}, it
5067 will be deleted after this call.
5070 @node Stack Smashing Protection
5071 @subsection Stack smashing protection
5072 @cindex stack smashing protection
5074 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
5075 This hook returns a @code{DECL} node for the external variable to use
5076 for the stack protection guard. This variable is initialized by the
5077 runtime to some random value and is used to initialize the guard value
5078 that is placed at the top of the local stack frame. The type of this
5079 variable must be @code{ptr_type_node}.
5081 The default version of this hook creates a variable called
5082 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
5085 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
5086 This hook returns a @code{CALL_EXPR} that alerts the runtime that the
5087 stack protect guard variable has been modified. This expression should
5088 involve a call to a @code{noreturn} function.
5090 The default version of this hook invokes a function called
5091 @samp{__stack_chk_fail}, taking no arguments. This function is
5092 normally defined in @file{libgcc2.c}.
5095 @deftypefn {Target Hook} bool TARGET_STACK_PROTECT_RUNTIME_ENABLED_P (void)
5096 Returns true if the target wants GCC's default stack protect runtime support, otherwise return false. The default implementation always returns true.
5099 @deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
5100 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
5103 @deftypefn {Common Target Hook} {vec<const char *>} TARGET_GET_VALID_OPTION_VALUES (int @var{option_code}, const char *@var{prefix})
5104 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.
5107 @node Miscellaneous Register Hooks
5108 @subsection Miscellaneous register hooks
5109 @cindex miscellaneous register hooks
5111 @deftypevr {Target Hook} bool TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
5112 Set to true if each call that binds to a local definition explicitly
5113 clobbers or sets all non-fixed registers modified by performing the call.
5114 That is, by the call pattern itself, or by code that might be inserted by the
5115 linker (e.g.@: stubs, veneers, branch islands), but not including those
5116 modifiable by the callee. The affected registers may be mentioned explicitly
5117 in the call pattern, or included as clobbers in CALL_INSN_FUNCTION_USAGE.
5118 The default version of this hook is set to false. The purpose of this hook
5119 is to enable the fipa-ra optimization.
5123 @section Implementing the Varargs Macros
5124 @cindex varargs implementation
5126 GCC comes with an implementation of @code{<varargs.h>} and
5127 @code{<stdarg.h>} that work without change on machines that pass arguments
5128 on the stack. Other machines require their own implementations of
5129 varargs, and the two machine independent header files must have
5130 conditionals to include it.
5132 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
5133 the calling convention for @code{va_start}. The traditional
5134 implementation takes just one argument, which is the variable in which
5135 to store the argument pointer. The ISO implementation of
5136 @code{va_start} takes an additional second argument. The user is
5137 supposed to write the last named argument of the function here.
5139 However, @code{va_start} should not use this argument. The way to find
5140 the end of the named arguments is with the built-in functions described
5143 @defmac __builtin_saveregs ()
5144 Use this built-in function to save the argument registers in memory so
5145 that the varargs mechanism can access them. Both ISO and traditional
5146 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
5147 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
5149 On some machines, @code{__builtin_saveregs} is open-coded under the
5150 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
5151 other machines, it calls a routine written in assembler language,
5152 found in @file{libgcc2.c}.
5154 Code generated for the call to @code{__builtin_saveregs} appears at the
5155 beginning of the function, as opposed to where the call to
5156 @code{__builtin_saveregs} is written, regardless of what the code is.
5157 This is because the registers must be saved before the function starts
5158 to use them for its own purposes.
5159 @c i rewrote the first sentence above to fix an overfull hbox. --mew
5163 @defmac __builtin_next_arg (@var{lastarg})
5164 This builtin returns the address of the first anonymous stack
5165 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
5166 returns the address of the location above the first anonymous stack
5167 argument. Use it in @code{va_start} to initialize the pointer for
5168 fetching arguments from the stack. Also use it in @code{va_start} to
5169 verify that the second parameter @var{lastarg} is the last named argument
5170 of the current function.
5173 @defmac __builtin_classify_type (@var{object})
5174 Since each machine has its own conventions for which data types are
5175 passed in which kind of register, your implementation of @code{va_arg}
5176 has to embody these conventions. The easiest way to categorize the
5177 specified data type is to use @code{__builtin_classify_type} together
5178 with @code{sizeof} and @code{__alignof__}.
5180 @code{__builtin_classify_type} ignores the value of @var{object},
5181 considering only its data type. It returns an integer describing what
5182 kind of type that is---integer, floating, pointer, structure, and so on.
5184 The file @file{typeclass.h} defines an enumeration that you can use to
5185 interpret the values of @code{__builtin_classify_type}.
5188 These machine description macros help implement varargs:
5190 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5191 If defined, this hook produces the machine-specific code for a call to
5192 @code{__builtin_saveregs}. This code will be moved to the very
5193 beginning of the function, before any parameter access are made. The
5194 return value of this function should be an RTX that contains the value
5195 to use as the return of @code{__builtin_saveregs}.
5198 @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})
5199 This target hook offers an alternative to using
5200 @code{__builtin_saveregs} and defining the hook
5201 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5202 register arguments into the stack so that all the arguments appear to
5203 have been passed consecutively on the stack. Once this is done, you can
5204 use the standard implementation of varargs that works for machines that
5205 pass all their arguments on the stack.
5207 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5208 structure, containing the values that are obtained after processing the
5209 named arguments. The argument @var{arg} describes the last of these named
5212 The target hook should do two things: first, push onto the stack all the
5213 argument registers @emph{not} used for the named arguments, and second,
5214 store the size of the data thus pushed into the @code{int}-valued
5215 variable pointed to by @var{pretend_args_size}. The value that you
5216 store here will serve as additional offset for setting up the stack
5219 Because you must generate code to push the anonymous arguments at
5220 compile time without knowing their data types,
5221 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5222 have just a single category of argument register and use it uniformly
5225 If the argument @var{second_time} is nonzero, it means that the
5226 arguments of the function are being analyzed for the second time. This
5227 happens for an inline function, which is not actually compiled until the
5228 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5229 not generate any instructions in this case.
5232 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca})
5233 Define this hook to return @code{true} if the location where a function
5234 argument is passed depends on whether or not it is a named argument.
5236 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5237 is set for varargs and stdarg functions. If this hook returns
5238 @code{true}, the @var{named} argument is always true for named
5239 arguments, and false for unnamed arguments. If it returns @code{false},
5240 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5241 then all arguments are treated as named. Otherwise, all named arguments
5242 except the last are treated as named.
5244 You need not define this hook if it always returns @code{false}.
5247 @deftypefn {Target Hook} void TARGET_CALL_ARGS (rtx, @var{tree})
5248 While generating RTL for a function call, this target hook is invoked once
5249 for each argument passed to the function, either a register returned by
5250 @code{TARGET_FUNCTION_ARG} or a memory location. It is called just
5251 before the point where argument registers are stored. The type of the
5252 function to be called is also passed as the second argument; it is
5253 @code{NULL_TREE} for libcalls. The @code{TARGET_END_CALL_ARGS} hook is
5254 invoked just after the code to copy the return reg has been emitted.
5255 This functionality can be used to perform special setup of call argument
5256 registers if a target needs it.
5257 For functions without arguments, the hook is called once with @code{pc_rtx}
5258 passed instead of an argument register.
5259 Most ports do not need to implement anything for this hook.
5262 @deftypefn {Target Hook} void TARGET_END_CALL_ARGS (void)
5263 This target hook is invoked while generating RTL for a function call,
5264 just after the point where the return reg is copied into a pseudo. It
5265 signals that all the call argument and return registers for the just
5266 emitted call are now no longer in use.
5267 Most ports do not need to implement anything for this hook.
5270 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca})
5271 If you need to conditionally change ABIs so that one works with
5272 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5273 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5274 defined, then define this hook to return @code{true} if
5275 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5276 Otherwise, you should not define this hook.
5279 @deftypefn {Target Hook} rtx TARGET_LOAD_BOUNDS_FOR_ARG (rtx @var{slot}, rtx @var{arg}, rtx @var{slot_no})
5280 This hook is used by expand pass to emit insn to load bounds of
5281 @var{arg} passed in @var{slot}. Expand pass uses this hook in case
5282 bounds of @var{arg} are not passed in register. If @var{slot} is a
5283 memory, then bounds are loaded as for regular pointer loaded from
5284 memory. If @var{slot} is not a memory then @var{slot_no} is an integer
5285 constant holding number of the target dependent special slot which
5286 should be used to obtain bounds. Hook returns RTX holding loaded bounds.
5289 @deftypefn {Target Hook} void TARGET_STORE_BOUNDS_FOR_ARG (rtx @var{arg}, rtx @var{slot}, rtx @var{bounds}, rtx @var{slot_no})
5290 This hook is used by expand pass to emit insns to store @var{bounds} of
5291 @var{arg} passed in @var{slot}. Expand pass uses this hook in case
5292 @var{bounds} of @var{arg} are not passed in register. If @var{slot} is a
5293 memory, then @var{bounds} are stored as for regular pointer stored in
5294 memory. If @var{slot} is not a memory then @var{slot_no} is an integer
5295 constant holding number of the target dependent special slot which
5296 should be used to store @var{bounds}.
5299 @deftypefn {Target Hook} rtx TARGET_LOAD_RETURNED_BOUNDS (rtx @var{slot})
5300 This hook is used by expand pass to emit insn to load bounds
5301 returned by function call in @var{slot}. Hook returns RTX holding
5305 @deftypefn {Target Hook} void TARGET_STORE_RETURNED_BOUNDS (rtx @var{slot}, rtx @var{bounds})
5306 This hook is used by expand pass to emit insn to store @var{bounds}
5307 returned by function call into @var{slot}.
5311 @section Support for Nested Functions
5312 @cindex support for nested functions
5313 @cindex trampolines for nested functions
5314 @cindex descriptors for nested functions
5315 @cindex nested functions, support for
5317 Taking the address of a nested function requires special compiler
5318 handling to ensure that the static chain register is loaded when
5319 the function is invoked via an indirect call.
5321 GCC has traditionally supported nested functions by creating an
5322 executable @dfn{trampoline} at run time when the address of a nested
5323 function is taken. This is a small piece of code which normally
5324 resides on the stack, in the stack frame of the containing function.
5325 The trampoline loads the static chain register and then jumps to the
5326 real address of the nested function.
5328 The use of trampolines requires an executable stack, which is a
5329 security risk. To avoid this problem, GCC also supports another
5330 strategy: using descriptors for nested functions. Under this model,
5331 taking the address of a nested function results in a pointer to a
5332 non-executable function descriptor object. Initializing the static chain
5333 from the descriptor is handled at indirect call sites.
5335 On some targets, including HPPA and IA-64, function descriptors may be
5336 mandated by the ABI or be otherwise handled in a target-specific way
5337 by the back end in its code generation strategy for indirect calls.
5338 GCC also provides its own generic descriptor implementation to support the
5339 @option{-fno-trampolines} option. In this case runtime detection of
5340 function descriptors at indirect call sites relies on descriptor
5341 pointers being tagged with a bit that is never set in bare function
5342 addresses. Since GCC's generic function descriptors are
5343 not ABI-compliant, this option is typically used only on a
5344 per-language basis (notably by Ada) or when it can otherwise be
5345 applied to the whole program.
5347 Define the following hook if your backend either implements ABI-specified
5348 descriptor support, or can use GCC's generic descriptor implementation
5349 for nested functions.
5351 @deftypevr {Target Hook} int TARGET_CUSTOM_FUNCTION_DESCRIPTORS
5352 If the target can use GCC's generic descriptor mechanism for nested
5353 functions, define this hook to a power of 2 representing an unused bit
5354 in function pointers which can be used to differentiate descriptors at
5355 run time. This value gives the number of bytes by which descriptor
5356 pointers are misaligned compared to function pointers. For example, on
5357 targets that require functions to be aligned to a 4-byte boundary, a
5358 value of either 1 or 2 is appropriate unless the architecture already
5359 reserves the bit for another purpose, such as on ARM.
5361 Define this hook to 0 if the target implements ABI support for
5362 function descriptors in its standard calling sequence, like for example
5365 Using descriptors for nested functions
5366 eliminates the need for trampolines that reside on the stack and require
5367 it to be made executable.
5370 The following macros tell GCC how to generate code to allocate and
5371 initialize an executable trampoline. You can also use this interface
5372 if your back end needs to create ABI-specified non-executable descriptors; in
5373 this case the "trampoline" created is the descriptor containing data only.
5375 The instructions in an executable trampoline must do two things: load
5376 a constant address into the static chain register, and jump to the real
5377 address of the nested function. On CISC machines such as the m68k,
5378 this requires two instructions, a move immediate and a jump. Then the
5379 two addresses exist in the trampoline as word-long immediate operands.
5380 On RISC machines, it is often necessary to load each address into a
5381 register in two parts. Then pieces of each address form separate
5384 The code generated to initialize the trampoline must store the variable
5385 parts---the static chain value and the function address---into the
5386 immediate operands of the instructions. On a CISC machine, this is
5387 simply a matter of copying each address to a memory reference at the
5388 proper offset from the start of the trampoline. On a RISC machine, it
5389 may be necessary to take out pieces of the address and store them
5392 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5393 This hook is called by @code{assemble_trampoline_template} to output,
5394 on the stream @var{f}, assembler code for a block of data that contains
5395 the constant parts of a trampoline. This code should not include a
5396 label---the label is taken care of automatically.
5398 If you do not define this hook, it means no template is needed
5399 for the target. Do not define this hook on systems where the block move
5400 code to copy the trampoline into place would be larger than the code
5401 to generate it on the spot.
5404 @defmac TRAMPOLINE_SECTION
5405 Return the section into which the trampoline template is to be placed
5406 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5409 @defmac TRAMPOLINE_SIZE
5410 A C expression for the size in bytes of the trampoline, as an integer.
5413 @defmac TRAMPOLINE_ALIGNMENT
5414 Alignment required for trampolines, in bits.
5416 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5417 is used for aligning trampolines.
5420 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5421 This hook is called to initialize a trampoline.
5422 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5423 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5424 RTX for the static chain value that should be passed to the function
5427 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5428 first thing this hook should do is emit a block move into @var{m_tramp}
5429 from the memory block returned by @code{assemble_trampoline_template}.
5430 Note that the block move need only cover the constant parts of the
5431 trampoline. If the target isolates the variable parts of the trampoline
5432 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5434 If the target requires any other actions, such as flushing caches or
5435 enabling stack execution, these actions should be performed after
5436 initializing the trampoline proper.
5439 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5440 This hook should perform any machine-specific adjustment in
5441 the address of the trampoline. Its argument contains the address of the
5442 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5443 the address to be used for a function call should be different from the
5444 address at which the template was stored, the different address should
5445 be returned; otherwise @var{addr} should be returned unchanged.
5446 If this hook is not defined, @var{addr} will be used for function calls.
5449 Implementing trampolines is difficult on many machines because they have
5450 separate instruction and data caches. Writing into a stack location
5451 fails to clear the memory in the instruction cache, so when the program
5452 jumps to that location, it executes the old contents.
5454 Here are two possible solutions. One is to clear the relevant parts of
5455 the instruction cache whenever a trampoline is set up. The other is to
5456 make all trampolines identical, by having them jump to a standard
5457 subroutine. The former technique makes trampoline execution faster; the
5458 latter makes initialization faster.
5460 To clear the instruction cache when a trampoline is initialized, define
5461 the following macro.
5463 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5464 If defined, expands to a C expression clearing the @emph{instruction
5465 cache} in the specified interval. The definition of this macro would
5466 typically be a series of @code{asm} statements. Both @var{beg} and
5467 @var{end} are both pointer expressions.
5470 To use a standard subroutine, define the following macro. In addition,
5471 you must make sure that the instructions in a trampoline fill an entire
5472 cache line with identical instructions, or else ensure that the
5473 beginning of the trampoline code is always aligned at the same point in
5474 its cache line. Look in @file{m68k.h} as a guide.
5476 @defmac TRANSFER_FROM_TRAMPOLINE
5477 Define this macro if trampolines need a special subroutine to do their
5478 work. The macro should expand to a series of @code{asm} statements
5479 which will be compiled with GCC@. They go in a library function named
5480 @code{__transfer_from_trampoline}.
5482 If you need to avoid executing the ordinary prologue code of a compiled
5483 C function when you jump to the subroutine, you can do so by placing a
5484 special label of your own in the assembler code. Use one @code{asm}
5485 statement to generate an assembler label, and another to make the label
5486 global. Then trampolines can use that label to jump directly to your
5487 special assembler code.
5491 @section Implicit Calls to Library Routines
5492 @cindex library subroutine names
5493 @cindex @file{libgcc.a}
5495 @c prevent bad page break with this line
5496 Here is an explanation of implicit calls to library routines.
5498 @defmac DECLARE_LIBRARY_RENAMES
5499 This macro, if defined, should expand to a piece of C code that will get
5500 expanded when compiling functions for libgcc.a. It can be used to
5501 provide alternate names for GCC's internal library functions if there
5502 are ABI-mandated names that the compiler should provide.
5505 @findex set_optab_libfunc
5506 @findex init_one_libfunc
5507 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5508 This hook should declare additional library routines or rename
5509 existing ones, using the functions @code{set_optab_libfunc} and
5510 @code{init_one_libfunc} defined in @file{optabs.c}.
5511 @code{init_optabs} calls this macro after initializing all the normal
5514 The default is to do nothing. Most ports don't need to define this hook.
5517 @deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX
5518 If false (the default), internal library routines start with two
5519 underscores. If set to true, these routines start with @code{__gnu_}
5520 instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This
5521 currently only affects functions defined in @file{libgcc2.c}. If this
5522 is set to true, the @file{tm.h} file must also
5523 @code{#define LIBGCC2_GNU_PREFIX}.
5526 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5527 This macro should return @code{true} if the library routine that
5528 implements the floating point comparison operator @var{comparison} in
5529 mode @var{mode} will return a boolean, and @var{false} if it will
5532 GCC's own floating point libraries return tristates from the
5533 comparison operators, so the default returns false always. Most ports
5534 don't need to define this macro.
5537 @defmac TARGET_LIB_INT_CMP_BIASED
5538 This macro should evaluate to @code{true} if the integer comparison
5539 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5540 operand is smaller than the second, 1 to indicate that they are equal,
5541 and 2 to indicate that the first operand is greater than the second.
5542 If this macro evaluates to @code{false} the comparison functions return
5543 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5544 in @file{libgcc.a}, you do not need to define this macro.
5547 @defmac TARGET_HAS_NO_HW_DIVIDE
5548 This macro should be defined if the target has no hardware divide
5549 instructions. If this macro is defined, GCC will use an algorithm which
5550 make use of simple logical and arithmetic operations for 64-bit
5551 division. If the macro is not defined, GCC will use an algorithm which
5552 make use of a 64-bit by 32-bit divide primitive.
5555 @cindex @code{EDOM}, implicit usage
5558 The value of @code{EDOM} on the target machine, as a C integer constant
5559 expression. If you don't define this macro, GCC does not attempt to
5560 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5561 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5564 If you do not define @code{TARGET_EDOM}, then compiled code reports
5565 domain errors by calling the library function and letting it report the
5566 error. If mathematical functions on your system use @code{matherr} when
5567 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5568 that @code{matherr} is used normally.
5571 @cindex @code{errno}, implicit usage
5572 @defmac GEN_ERRNO_RTX
5573 Define this macro as a C expression to create an rtl expression that
5574 refers to the global ``variable'' @code{errno}. (On certain systems,
5575 @code{errno} may not actually be a variable.) If you don't define this
5576 macro, a reasonable default is used.
5579 @deftypefn {Target Hook} bool TARGET_LIBC_HAS_FUNCTION (enum function_class @var{fn_class})
5580 This hook determines whether a function from a class of functions
5581 @var{fn_class} is present in the target C library.
5584 @deftypefn {Target Hook} bool TARGET_LIBC_HAS_FAST_FUNCTION (int @var{fcode})
5585 This hook determines whether a function from a class of functions
5586 @code{(enum function_class)}@var{fcode} has a fast implementation.
5589 @defmac NEXT_OBJC_RUNTIME
5590 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
5591 by default. This calling convention involves passing the object, the selector
5592 and the method arguments all at once to the method-lookup library function.
5593 This is the usual setting when targeting Darwin/Mac OS X systems, which have
5594 the NeXT runtime installed.
5596 If the macro is set to 0, the "GNU" Objective-C message sending convention
5597 will be used by default. This convention passes just the object and the
5598 selector to the method-lookup function, which returns a pointer to the method.
5600 In either case, it remains possible to select code-generation for the alternate
5601 scheme, by means of compiler command line switches.
5604 @node Addressing Modes
5605 @section Addressing Modes
5606 @cindex addressing modes
5608 @c prevent bad page break with this line
5609 This is about addressing modes.
5611 @defmac HAVE_PRE_INCREMENT
5612 @defmacx HAVE_PRE_DECREMENT
5613 @defmacx HAVE_POST_INCREMENT
5614 @defmacx HAVE_POST_DECREMENT
5615 A C expression that is nonzero if the machine supports pre-increment,
5616 pre-decrement, post-increment, or post-decrement addressing respectively.
5619 @defmac HAVE_PRE_MODIFY_DISP
5620 @defmacx HAVE_POST_MODIFY_DISP
5621 A C expression that is nonzero if the machine supports pre- or
5622 post-address side-effect generation involving constants other than
5623 the size of the memory operand.
5626 @defmac HAVE_PRE_MODIFY_REG
5627 @defmacx HAVE_POST_MODIFY_REG
5628 A C expression that is nonzero if the machine supports pre- or
5629 post-address side-effect generation involving a register displacement.
5632 @defmac CONSTANT_ADDRESS_P (@var{x})
5633 A C expression that is 1 if the RTX @var{x} is a constant which
5634 is a valid address. On most machines the default definition of
5635 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5636 is acceptable, but a few machines are more restrictive as to which
5637 constant addresses are supported.
5640 @defmac CONSTANT_P (@var{x})
5641 @code{CONSTANT_P}, which is defined by target-independent code,
5642 accepts integer-values expressions whose values are not explicitly
5643 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5644 expressions and @code{const} arithmetic expressions, in addition to
5645 @code{const_int} and @code{const_double} expressions.
5648 @defmac MAX_REGS_PER_ADDRESS
5649 A number, the maximum number of registers that can appear in a valid
5650 memory address. Note that it is up to you to specify a value equal to
5651 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5655 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5656 A function that returns whether @var{x} (an RTX) is a legitimate memory
5657 address on the target machine for a memory operand of mode @var{mode}.
5659 Legitimate addresses are defined in two variants: a strict variant and a
5660 non-strict one. The @var{strict} parameter chooses which variant is
5661 desired by the caller.
5663 The strict variant is used in the reload pass. It must be defined so
5664 that any pseudo-register that has not been allocated a hard register is
5665 considered a memory reference. This is because in contexts where some
5666 kind of register is required, a pseudo-register with no hard register
5667 must be rejected. For non-hard registers, the strict variant should look
5668 up the @code{reg_renumber} array; it should then proceed using the hard
5669 register number in the array, or treat the pseudo as a memory reference
5670 if the array holds @code{-1}.
5672 The non-strict variant is used in other passes. It must be defined to
5673 accept all pseudo-registers in every context where some kind of
5674 register is required.
5676 Normally, constant addresses which are the sum of a @code{symbol_ref}
5677 and an integer are stored inside a @code{const} RTX to mark them as
5678 constant. Therefore, there is no need to recognize such sums
5679 specifically as legitimate addresses. Normally you would simply
5680 recognize any @code{const} as legitimate.
5682 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5683 sums that are not marked with @code{const}. It assumes that a naked
5684 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5685 naked constant sums as illegitimate addresses, so that none of them will
5686 be given to @code{PRINT_OPERAND_ADDRESS}.
5688 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5689 On some machines, whether a symbolic address is legitimate depends on
5690 the section that the address refers to. On these machines, define the
5691 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5692 into the @code{symbol_ref}, and then check for it here. When you see a
5693 @code{const}, you will have to look inside it to find the
5694 @code{symbol_ref} in order to determine the section. @xref{Assembler
5697 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5698 Some ports are still using a deprecated legacy substitute for
5699 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5703 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5707 and should @code{goto @var{label}} if the address @var{x} is a valid
5708 address on the target machine for a memory operand of mode @var{mode}.
5710 @findex REG_OK_STRICT
5711 Compiler source files that want to use the strict variant of this
5712 macro define the macro @code{REG_OK_STRICT}. You should use an
5713 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5714 that case and the non-strict variant otherwise.
5716 Using the hook is usually simpler because it limits the number of
5717 files that are recompiled when changes are made.
5720 @defmac TARGET_MEM_CONSTRAINT
5721 A single character to be used instead of the default @code{'m'}
5722 character for general memory addresses. This defines the constraint
5723 letter which matches the memory addresses accepted by
5724 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5725 support new address formats in your back end without changing the
5726 semantics of the @code{'m'} constraint. This is necessary in order to
5727 preserve functionality of inline assembly constructs using the
5728 @code{'m'} constraint.
5731 @defmac FIND_BASE_TERM (@var{x})
5732 A C expression to determine the base term of address @var{x},
5733 or to provide a simplified version of @var{x} from which @file{alias.c}
5734 can easily find the base term. This macro is used in only two places:
5735 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5737 It is always safe for this macro to not be defined. It exists so
5738 that alias analysis can understand machine-dependent addresses.
5740 The typical use of this macro is to handle addresses containing
5741 a label_ref or symbol_ref within an UNSPEC@.
5744 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode})
5745 This hook is given an invalid memory address @var{x} for an
5746 operand of mode @var{mode} and should try to return a valid memory
5749 @findex break_out_memory_refs
5750 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5751 and @var{oldx} will be the operand that was given to that function to produce
5754 The code of the hook should not alter the substructure of
5755 @var{x}. If it transforms @var{x} into a more legitimate form, it
5756 should return the new @var{x}.
5758 It is not necessary for this hook to come up with a legitimate address,
5759 with the exception of native TLS addresses (@pxref{Emulated TLS}).
5760 The compiler has standard ways of doing so in all cases. In fact, if
5761 the target supports only emulated TLS, it
5762 is safe to omit this hook or make it return @var{x} if it cannot find
5763 a valid way to legitimize the address. But often a machine-dependent
5764 strategy can generate better code.
5767 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5768 A C compound statement that attempts to replace @var{x}, which is an address
5769 that needs reloading, with a valid memory address for an operand of mode
5770 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5771 It is not necessary to define this macro, but it might be useful for
5772 performance reasons.
5774 For example, on the i386, it is sometimes possible to use a single
5775 reload register instead of two by reloading a sum of two pseudo
5776 registers into a register. On the other hand, for number of RISC
5777 processors offsets are limited so that often an intermediate address
5778 needs to be generated in order to address a stack slot. By defining
5779 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5780 generated for adjacent some stack slots can be made identical, and thus
5783 @emph{Note}: This macro should be used with caution. It is necessary
5784 to know something of how reload works in order to effectively use this,
5785 and it is quite easy to produce macros that build in too much knowledge
5786 of reload internals.
5788 @emph{Note}: This macro must be able to reload an address created by a
5789 previous invocation of this macro. If it fails to handle such addresses
5790 then the compiler may generate incorrect code or abort.
5793 The macro definition should use @code{push_reload} to indicate parts that
5794 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5795 suitable to be passed unaltered to @code{push_reload}.
5797 The code generated by this macro must not alter the substructure of
5798 @var{x}. If it transforms @var{x} into a more legitimate form, it
5799 should assign @var{x} (which will always be a C variable) a new value.
5800 This also applies to parts that you change indirectly by calling
5803 @findex strict_memory_address_p
5804 The macro definition may use @code{strict_memory_address_p} to test if
5805 the address has become legitimate.
5808 If you want to change only a part of @var{x}, one standard way of doing
5809 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5810 single level of rtl. Thus, if the part to be changed is not at the
5811 top level, you'll need to replace first the top level.
5812 It is not necessary for this macro to come up with a legitimate
5813 address; but often a machine-dependent strategy can generate better code.
5816 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr}, addr_space_t @var{addrspace})
5817 This hook returns @code{true} if memory address @var{addr} in address
5818 space @var{addrspace} can have
5819 different meanings depending on the machine mode of the memory
5820 reference it is used for or if the address is valid for some modes
5823 Autoincrement and autodecrement addresses typically have mode-dependent
5824 effects because the amount of the increment or decrement is the size
5825 of the operand being addressed. Some machines have other mode-dependent
5826 addresses. Many RISC machines have no mode-dependent addresses.
5828 You may assume that @var{addr} is a valid address for the machine.
5830 The default version of this hook returns @code{false}.
5833 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (machine_mode @var{mode}, rtx @var{x})
5834 This hook returns true if @var{x} is a legitimate constant for a
5835 @var{mode}-mode immediate operand on the target machine. You can assume that
5836 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5838 The default definition returns true.
5841 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5842 This hook is used to undo the possibly obfuscating effects of the
5843 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5844 macros. Some backend implementations of these macros wrap symbol
5845 references inside an @code{UNSPEC} rtx to represent PIC or similar
5846 addressing modes. This target hook allows GCC's optimizers to understand
5847 the semantics of these opaque @code{UNSPEC}s by converting them back
5848 into their original form.
5851 @deftypefn {Target Hook} bool TARGET_CONST_NOT_OK_FOR_DEBUG_P (rtx @var{x})
5852 This hook should return true if @var{x} should not be emitted into
5856 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (machine_mode @var{mode}, rtx @var{x})
5857 This hook should return true if @var{x} is of a form that cannot (or
5858 should not) be spilled to the constant pool. @var{mode} is the mode
5861 The default version of this hook returns false.
5863 The primary reason to define this hook is to prevent reload from
5864 deciding that a non-legitimate constant would be better reloaded
5865 from the constant pool instead of spilling and reloading a register
5866 holding the constant. This restriction is often true of addresses
5867 of TLS symbols for various targets.
5870 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (machine_mode @var{mode}, const_rtx @var{x})
5871 This hook should return true if pool entries for constant @var{x} can
5872 be placed in an @code{object_block} structure. @var{mode} is the mode
5875 The default version returns false for all constants.
5878 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_DECL_P (const_tree @var{decl})
5879 This hook should return true if pool entries for @var{decl} should
5880 be placed in an @code{object_block} structure.
5882 The default version returns true for all decls.
5885 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (tree @var{fndecl})
5886 This hook should return the DECL of a function that implements the
5887 reciprocal of the machine-specific builtin function @var{fndecl}, or
5888 @code{NULL_TREE} if such a function is not available.
5891 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5892 This hook should return the DECL of a function @var{f} that given an
5893 address @var{addr} as an argument returns a mask @var{m} that can be
5894 used to extract from two vectors the relevant data that resides in
5895 @var{addr} in case @var{addr} is not properly aligned.
5897 The autovectorizer, when vectorizing a load operation from an address
5898 @var{addr} that may be unaligned, will generate two vector loads from
5899 the two aligned addresses around @var{addr}. It then generates a
5900 @code{REALIGN_LOAD} operation to extract the relevant data from the
5901 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5902 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5903 the third argument, @var{OFF}, defines how the data will be extracted
5904 from these two vectors: if @var{OFF} is 0, then the returned vector is
5905 @var{v2}; otherwise, the returned vector is composed from the last
5906 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5907 @var{OFF} elements of @var{v2}.
5909 If this hook is defined, the autovectorizer will generate a call
5910 to @var{f} (using the DECL tree that this hook returns) and will
5911 use the return value of @var{f} as the argument @var{OFF} to
5912 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5913 should comply with the semantics expected by @code{REALIGN_LOAD}
5915 If this hook is not defined, then @var{addr} will be used as
5916 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5917 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5920 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5921 Returns cost of different scalar or vector statements for vectorization cost model.
5922 For vector memory operations the cost may depend on type (@var{vectype}) and
5923 misalignment value (@var{misalign}).
5926 @deftypefn {Target Hook} poly_uint64 TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT (const_tree @var{type})
5927 This hook returns the preferred alignment in bits for accesses to
5928 vectors of type @var{type} in vectorized code. This might be less than
5929 or greater than the ABI-defined value returned by
5930 @code{TARGET_VECTOR_ALIGNMENT}. It can be equal to the alignment of
5931 a single element, in which case the vectorizer will not try to optimize
5934 The default hook returns @code{TYPE_ALIGN (@var{type})}, which is
5935 correct for most targets.
5938 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5939 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.
5942 @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})
5943 This hook is used to test whether the target can permute up to two
5944 vectors of mode @var{mode} using the permutation vector @code{sel}, and
5945 also to emit such a permutation. In the former case @var{in0}, @var{in1}
5946 and @var{out} are all null. In the latter case @var{in0} and @var{in1} are
5947 the source vectors and @var{out} is the destination vector; all three are
5948 registers of mode @var{mode}. @var{in1} is the same as @var{in0} if
5949 @var{sel} describes a permutation on one vector instead of two.
5951 Return true if the operation is possible, emitting instructions for it
5952 if rtxes are provided.
5954 @cindex @code{vec_perm@var{m}} instruction pattern
5955 If the hook returns false for a mode with multibyte elements, GCC will
5956 try the equivalent byte operation. If that also fails, it will try forcing
5957 the selector into a register and using the @var{vec_perm@var{mode}}
5958 instruction pattern. There is no need for the hook to handle these two
5959 implementation approaches itself.
5962 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5963 This hook should return the DECL of a function that implements conversion of the
5964 input vector of type @var{src_type} to type @var{dest_type}.
5965 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5966 specifies how the conversion is to be applied
5967 (truncation, rounding, etc.).
5969 If this hook is defined, the autovectorizer will use the
5970 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5971 conversion. Otherwise, it will return @code{NULL_TREE}.
5974 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (unsigned @var{code}, tree @var{vec_type_out}, tree @var{vec_type_in})
5975 This hook should return the decl of a function that implements the
5976 vectorized variant of the function with the @code{combined_fn} code
5977 @var{code} or @code{NULL_TREE} if such a function is not available.
5978 The return type of the vectorized function shall be of vector type
5979 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5982 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5983 This hook should return the decl of a function that implements the
5984 vectorized variant of target built-in function @code{fndecl}. The
5985 return type of the vectorized function shall be of vector type
5986 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5989 @deftypefn {Target Hook} bool TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT (machine_mode @var{mode}, const_tree @var{type}, int @var{misalignment}, bool @var{is_packed})
5990 This hook should return true if the target supports misaligned vector
5991 store/load of a specific factor denoted in the @var{misalignment}
5992 parameter. The vector store/load should be of machine mode @var{mode} and
5993 the elements in the vectors should be of type @var{type}. @var{is_packed}
5994 parameter is true if the memory access is defined in a packed struct.
5997 @deftypefn {Target Hook} machine_mode TARGET_VECTORIZE_PREFERRED_SIMD_MODE (scalar_mode @var{mode})
5998 This hook should return the preferred mode for vectorizing scalar
5999 mode @var{mode}. The default is
6000 equal to @code{word_mode}, because the vectorizer can do some
6001 transformations even in absence of specialized @acronym{SIMD} hardware.
6004 @deftypefn {Target Hook} machine_mode TARGET_VECTORIZE_SPLIT_REDUCTION (machine_mode)
6005 This hook should return the preferred mode to split the final reduction
6006 step on @var{mode} to. The reduction is then carried out reducing upper
6007 against lower halves of vectors recursively until the specified mode is
6008 reached. The default is @var{mode} which means no splitting.
6011 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_MODES (vector_modes *@var{modes}, bool @var{all})
6012 If using the mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}
6013 is not the only approach worth considering, this hook should add one mode to
6014 @var{modes} for each useful alternative approach. These modes are then
6015 passed to @code{TARGET_VECTORIZE_RELATED_MODE} to obtain the vector mode
6016 for a given element mode.
6018 The modes returned in @var{modes} should use the smallest element mode
6019 possible for the vectorization approach that they represent, preferring
6020 integer modes over floating-poing modes in the event of a tie. The first
6021 mode should be the @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE} for its
6024 If @var{all} is true, add suitable vector modes even when they are generally
6025 not expected to be worthwhile.
6027 The hook returns a bitmask of flags that control how the modes in
6028 @var{modes} are used. The flags are:
6030 @item VECT_COMPARE_COSTS
6031 Tells the loop vectorizer to try all the provided modes and pick the one
6032 with the lowest cost. By default the vectorizer will choose the first
6036 The hook does not need to do anything if the vector returned by
6037 @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE} is the only one relevant
6038 for autovectorization. The default implementation adds no modes and
6042 @deftypefn {Target Hook} opt_machine_mode TARGET_VECTORIZE_RELATED_MODE (machine_mode @var{vector_mode}, scalar_mode @var{element_mode}, poly_uint64 @var{nunits})
6043 If a piece of code is using vector mode @var{vector_mode} and also wants
6044 to operate on elements of mode @var{element_mode}, return the vector mode
6045 it should use for those elements. If @var{nunits} is nonzero, ensure that
6046 the mode has exactly @var{nunits} elements, otherwise pick whichever vector
6047 size pairs the most naturally with @var{vector_mode}. Return an empty
6048 @code{opt_machine_mode} if there is no supported vector mode with the
6049 required properties.
6051 There is no prescribed way of handling the case in which @var{nunits}
6052 is zero. One common choice is to pick a vector mode with the same size
6053 as @var{vector_mode}; this is the natural choice if the target has a
6054 fixed vector size. Another option is to choose a vector mode with the
6055 same number of elements as @var{vector_mode}; this is the natural choice
6056 if the target has a fixed number of elements. Alternatively, the hook
6057 might choose a middle ground, such as trying to keep the number of
6058 elements as similar as possible while applying maximum and minimum
6061 The default implementation uses @code{mode_for_vector} to find the
6062 requested mode, returning a mode with the same size as @var{vector_mode}
6063 when @var{nunits} is zero. This is the correct behavior for most targets.
6066 @deftypefn {Target Hook} opt_machine_mode TARGET_VECTORIZE_GET_MASK_MODE (machine_mode @var{mode})
6067 Return the mode to use for a vector mask that holds one boolean
6068 result for each element of vector mode @var{mode}. The returned mask mode
6069 can be a vector of integers (class @code{MODE_VECTOR_INT}), a vector of
6070 booleans (class @code{MODE_VECTOR_BOOL}) or a scalar integer (class
6071 @code{MODE_INT}). Return an empty @code{opt_machine_mode} if no such
6074 The default implementation returns a @code{MODE_VECTOR_INT} with the
6075 same size and number of elements as @var{mode}, if such a mode exists.
6078 @deftypefn {Target Hook} bool TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE (unsigned @var{ifn})
6079 This hook returns true if masked internal function @var{ifn} (really of
6080 type @code{internal_fn}) should be considered expensive when the mask is
6081 all zeros. GCC can then try to branch around the instruction instead.
6084 @deftypefn {Target Hook} {void *} TARGET_VECTORIZE_INIT_COST (class loop *@var{loop_info})
6085 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.
6088 @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})
6089 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.
6092 @deftypefn {Target Hook} void TARGET_VECTORIZE_FINISH_COST (void *@var{data}, unsigned *@var{prologue_cost}, unsigned *@var{body_cost}, unsigned *@var{epilogue_cost})
6093 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.
6096 @deftypefn {Target Hook} void TARGET_VECTORIZE_DESTROY_COST_DATA (void *@var{data})
6097 This hook should release @var{data} and any related data structures allocated by TARGET_VECTORIZE_INIT_COST. The default releases the accumulator.
6100 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_GATHER (const_tree @var{mem_vectype}, const_tree @var{index_type}, int @var{scale})
6101 Target builtin that implements vector gather operation. @var{mem_vectype}
6102 is the vector type of the load and @var{index_type} is scalar type of
6103 the index, scaled by @var{scale}.
6104 The default is @code{NULL_TREE} which means to not vectorize gather
6108 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_SCATTER (const_tree @var{vectype}, const_tree @var{index_type}, int @var{scale})
6109 Target builtin that implements vector scatter operation. @var{vectype}
6110 is the vector type of the store and @var{index_type} is scalar type of
6111 the index, scaled by @var{scale}.
6112 The default is @code{NULL_TREE} which means to not vectorize scatter
6116 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN (struct cgraph_node *@var{}, struct cgraph_simd_clone *@var{}, @var{tree}, @var{int})
6117 This hook should set @var{vecsize_mangle}, @var{vecsize_int}, @var{vecsize_float}
6118 fields in @var{simd_clone} structure pointed by @var{clone_info} argument and also
6119 @var{simdlen} field if it was previously 0.
6120 The hook should return 0 if SIMD clones shouldn't be emitted,
6121 or number of @var{vecsize_mangle} variants that should be emitted.
6124 @deftypefn {Target Hook} void TARGET_SIMD_CLONE_ADJUST (struct cgraph_node *@var{})
6125 This hook should add implicit @code{attribute(target("..."))} attribute
6126 to SIMD clone @var{node} if needed.
6129 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_USABLE (struct cgraph_node *@var{})
6130 This hook should return -1 if SIMD clone @var{node} shouldn't be used
6131 in vectorized loops in current function, or non-negative number if it is
6132 usable. In that case, the smaller the number is, the more desirable it is
6136 @deftypefn {Target Hook} int TARGET_SIMT_VF (void)
6137 Return number of threads in SIMT thread group on the target.
6140 @deftypefn {Target Hook} int TARGET_OMP_DEVICE_KIND_ARCH_ISA (enum omp_device_kind_arch_isa @var{trait}, const char *@var{name})
6141 Return 1 if @var{trait} @var{name} is present in the OpenMP context's
6142 device trait set, return 0 if not present in any OpenMP context in the
6143 whole translation unit, or -1 if not present in the current OpenMP context
6144 but might be present in another OpenMP context in the same TU.
6147 @deftypefn {Target Hook} bool TARGET_GOACC_VALIDATE_DIMS (tree @var{decl}, int *@var{dims}, int @var{fn_level}, unsigned @var{used})
6148 This hook should check the launch dimensions provided for an OpenACC
6149 compute region, or routine. Defaulted values are represented as -1
6150 and non-constant values as 0. The @var{fn_level} is negative for the
6151 function corresponding to the compute region. For a routine is is the
6152 outermost level at which partitioned execution may be spawned. The hook
6153 should verify non-default values. If DECL is NULL, global defaults
6154 are being validated and unspecified defaults should be filled in.
6155 Diagnostics should be issued as appropriate. Return
6156 true, if changes have been made. You must override this hook to
6157 provide dimensions larger than 1.
6160 @deftypefn {Target Hook} int TARGET_GOACC_DIM_LIMIT (int @var{axis})
6161 This hook should return the maximum size of a particular dimension,
6162 or zero if unbounded.
6165 @deftypefn {Target Hook} bool TARGET_GOACC_FORK_JOIN (gcall *@var{call}, const int *@var{dims}, bool @var{is_fork})
6166 This hook can be used to convert IFN_GOACC_FORK and IFN_GOACC_JOIN
6167 function calls to target-specific gimple, or indicate whether they
6168 should be retained. It is executed during the oacc_device_lower pass.
6169 It should return true, if the call should be retained. It should
6170 return false, if it is to be deleted (either because target-specific
6171 gimple has been inserted before it, or there is no need for it).
6172 The default hook returns false, if there are no RTL expanders for them.
6175 @deftypefn {Target Hook} void TARGET_GOACC_REDUCTION (gcall *@var{call})
6176 This hook is used by the oacc_transform pass to expand calls to the
6177 @var{GOACC_REDUCTION} internal function, into a sequence of gimple
6178 instructions. @var{call} is gimple statement containing the call to
6179 the function. This hook removes statement @var{call} after the
6180 expanded sequence has been inserted. This hook is also responsible
6181 for allocating any storage for reductions when necessary.
6184 @deftypefn {Target Hook} tree TARGET_PREFERRED_ELSE_VALUE (unsigned @var{ifn}, tree @var{type}, unsigned @var{nops}, tree *@var{ops})
6185 This hook returns the target's preferred final argument for a call
6186 to conditional internal function @var{ifn} (really of type
6187 @code{internal_fn}). @var{type} specifies the return type of the
6188 function and @var{ops} are the operands to the conditional operation,
6189 of which there are @var{nops}.
6191 For example, if @var{ifn} is @code{IFN_COND_ADD}, the hook returns
6192 a value of type @var{type} that should be used when @samp{@var{ops}[0]}
6193 and @samp{@var{ops}[1]} are conditionally added together.
6195 This hook is only relevant if the target supports conditional patterns
6196 like @code{cond_add@var{m}}. The default implementation returns a zero
6197 constant of type @var{type}.
6200 @node Anchored Addresses
6201 @section Anchored Addresses
6202 @cindex anchored addresses
6203 @cindex @option{-fsection-anchors}
6205 GCC usually addresses every static object as a separate entity.
6206 For example, if we have:
6210 int foo (void) @{ return a + b + c; @}
6213 the code for @code{foo} will usually calculate three separate symbolic
6214 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
6215 it would be better to calculate just one symbolic address and access
6216 the three variables relative to it. The equivalent pseudocode would
6222 register int *xr = &x;
6223 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
6227 (which isn't valid C). We refer to shared addresses like @code{x} as
6228 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
6230 The hooks below describe the target properties that GCC needs to know
6231 in order to make effective use of section anchors. It won't use
6232 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
6233 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
6235 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
6236 The minimum offset that should be applied to a section anchor.
6237 On most targets, it should be the smallest offset that can be
6238 applied to a base register while still giving a legitimate address
6239 for every mode. The default value is 0.
6242 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
6243 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
6244 offset that should be applied to section anchors. The default
6248 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
6249 Write the assembly code to define section anchor @var{x}, which is a
6250 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
6251 The hook is called with the assembly output position set to the beginning
6252 of @code{SYMBOL_REF_BLOCK (@var{x})}.
6254 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
6255 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
6256 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
6257 is @code{NULL}, which disables the use of section anchors altogether.
6260 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
6261 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
6262 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
6263 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
6265 The default version is correct for most targets, but you might need to
6266 intercept this hook to handle things like target-specific attributes
6267 or target-specific sections.
6270 @node Condition Code
6271 @section Condition Code Status
6272 @cindex condition code status
6274 The macros in this section can be split in two families, according to the
6275 two ways of representing condition codes in GCC.
6277 The first representation is the so called @code{(cc0)} representation
6278 (@pxref{Jump Patterns}), where all instructions can have an implicit
6279 clobber of the condition codes. The second is the condition code
6280 register representation, which provides better schedulability for
6281 architectures that do have a condition code register, but on which
6282 most instructions do not affect it. The latter category includes
6285 The implicit clobbering poses a strong restriction on the placement of
6286 the definition and use of the condition code. In the past the definition
6287 and use were always adjacent. However, recent changes to support trapping
6288 arithmatic may result in the definition and user being in different blocks.
6289 Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally,
6290 the definition may be the source of exception handling edges.
6292 These restrictions can prevent important
6293 optimizations on some machines. For example, on the IBM RS/6000, there
6294 is a delay for taken branches unless the condition code register is set
6295 three instructions earlier than the conditional branch. The instruction
6296 scheduler cannot perform this optimization if it is not permitted to
6297 separate the definition and use of the condition code register.
6299 For this reason, it is possible and suggested to use a register to
6300 represent the condition code for new ports. If there is a specific
6301 condition code register in the machine, use a hard register. If the
6302 condition code or comparison result can be placed in any general register,
6303 or if there are multiple condition registers, use a pseudo register.
6304 Registers used to store the condition code value will usually have a mode
6305 that is in class @code{MODE_CC}.
6307 Alternatively, you can use @code{BImode} if the comparison operator is
6308 specified already in the compare instruction. In this case, you are not
6309 interested in most macros in this section.
6312 * CC0 Condition Codes:: Old style representation of condition codes.
6313 * MODE_CC Condition Codes:: Modern representation of condition codes.
6316 @node CC0 Condition Codes
6317 @subsection Representation of condition codes using @code{(cc0)}
6321 The file @file{conditions.h} defines a variable @code{cc_status} to
6322 describe how the condition code was computed (in case the interpretation of
6323 the condition code depends on the instruction that it was set by). This
6324 variable contains the RTL expressions on which the condition code is
6325 currently based, and several standard flags.
6327 Sometimes additional machine-specific flags must be defined in the machine
6328 description header file. It can also add additional machine-specific
6329 information by defining @code{CC_STATUS_MDEP}.
6331 @defmac CC_STATUS_MDEP
6332 C code for a data type which is used for declaring the @code{mdep}
6333 component of @code{cc_status}. It defaults to @code{int}.
6335 This macro is not used on machines that do not use @code{cc0}.
6338 @defmac CC_STATUS_MDEP_INIT
6339 A C expression to initialize the @code{mdep} field to ``empty''.
6340 The default definition does nothing, since most machines don't use
6341 the field anyway. If you want to use the field, you should probably
6342 define this macro to initialize it.
6344 This macro is not used on machines that do not use @code{cc0}.
6347 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
6348 A C compound statement to set the components of @code{cc_status}
6349 appropriately for an insn @var{insn} whose body is @var{exp}. It is
6350 this macro's responsibility to recognize insns that set the condition
6351 code as a byproduct of other activity as well as those that explicitly
6354 This macro is not used on machines that do not use @code{cc0}.
6356 If there are insns that do not set the condition code but do alter
6357 other machine registers, this macro must check to see whether they
6358 invalidate the expressions that the condition code is recorded as
6359 reflecting. For example, on the 68000, insns that store in address
6360 registers do not set the condition code, which means that usually
6361 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
6362 insns. But suppose that the previous insn set the condition code
6363 based on location @samp{a4@@(102)} and the current insn stores a new
6364 value in @samp{a4}. Although the condition code is not changed by
6365 this, it will no longer be true that it reflects the contents of
6366 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
6367 @code{cc_status} in this case to say that nothing is known about the
6368 condition code value.
6370 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
6371 with the results of peephole optimization: insns whose patterns are
6372 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
6373 constants which are just the operands. The RTL structure of these
6374 insns is not sufficient to indicate what the insns actually do. What
6375 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
6376 @code{CC_STATUS_INIT}.
6378 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
6379 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
6380 @samp{cc}. This avoids having detailed information about patterns in
6381 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
6384 @node MODE_CC Condition Codes
6385 @subsection Representation of condition codes using registers
6389 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
6390 On many machines, the condition code may be produced by other instructions
6391 than compares, for example the branch can use directly the condition
6392 code set by a subtract instruction. However, on some machines
6393 when the condition code is set this way some bits (such as the overflow
6394 bit) are not set in the same way as a test instruction, so that a different
6395 branch instruction must be used for some conditional branches. When
6396 this happens, use the machine mode of the condition code register to
6397 record different formats of the condition code register. Modes can
6398 also be used to record which compare instruction (e.g.@: a signed or an
6399 unsigned comparison) produced the condition codes.
6401 If other modes than @code{CCmode} are required, add them to
6402 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
6403 a mode given an operand of a compare. This is needed because the modes
6404 have to be chosen not only during RTL generation but also, for example,
6405 by instruction combination. The result of @code{SELECT_CC_MODE} should
6406 be consistent with the mode used in the patterns; for example to support
6407 the case of the add on the SPARC discussed above, we have the pattern
6413 (plus:SI (match_operand:SI 0 "register_operand" "%r")
6414 (match_operand:SI 1 "arith_operand" "rI"))
6421 together with a @code{SELECT_CC_MODE} that returns @code{CCNZmode}
6422 for comparisons whose argument is a @code{plus}:
6425 #define SELECT_CC_MODE(OP,X,Y) \
6426 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
6427 ? ((OP == LT || OP == LE || OP == GT || OP == GE) \
6428 ? CCFPEmode : CCFPmode) \
6429 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
6430 || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \
6431 ? CCNZmode : CCmode))
6434 Another reason to use modes is to retain information on which operands
6435 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
6438 You should define this macro if and only if you define extra CC modes
6439 in @file{@var{machine}-modes.def}.
6442 @deftypefn {Target Hook} void TARGET_CANONICALIZE_COMPARISON (int *@var{code}, rtx *@var{op0}, rtx *@var{op1}, bool @var{op0_preserve_value})
6443 On some machines not all possible comparisons are defined, but you can
6444 convert an invalid comparison into a valid one. For example, the Alpha
6445 does not have a @code{GT} comparison, but you can use an @code{LT}
6446 comparison instead and swap the order of the operands.
6448 On such machines, implement this hook to do any required conversions.
6449 @var{code} is the initial comparison code and @var{op0} and @var{op1}
6450 are the left and right operands of the comparison, respectively. If
6451 @var{op0_preserve_value} is @code{true} the implementation is not
6452 allowed to change the value of @var{op0} since the value might be used
6453 in RTXs which aren't comparisons. E.g. the implementation is not
6454 allowed to swap operands in that case.
6456 GCC will not assume that the comparison resulting from this macro is
6457 valid but will see if the resulting insn matches a pattern in the
6460 You need not to implement this hook if it would never change the
6461 comparison code or operands.
6464 @defmac REVERSIBLE_CC_MODE (@var{mode})
6465 A C expression whose value is one if it is always safe to reverse a
6466 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6467 can ever return @var{mode} for a floating-point inequality comparison,
6468 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6470 You need not define this macro if it would always returns zero or if the
6471 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6472 For example, here is the definition used on the SPARC, where floating-point
6473 inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}:
6476 #define REVERSIBLE_CC_MODE(MODE) \
6477 ((MODE) != CCFPEmode && (MODE) != CCFPmode)
6481 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6482 A C expression whose value is reversed condition code of the @var{code} for
6483 comparison done in CC_MODE @var{mode}. The macro is used only in case
6484 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6485 machine has some non-standard way how to reverse certain conditionals. For
6486 instance in case all floating point conditions are non-trapping, compiler may
6487 freely convert unordered compares to ordered ones. Then definition may look
6491 #define REVERSE_CONDITION(CODE, MODE) \
6492 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6493 : reverse_condition_maybe_unordered (CODE))
6497 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6498 On targets which do not use @code{(cc0)}, and which use a hard
6499 register rather than a pseudo-register to hold condition codes, the
6500 regular CSE passes are often not able to identify cases in which the
6501 hard register is set to a common value. Use this hook to enable a
6502 small pass which optimizes such cases. This hook should return true
6503 to enable this pass, and it should set the integers to which its
6504 arguments point to the hard register numbers used for condition codes.
6505 When there is only one such register, as is true on most systems, the
6506 integer pointed to by @var{p2} should be set to
6507 @code{INVALID_REGNUM}.
6509 The default version of this hook returns false.
6512 @deftypefn {Target Hook} machine_mode TARGET_CC_MODES_COMPATIBLE (machine_mode @var{m1}, machine_mode @var{m2})
6513 On targets which use multiple condition code modes in class
6514 @code{MODE_CC}, it is sometimes the case that a comparison can be
6515 validly done in more than one mode. On such a system, define this
6516 target hook to take two mode arguments and to return a mode in which
6517 both comparisons may be validly done. If there is no such mode,
6518 return @code{VOIDmode}.
6520 The default version of this hook checks whether the modes are the
6521 same. If they are, it returns that mode. If they are different, it
6522 returns @code{VOIDmode}.
6525 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
6526 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.
6530 @section Describing Relative Costs of Operations
6531 @cindex costs of instructions
6532 @cindex relative costs
6533 @cindex speed of instructions
6535 These macros let you describe the relative speed of various operations
6536 on the target machine.
6538 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6539 A C expression for the cost of moving data of mode @var{mode} from a
6540 register in class @var{from} to one in class @var{to}. The classes are
6541 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6542 value of 2 is the default; other values are interpreted relative to
6545 It is not required that the cost always equal 2 when @var{from} is the
6546 same as @var{to}; on some machines it is expensive to move between
6547 registers if they are not general registers.
6549 If reload sees an insn consisting of a single @code{set} between two
6550 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6551 classes returns a value of 2, reload does not check to ensure that the
6552 constraints of the insn are met. Setting a cost of other than 2 will
6553 allow reload to verify that the constraints are met. You should do this
6554 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6556 These macros are obsolete, new ports should use the target hook
6557 @code{TARGET_REGISTER_MOVE_COST} instead.
6560 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6561 This target hook should return the cost of moving data of mode @var{mode}
6562 from a register in class @var{from} to one in class @var{to}. The classes
6563 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6564 A value of 2 is the default; other values are interpreted relative to
6567 It is not required that the cost always equal 2 when @var{from} is the
6568 same as @var{to}; on some machines it is expensive to move between
6569 registers if they are not general registers.
6571 If reload sees an insn consisting of a single @code{set} between two
6572 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6573 classes returns a value of 2, reload does not check to ensure that the
6574 constraints of the insn are met. Setting a cost of other than 2 will
6575 allow reload to verify that the constraints are met. You should do this
6576 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6578 The default version of this function returns 2.
6581 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6582 A C expression for the cost of moving data of mode @var{mode} between a
6583 register of class @var{class} and memory; @var{in} is zero if the value
6584 is to be written to memory, nonzero if it is to be read in. This cost
6585 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6586 registers and memory is more expensive than between two registers, you
6587 should define this macro to express the relative cost.
6589 If you do not define this macro, GCC uses a default cost of 4 plus
6590 the cost of copying via a secondary reload register, if one is
6591 needed. If your machine requires a secondary reload register to copy
6592 between memory and a register of @var{class} but the reload mechanism is
6593 more complex than copying via an intermediate, define this macro to
6594 reflect the actual cost of the move.
6596 GCC defines the function @code{memory_move_secondary_cost} if
6597 secondary reloads are needed. It computes the costs due to copying via
6598 a secondary register. If your machine copies from memory using a
6599 secondary register in the conventional way but the default base value of
6600 4 is not correct for your machine, define this macro to add some other
6601 value to the result of that function. The arguments to that function
6602 are the same as to this macro.
6604 These macros are obsolete, new ports should use the target hook
6605 @code{TARGET_MEMORY_MOVE_COST} instead.
6608 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6609 This target hook should return the cost of moving data of mode @var{mode}
6610 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6611 if the value is to be written to memory, @code{true} if it is to be read in.
6612 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6613 If moving between registers and memory is more expensive than between two
6614 registers, you should add this target hook to express the relative cost.
6616 If you do not add this target hook, GCC uses a default cost of 4 plus
6617 the cost of copying via a secondary reload register, if one is
6618 needed. If your machine requires a secondary reload register to copy
6619 between memory and a register of @var{rclass} but the reload mechanism is
6620 more complex than copying via an intermediate, use this target hook to
6621 reflect the actual cost of the move.
6623 GCC defines the function @code{memory_move_secondary_cost} if
6624 secondary reloads are needed. It computes the costs due to copying via
6625 a secondary register. If your machine copies from memory using a
6626 secondary register in the conventional way but the default base value of
6627 4 is not correct for your machine, use this target hook to add some other
6628 value to the result of that function. The arguments to that function
6629 are the same as to this target hook.
6632 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6633 A C expression for the cost of a branch instruction. A value of 1 is
6634 the default; other values are interpreted relative to that. Parameter
6635 @var{speed_p} is true when the branch in question should be optimized
6636 for speed. When it is false, @code{BRANCH_COST} should return a value
6637 optimal for code size rather than performance. @var{predictable_p} is
6638 true for well-predicted branches. On many architectures the
6639 @code{BRANCH_COST} can be reduced then.
6642 Here are additional macros which do not specify precise relative costs,
6643 but only that certain actions are more expensive than GCC would
6646 @defmac SLOW_BYTE_ACCESS
6647 Define this macro as a C expression which is nonzero if accessing less
6648 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6649 faster than accessing a word of memory, i.e., if such access
6650 require more than one instruction or if there is no difference in cost
6651 between byte and (aligned) word loads.
6653 When this macro is not defined, the compiler will access a field by
6654 finding the smallest containing object; when it is defined, a fullword
6655 load will be used if alignment permits. Unless bytes accesses are
6656 faster than word accesses, using word accesses is preferable since it
6657 may eliminate subsequent memory access if subsequent accesses occur to
6658 other fields in the same word of the structure, but to different bytes.
6661 @deftypefn {Target Hook} bool TARGET_SLOW_UNALIGNED_ACCESS (machine_mode @var{mode}, unsigned int @var{align})
6662 This hook returns true if memory accesses described by the
6663 @var{mode} and @var{alignment} parameters have a cost many times greater
6664 than aligned accesses, for example if they are emulated in a trap handler.
6665 This hook is invoked only for unaligned accesses, i.e.@: when
6666 @code{@var{alignment} < GET_MODE_ALIGNMENT (@var{mode})}.
6668 When this hook returns true, the compiler will act as if
6669 @code{STRICT_ALIGNMENT} were true when generating code for block
6670 moves. This can cause significantly more instructions to be produced.
6671 Therefore, do not make this hook return true if unaligned accesses only
6672 add a cycle or two to the time for a memory access.
6674 The hook must return true whenever @code{STRICT_ALIGNMENT} is true.
6675 The default implementation returns @code{STRICT_ALIGNMENT}.
6678 @defmac MOVE_RATIO (@var{speed})
6679 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6680 which a sequence of insns should be generated instead of a
6681 string move insn or a library call. Increasing the value will always
6682 make code faster, but eventually incurs high cost in increased code size.
6684 Note that on machines where the corresponding move insn is a
6685 @code{define_expand} that emits a sequence of insns, this macro counts
6686 the number of such sequences.
6688 The parameter @var{speed} is true if the code is currently being
6689 optimized for speed rather than size.
6691 If you don't define this, a reasonable default is used.
6694 @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})
6695 GCC will attempt several strategies when asked to copy between
6696 two areas of memory, or to set, clear or store to memory, for example
6697 when copying a @code{struct}. The @code{by_pieces} infrastructure
6698 implements such memory operations as a sequence of load, store or move
6699 insns. Alternate strategies are to expand the
6700 @code{cpymem} or @code{setmem} optabs, to emit a library call, or to emit
6701 unit-by-unit, loop-based operations.
6703 This target hook should return true if, for a memory operation with a
6704 given @var{size} and @var{alignment}, using the @code{by_pieces}
6705 infrastructure is expected to result in better code generation.
6706 Both @var{size} and @var{alignment} are measured in terms of storage
6709 The parameter @var{op} is one of: @code{CLEAR_BY_PIECES},
6710 @code{MOVE_BY_PIECES}, @code{SET_BY_PIECES}, @code{STORE_BY_PIECES} or
6711 @code{COMPARE_BY_PIECES}. These describe the type of memory operation
6712 under consideration.
6714 The parameter @var{speed_p} is true if the code is currently being
6715 optimized for speed rather than size.
6717 Returning true for higher values of @var{size} can improve code generation
6718 for speed if the target does not provide an implementation of the
6719 @code{cpymem} or @code{setmem} standard names, if the @code{cpymem} or
6720 @code{setmem} implementation would be more expensive than a sequence of
6721 insns, or if the overhead of a library call would dominate that of
6722 the body of the memory operation.
6724 Returning true for higher values of @code{size} may also cause an increase
6725 in code size, for example where the number of insns emitted to perform a
6726 move would be greater than that of a library call.
6729 @deftypefn {Target Hook} int TARGET_COMPARE_BY_PIECES_BRANCH_RATIO (machine_mode @var{mode})
6730 When expanding a block comparison in MODE, gcc can try to reduce the
6731 number of branches at the expense of more memory operations. This hook
6732 allows the target to override the default choice. It should return the
6733 factor by which branches should be reduced over the plain expansion with
6734 one comparison per @var{mode}-sized piece. A port can also prevent a
6735 particular mode from being used for block comparisons by returning a
6736 negative number from this hook.
6739 @defmac MOVE_MAX_PIECES
6740 A C expression used by @code{move_by_pieces} to determine the largest unit
6741 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6744 @defmac STORE_MAX_PIECES
6745 A C expression used by @code{store_by_pieces} to determine the largest unit
6746 a store used to memory is. Defaults to @code{MOVE_MAX_PIECES}, or two times
6747 the size of @code{HOST_WIDE_INT}, whichever is smaller.
6750 @defmac COMPARE_MAX_PIECES
6751 A C expression used by @code{compare_by_pieces} to determine the largest unit
6752 a load or store used to compare memory is. Defaults to
6753 @code{MOVE_MAX_PIECES}.
6756 @defmac CLEAR_RATIO (@var{speed})
6757 The threshold of number of scalar move insns, @emph{below} which a sequence
6758 of insns should be generated to clear memory instead of a string clear insn
6759 or a library call. Increasing the value will always make code faster, but
6760 eventually incurs high cost in increased code size.
6762 The parameter @var{speed} is true if the code is currently being
6763 optimized for speed rather than size.
6765 If you don't define this, a reasonable default is used.
6768 @defmac SET_RATIO (@var{speed})
6769 The threshold of number of scalar move insns, @emph{below} which a sequence
6770 of insns should be generated to set memory to a constant value, instead of
6771 a block set insn or a library call.
6772 Increasing the value will always make code faster, but
6773 eventually incurs high cost in increased code size.
6775 The parameter @var{speed} is true if the code is currently being
6776 optimized for speed rather than size.
6778 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6781 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6782 A C expression used to determine whether a load postincrement is a good
6783 thing to use for a given mode. Defaults to the value of
6784 @code{HAVE_POST_INCREMENT}.
6787 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6788 A C expression used to determine whether a load postdecrement is a good
6789 thing to use for a given mode. Defaults to the value of
6790 @code{HAVE_POST_DECREMENT}.
6793 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6794 A C expression used to determine whether a load preincrement is a good
6795 thing to use for a given mode. Defaults to the value of
6796 @code{HAVE_PRE_INCREMENT}.
6799 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6800 A C expression used to determine whether a load predecrement is a good
6801 thing to use for a given mode. Defaults to the value of
6802 @code{HAVE_PRE_DECREMENT}.
6805 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6806 A C expression used to determine whether a store postincrement is a good
6807 thing to use for a given mode. Defaults to the value of
6808 @code{HAVE_POST_INCREMENT}.
6811 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6812 A C expression used to determine whether a store postdecrement is a good
6813 thing to use for a given mode. Defaults to the value of
6814 @code{HAVE_POST_DECREMENT}.
6817 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6818 This macro is used to determine whether a store preincrement is a good
6819 thing to use for a given mode. Defaults to the value of
6820 @code{HAVE_PRE_INCREMENT}.
6823 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6824 This macro is used to determine whether a store predecrement is a good
6825 thing to use for a given mode. Defaults to the value of
6826 @code{HAVE_PRE_DECREMENT}.
6829 @defmac NO_FUNCTION_CSE
6830 Define this macro to be true if it is as good or better to call a constant
6831 function address than to call an address kept in a register.
6834 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
6835 Define this macro if a non-short-circuit operation produced by
6836 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6837 @code{BRANCH_COST} is greater than or equal to the value 2.
6840 @deftypefn {Target Hook} bool TARGET_OPTAB_SUPPORTED_P (int @var{op}, machine_mode @var{mode1}, machine_mode @var{mode2}, optimization_type @var{opt_type})
6841 Return true if the optimizers should use optab @var{op} with
6842 modes @var{mode1} and @var{mode2} for optimization type @var{opt_type}.
6843 The optab is known to have an associated @file{.md} instruction
6844 whose C condition is true. @var{mode2} is only meaningful for conversion
6845 optabs; for direct optabs it is a copy of @var{mode1}.
6847 For example, when called with @var{op} equal to @code{rint_optab} and
6848 @var{mode1} equal to @code{DFmode}, the hook should say whether the
6849 optimizers should use optab @code{rintdf2}.
6851 The default hook returns true for all inputs.
6854 @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})
6855 This target hook describes the relative costs of RTL expressions.
6857 The cost may depend on the precise form of the expression, which is
6858 available for examination in @var{x}, and the fact that @var{x} appears
6859 as operand @var{opno} of an expression with rtx code @var{outer_code}.
6860 That is, the hook can assume that there is some rtx @var{y} such
6861 that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that
6862 either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or
6863 (b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}.
6865 @var{mode} is @var{x}'s machine mode, or for cases like @code{const_int} that
6866 do not have a mode, the mode in which @var{x} is used.
6868 In implementing this hook, you can use the construct
6869 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6872 On entry to the hook, @code{*@var{total}} contains a default estimate
6873 for the cost of the expression. The hook should modify this value as
6874 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6875 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6876 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6878 When optimizing for code size, i.e.@: when @code{speed} is
6879 false, this target hook should be used to estimate the relative
6880 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6882 The hook returns true when all subexpressions of @var{x} have been
6883 processed, and false when @code{rtx_cost} should recurse.
6886 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, machine_mode @var{mode}, addr_space_t @var{as}, bool @var{speed})
6887 This hook computes the cost of an addressing mode that contains
6888 @var{address}. If not defined, the cost is computed from
6889 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6891 For most CISC machines, the default cost is a good approximation of the
6892 true cost of the addressing mode. However, on RISC machines, all
6893 instructions normally have the same length and execution time. Hence
6894 all addresses will have equal costs.
6896 In cases where more than one form of an address is known, the form with
6897 the lowest cost will be used. If multiple forms have the same, lowest,
6898 cost, the one that is the most complex will be used.
6900 For example, suppose an address that is equal to the sum of a register
6901 and a constant is used twice in the same basic block. When this macro
6902 is not defined, the address will be computed in a register and memory
6903 references will be indirect through that register. On machines where
6904 the cost of the addressing mode containing the sum is no higher than
6905 that of a simple indirect reference, this will produce an additional
6906 instruction and possibly require an additional register. Proper
6907 specification of this macro eliminates this overhead for such machines.
6909 This hook is never called with an invalid address.
6911 On machines where an address involving more than one register is as
6912 cheap as an address computation involving only one register, defining
6913 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6914 be live over a region of code where only one would have been if
6915 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6916 should be considered in the definition of this macro. Equivalent costs
6917 should probably only be given to addresses with different numbers of
6918 registers on machines with lots of registers.
6921 @deftypefn {Target Hook} int TARGET_INSN_COST (rtx_insn *@var{insn}, bool @var{speed})
6922 This target hook describes the relative costs of RTL instructions.
6924 In implementing this hook, you can use the construct
6925 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6928 When optimizing for code size, i.e.@: when @code{speed} is
6929 false, this target hook should be used to estimate the relative
6930 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6933 @deftypefn {Target Hook} {unsigned int} TARGET_MAX_NOCE_IFCVT_SEQ_COST (edge @var{e})
6934 This hook returns a value in the same units as @code{TARGET_RTX_COSTS},
6935 giving the maximum acceptable cost for a sequence generated by the RTL
6936 if-conversion pass when conditional execution is not available.
6937 The RTL if-conversion pass attempts to convert conditional operations
6938 that would require a branch to a series of unconditional operations and
6939 @code{mov@var{mode}cc} insns. This hook returns the maximum cost of the
6940 unconditional instructions and the @code{mov@var{mode}cc} insns.
6941 RTL if-conversion is cancelled if the cost of the converted sequence
6942 is greater than the value returned by this hook.
6944 @code{e} is the edge between the basic block containing the conditional
6945 branch to the basic block which would be executed if the condition
6948 The default implementation of this hook uses the
6949 @code{max-rtl-if-conversion-[un]predictable} parameters if they are set,
6950 and uses a multiple of @code{BRANCH_COST} otherwise.
6953 @deftypefn {Target Hook} bool TARGET_NOCE_CONVERSION_PROFITABLE_P (rtx_insn *@var{seq}, struct noce_if_info *@var{if_info})
6954 This hook returns true if the instruction sequence @code{seq} is a good
6955 candidate as a replacement for the if-convertible sequence described in
6959 @deftypefn {Target Hook} bool TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P (void)
6960 This predicate controls the use of the eager delay slot filler to disallow
6961 speculatively executed instructions being placed in delay slots. Targets
6962 such as certain MIPS architectures possess both branches with and without
6963 delay slots. As the eager delay slot filler can decrease performance,
6964 disabling it is beneficial when ordinary branches are available. Use of
6965 delay slot branches filled using the basic filler is often still desirable
6966 as the delay slot can hide a pipeline bubble.
6969 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_ESTIMATED_POLY_VALUE (poly_int64 @var{val})
6970 Return an estimate of the runtime value of @var{val}, for use in
6971 things like cost calculations or profiling frequencies. The default
6972 implementation returns the lowest possible value of @var{val}.
6976 @section Adjusting the Instruction Scheduler
6978 The instruction scheduler may need a fair amount of machine-specific
6979 adjustment in order to produce good code. GCC provides several target
6980 hooks for this purpose. It is usually enough to define just a few of
6981 them: try the first ones in this list first.
6983 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6984 This hook returns the maximum number of instructions that can ever
6985 issue at the same time on the target machine. The default is one.
6986 Although the insn scheduler can define itself the possibility of issue
6987 an insn on the same cycle, the value can serve as an additional
6988 constraint to issue insns on the same simulated processor cycle (see
6989 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6990 This value must be constant over the entire compilation. If you need
6991 it to vary depending on what the instructions are, you must use
6992 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6995 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx_insn *@var{insn}, int @var{more})
6996 This hook is executed by the scheduler after it has scheduled an insn
6997 from the ready list. It should return the number of insns which can
6998 still be issued in the current cycle. The default is
6999 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
7000 @code{USE}, which normally are not counted against the issue rate.
7001 You should define this hook if some insns take more machine resources
7002 than others, so that fewer insns can follow them in the same cycle.
7003 @var{file} is either a null pointer, or a stdio stream to write any
7004 debug output to. @var{verbose} is the verbose level provided by
7005 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
7009 @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})
7010 This function corrects the value of @var{cost} based on the
7011 relationship between @var{insn} and @var{dep_insn} through a
7012 dependence of type dep_type, and strength @var{dw}. It should return the new
7013 value. The default is to make no adjustment to @var{cost}. This can be
7014 used for example to specify to the scheduler using the traditional pipeline
7015 description that an output- or anti-dependence does not incur the same cost
7016 as a data-dependence. If the scheduler using the automaton based pipeline
7017 description, the cost of anti-dependence is zero and the cost of
7018 output-dependence is maximum of one and the difference of latency
7019 times of the first and the second insns. If these values are not
7020 acceptable, you could use the hook to modify them too. See also
7021 @pxref{Processor pipeline description}.
7024 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx_insn *@var{insn}, int @var{priority})
7025 This hook adjusts the integer scheduling priority @var{priority} of
7026 @var{insn}. It should return the new priority. Increase the priority to
7027 execute @var{insn} earlier, reduce the priority to execute @var{insn}
7028 later. Do not define this hook if you do not need to adjust the
7029 scheduling priorities of insns.
7032 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock})
7033 This hook is executed by the scheduler after it has scheduled the ready
7034 list, to allow the machine description to reorder it (for example to
7035 combine two small instructions together on @samp{VLIW} machines).
7036 @var{file} is either a null pointer, or a stdio stream to write any
7037 debug output to. @var{verbose} is the verbose level provided by
7038 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
7039 list of instructions that are ready to be scheduled. @var{n_readyp} is
7040 a pointer to the number of elements in the ready list. The scheduler
7041 reads the ready list in reverse order, starting with
7042 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
7043 is the timer tick of the scheduler. You may modify the ready list and
7044 the number of ready insns. The return value is the number of insns that
7045 can issue this cycle; normally this is just @code{issue_rate}. See also
7046 @samp{TARGET_SCHED_REORDER2}.
7049 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock})
7050 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
7051 function is called whenever the scheduler starts a new cycle. This one
7052 is called once per iteration over a cycle, immediately after
7053 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
7054 return the number of insns to be scheduled in the same cycle. Defining
7055 this hook can be useful if there are frequent situations where
7056 scheduling one insn causes other insns to become ready in the same
7057 cycle. These other insns can then be taken into account properly.
7060 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_P (void)
7061 This hook is used to check whether target platform supports macro fusion.
7064 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_PAIR_P (rtx_insn *@var{prev}, rtx_insn *@var{curr})
7065 This hook is used to check whether two insns should be macro fused for
7066 a target microarchitecture. If this hook returns true for the given insn pair
7067 (@var{prev} and @var{curr}), the scheduler will put them into a sched
7068 group, and they will not be scheduled apart. The two insns will be either
7069 two SET insns or a compare and a conditional jump and this hook should
7070 validate any dependencies needed to fuse the two insns together.
7073 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx_insn *@var{head}, rtx_insn *@var{tail})
7074 This hook is called after evaluation forward dependencies of insns in
7075 chain given by two parameter values (@var{head} and @var{tail}
7076 correspondingly) but before insns scheduling of the insn chain. For
7077 example, it can be used for better insn classification if it requires
7078 analysis of dependencies. This hook can use backward and forward
7079 dependencies of the insn scheduler because they are already
7083 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
7084 This hook is executed by the scheduler at the beginning of each block of
7085 instructions that are to be scheduled. @var{file} is either a null
7086 pointer, or a stdio stream to write any debug output to. @var{verbose}
7087 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
7088 @var{max_ready} is the maximum number of insns in the current scheduling
7089 region that can be live at the same time. This can be used to allocate
7090 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
7093 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
7094 This hook is executed by the scheduler at the end of each block of
7095 instructions that are to be scheduled. It can be used to perform
7096 cleanup of any actions done by the other scheduling hooks. @var{file}
7097 is either a null pointer, or a stdio stream to write any debug output
7098 to. @var{verbose} is the verbose level provided by
7099 @option{-fsched-verbose-@var{n}}.
7102 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
7103 This hook is executed by the scheduler after function level initializations.
7104 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
7105 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
7106 @var{old_max_uid} is the maximum insn uid when scheduling begins.
7109 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
7110 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
7111 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
7112 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
7115 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
7116 The hook returns an RTL insn. The automaton state used in the
7117 pipeline hazard recognizer is changed as if the insn were scheduled
7118 when the new simulated processor cycle starts. Usage of the hook may
7119 simplify the automaton pipeline description for some @acronym{VLIW}
7120 processors. If the hook is defined, it is used only for the automaton
7121 based pipeline description. The default is not to change the state
7122 when the new simulated processor cycle starts.
7125 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
7126 The hook can be used to initialize data used by the previous hook.
7129 @deftypefn {Target Hook} {rtx_insn *} TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
7130 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
7131 to changed the state as if the insn were scheduled when the new
7132 simulated processor cycle finishes.
7135 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
7136 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
7137 used to initialize data used by the previous hook.
7140 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
7141 The hook to notify target that the current simulated cycle is about to finish.
7142 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
7143 to change the state in more complicated situations - e.g., when advancing
7144 state on a single insn is not enough.
7147 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
7148 The hook to notify target that new simulated cycle has just started.
7149 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
7150 to change the state in more complicated situations - e.g., when advancing
7151 state on a single insn is not enough.
7154 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
7155 This hook controls better choosing an insn from the ready insn queue
7156 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
7157 chooses the first insn from the queue. If the hook returns a positive
7158 value, an additional scheduler code tries all permutations of
7159 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
7160 subsequent ready insns to choose an insn whose issue will result in
7161 maximal number of issued insns on the same cycle. For the
7162 @acronym{VLIW} processor, the code could actually solve the problem of
7163 packing simple insns into the @acronym{VLIW} insn. Of course, if the
7164 rules of @acronym{VLIW} packing are described in the automaton.
7166 This code also could be used for superscalar @acronym{RISC}
7167 processors. Let us consider a superscalar @acronym{RISC} processor
7168 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
7169 @var{B}, some insns can be executed only in pipelines @var{B} or
7170 @var{C}, and one insn can be executed in pipeline @var{B}. The
7171 processor may issue the 1st insn into @var{A} and the 2nd one into
7172 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
7173 until the next cycle. If the scheduler issues the 3rd insn the first,
7174 the processor could issue all 3 insns per cycle.
7176 Actually this code demonstrates advantages of the automaton based
7177 pipeline hazard recognizer. We try quickly and easy many insn
7178 schedules to choose the best one.
7180 The default is no multipass scheduling.
7183 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx_insn *@var{insn}, int @var{ready_index})
7185 This hook controls what insns from the ready insn queue will be
7186 considered for the multipass insn scheduling. If the hook returns
7187 zero for @var{insn}, the insn will be considered in multipass scheduling.
7188 Positive return values will remove @var{insn} from consideration on
7189 the current round of multipass scheduling.
7190 Negative return values will remove @var{insn} from consideration for given
7192 Backends should be careful about returning non-zero for highest priority
7193 instruction at position 0 in the ready list. @var{ready_index} is passed
7194 to allow backends make correct judgements.
7196 The default is that any ready insns can be chosen to be issued.
7199 @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})
7200 This hook prepares the target backend for a new round of multipass
7204 @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})
7205 This hook is called when multipass scheduling evaluates instruction INSN.
7208 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, signed char *@var{ready_try}, int @var{n_ready})
7209 This is called when multipass scheduling backtracks from evaluation of
7213 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
7214 This hook notifies the target about the result of the concluded current
7215 round of multipass scheduling.
7218 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
7219 This hook initializes target-specific data used in multipass scheduling.
7222 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
7223 This hook finalizes target-specific data used in multipass scheduling.
7226 @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})
7227 This hook is called by the insn scheduler before issuing @var{insn}
7228 on cycle @var{clock}. If the hook returns nonzero,
7229 @var{insn} is not issued on this processor cycle. Instead,
7230 the processor cycle is advanced. If *@var{sort_p}
7231 is zero, the insn ready queue is not sorted on the new cycle
7232 start as usually. @var{dump} and @var{verbose} specify the file and
7233 verbosity level to use for debugging output.
7234 @var{last_clock} and @var{clock} are, respectively, the
7235 processor cycle on which the previous insn has been issued,
7236 and the current processor cycle.
7239 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
7240 This hook is used to define which dependences are considered costly by
7241 the target, so costly that it is not advisable to schedule the insns that
7242 are involved in the dependence too close to one another. The parameters
7243 to this hook are as follows: The first parameter @var{_dep} is the dependence
7244 being evaluated. The second parameter @var{cost} is the cost of the
7245 dependence as estimated by the scheduler, and the third
7246 parameter @var{distance} is the distance in cycles between the two insns.
7247 The hook returns @code{true} if considering the distance between the two
7248 insns the dependence between them is considered costly by the target,
7249 and @code{false} otherwise.
7251 Defining this hook can be useful in multiple-issue out-of-order machines,
7252 where (a) it's practically hopeless to predict the actual data/resource
7253 delays, however: (b) there's a better chance to predict the actual grouping
7254 that will be formed, and (c) correctly emulating the grouping can be very
7255 important. In such targets one may want to allow issuing dependent insns
7256 closer to one another---i.e., closer than the dependence distance; however,
7257 not in cases of ``costly dependences'', which this hooks allows to define.
7260 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
7261 This hook is called by the insn scheduler after emitting a new instruction to
7262 the instruction stream. The hook notifies a target backend to extend its
7263 per instruction data structures.
7266 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
7267 Return a pointer to a store large enough to hold target scheduling context.
7270 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
7271 Initialize store pointed to by @var{tc} to hold target scheduling context.
7272 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
7273 beginning of the block. Otherwise, copy the current context into @var{tc}.
7276 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
7277 Copy target scheduling context pointed to by @var{tc} to the current context.
7280 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
7281 Deallocate internal data in target scheduling context pointed to by @var{tc}.
7284 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
7285 Deallocate a store for target scheduling context pointed to by @var{tc}.
7288 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx_insn *@var{insn}, unsigned int @var{dep_status}, rtx *@var{new_pat})
7289 This hook is called by the insn scheduler when @var{insn} has only
7290 speculative dependencies and therefore can be scheduled speculatively.
7291 The hook is used to check if the pattern of @var{insn} has a speculative
7292 version and, in case of successful check, to generate that speculative
7293 pattern. The hook should return 1, if the instruction has a speculative form,
7294 or @minus{}1, if it doesn't. @var{request} describes the type of requested
7295 speculation. If the return value equals 1 then @var{new_pat} is assigned
7296 the generated speculative pattern.
7299 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (unsigned int @var{dep_status})
7300 This hook is called by the insn scheduler during generation of recovery code
7301 for @var{insn}. It should return @code{true}, if the corresponding check
7302 instruction should branch to recovery code, or @code{false} otherwise.
7305 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx_insn *@var{insn}, rtx_insn *@var{label}, unsigned int @var{ds})
7306 This hook is called by the insn scheduler to generate a pattern for recovery
7307 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
7308 speculative instruction for which the check should be generated.
7309 @var{label} is either a label of a basic block, where recovery code should
7310 be emitted, or a null pointer, when requested check doesn't branch to
7311 recovery code (a simple check). If @var{mutate_p} is nonzero, then
7312 a pattern for a branchy check corresponding to a simple check denoted by
7313 @var{insn} should be generated. In this case @var{label} can't be null.
7316 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
7317 This hook is used by the insn scheduler to find out what features should be
7319 The structure *@var{spec_info} should be filled in by the target.
7320 The structure describes speculation types that can be used in the scheduler.
7323 @deftypefn {Target Hook} bool TARGET_SCHED_CAN_SPECULATE_INSN (rtx_insn *@var{insn})
7324 Some instructions should never be speculated by the schedulers, usually
7325 because the instruction is too expensive to get this wrong. Often such
7326 instructions have long latency, and often they are not fully modeled in the
7327 pipeline descriptions. This hook should return @code{false} if @var{insn}
7328 should not be speculated.
7331 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
7332 This hook is called by the swing modulo scheduler to calculate a
7333 resource-based lower bound which is based on the resources available in
7334 the machine and the resources required by each instruction. The target
7335 backend can use @var{g} to calculate such bound. A very simple lower
7336 bound will be used in case this hook is not implemented: the total number
7337 of instructions divided by the issue rate.
7340 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx_insn *@var{insn}, int @var{x})
7341 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
7342 is supported in hardware and the condition specified in the parameter is true.
7345 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx_insn *@var{insn}, int @var{x})
7346 This hook is called by Haifa Scheduler. It performs the operation specified
7347 in its second parameter.
7350 @deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE
7351 True if the processor has an exposed pipeline, which means that not just
7352 the order of instructions is important for correctness when scheduling, but
7353 also the latencies of operations.
7356 @deftypefn {Target Hook} int TARGET_SCHED_REASSOCIATION_WIDTH (unsigned int @var{opc}, machine_mode @var{mode})
7357 This hook is called by tree reassociator to determine a level of
7358 parallelism required in output calculations chain.
7361 @deftypefn {Target Hook} void TARGET_SCHED_FUSION_PRIORITY (rtx_insn *@var{insn}, int @var{max_pri}, int *@var{fusion_pri}, int *@var{pri})
7362 This hook is called by scheduling fusion pass. It calculates fusion
7363 priorities for each instruction passed in by parameter. The priorities
7364 are returned via pointer parameters.
7366 @var{insn} is the instruction whose priorities need to be calculated.
7367 @var{max_pri} is the maximum priority can be returned in any cases.
7368 @var{fusion_pri} is the pointer parameter through which @var{insn}'s
7369 fusion priority should be calculated and returned.
7370 @var{pri} is the pointer parameter through which @var{insn}'s priority
7371 should be calculated and returned.
7373 Same @var{fusion_pri} should be returned for instructions which should
7374 be scheduled together. Different @var{pri} should be returned for
7375 instructions with same @var{fusion_pri}. @var{fusion_pri} is the major
7376 sort key, @var{pri} is the minor sort key. All instructions will be
7377 scheduled according to the two priorities. All priorities calculated
7378 should be between 0 (exclusive) and @var{max_pri} (inclusive). To avoid
7379 false dependencies, @var{fusion_pri} of instructions which need to be
7380 scheduled together should be smaller than @var{fusion_pri} of irrelevant
7383 Given below example:
7396 On targets like ARM/AArch64, the two pairs of consecutive loads should be
7397 merged. Since peephole2 pass can't help in this case unless consecutive
7398 loads are actually next to each other in instruction flow. That's where
7399 this scheduling fusion pass works. This hook calculates priority for each
7400 instruction based on its fustion type, like:
7403 ldr r10, [r1, 4] ; fusion_pri=99, pri=96
7404 add r4, r4, r10 ; fusion_pri=100, pri=100
7405 ldr r15, [r2, 8] ; fusion_pri=98, pri=92
7406 sub r5, r5, r15 ; fusion_pri=100, pri=100
7407 ldr r11, [r1, 0] ; fusion_pri=99, pri=100
7408 add r4, r4, r11 ; fusion_pri=100, pri=100
7409 ldr r16, [r2, 12] ; fusion_pri=98, pri=88
7410 sub r5, r5, r16 ; fusion_pri=100, pri=100
7413 Scheduling fusion pass then sorts all ready to issue instructions according
7414 to the priorities. As a result, instructions of same fusion type will be
7415 pushed together in instruction flow, like:
7428 Now peephole2 pass can simply merge the two pairs of loads.
7430 Since scheduling fusion pass relies on peephole2 to do real fusion
7431 work, it is only enabled by default when peephole2 is in effect.
7433 This is firstly introduced on ARM/AArch64 targets, please refer to
7434 the hook implementation for how different fusion types are supported.
7437 @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})
7438 Define this hook for enabling divmod transform if the port does not have
7439 hardware divmod insn but defines target-specific divmod libfuncs.
7443 @section Dividing the Output into Sections (Texts, Data, @dots{})
7444 @c the above section title is WAY too long. maybe cut the part between
7445 @c the (...)? --mew 10feb93
7447 An object file is divided into sections containing different types of
7448 data. In the most common case, there are three sections: the @dfn{text
7449 section}, which holds instructions and read-only data; the @dfn{data
7450 section}, which holds initialized writable data; and the @dfn{bss
7451 section}, which holds uninitialized data. Some systems have other kinds
7454 @file{varasm.c} provides several well-known sections, such as
7455 @code{text_section}, @code{data_section} and @code{bss_section}.
7456 The normal way of controlling a @code{@var{foo}_section} variable
7457 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
7458 as described below. The macros are only read once, when @file{varasm.c}
7459 initializes itself, so their values must be run-time constants.
7460 They may however depend on command-line flags.
7462 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
7463 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
7464 to be string literals.
7466 Some assemblers require a different string to be written every time a
7467 section is selected. If your assembler falls into this category, you
7468 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
7469 @code{get_unnamed_section} to set up the sections.
7471 You must always create a @code{text_section}, either by defining
7472 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
7473 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
7474 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
7475 create a distinct @code{readonly_data_section}, the default is to
7476 reuse @code{text_section}.
7478 All the other @file{varasm.c} sections are optional, and are null
7479 if the target does not provide them.
7481 @defmac TEXT_SECTION_ASM_OP
7482 A C expression whose value is a string, including spacing, containing the
7483 assembler operation that should precede instructions and read-only data.
7484 Normally @code{"\t.text"} is right.
7487 @defmac HOT_TEXT_SECTION_NAME
7488 If defined, a C string constant for the name of the section containing most
7489 frequently executed functions of the program. If not defined, GCC will provide
7490 a default definition if the target supports named sections.
7493 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
7494 If defined, a C string constant for the name of the section containing unlikely
7495 executed functions in the program.
7498 @defmac DATA_SECTION_ASM_OP
7499 A C expression whose value is a string, including spacing, containing the
7500 assembler operation to identify the following data as writable initialized
7501 data. Normally @code{"\t.data"} is right.
7504 @defmac SDATA_SECTION_ASM_OP
7505 If defined, a C expression whose value is a string, including spacing,
7506 containing the assembler operation to identify the following data as
7507 initialized, writable small data.
7510 @defmac READONLY_DATA_SECTION_ASM_OP
7511 A C expression whose value is a string, including spacing, containing the
7512 assembler operation to identify the following data as read-only initialized
7516 @defmac BSS_SECTION_ASM_OP
7517 If defined, a C expression whose value is a string, including spacing,
7518 containing the assembler operation to identify the following data as
7519 uninitialized global data. If not defined, and
7520 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
7521 uninitialized global data will be output in the data section if
7522 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
7526 @defmac SBSS_SECTION_ASM_OP
7527 If defined, a C expression whose value is a string, including spacing,
7528 containing the assembler operation to identify the following data as
7529 uninitialized, writable small data.
7532 @defmac TLS_COMMON_ASM_OP
7533 If defined, a C expression whose value is a string containing the
7534 assembler operation to identify the following data as thread-local
7535 common data. The default is @code{".tls_common"}.
7538 @defmac TLS_SECTION_ASM_FLAG
7539 If defined, a C expression whose value is a character constant
7540 containing the flag used to mark a section as a TLS section. The
7541 default is @code{'T'}.
7544 @defmac INIT_SECTION_ASM_OP
7545 If defined, a C expression whose value is a string, including spacing,
7546 containing the assembler operation to identify the following data as
7547 initialization code. If not defined, GCC will assume such a section does
7548 not exist. This section has no corresponding @code{init_section}
7549 variable; it is used entirely in runtime code.
7552 @defmac FINI_SECTION_ASM_OP
7553 If defined, a C expression whose value is a string, including spacing,
7554 containing the assembler operation to identify the following data as
7555 finalization code. If not defined, GCC will assume such a section does
7556 not exist. This section has no corresponding @code{fini_section}
7557 variable; it is used entirely in runtime code.
7560 @defmac INIT_ARRAY_SECTION_ASM_OP
7561 If defined, a C expression whose value is a string, including spacing,
7562 containing the assembler operation to identify the following data as
7563 part of the @code{.init_array} (or equivalent) section. If not
7564 defined, GCC will assume such a section does not exist. Do not define
7565 both this macro and @code{INIT_SECTION_ASM_OP}.
7568 @defmac FINI_ARRAY_SECTION_ASM_OP
7569 If defined, a C expression whose value is a string, including spacing,
7570 containing the assembler operation to identify the following data as
7571 part of the @code{.fini_array} (or equivalent) section. If not
7572 defined, GCC will assume such a section does not exist. Do not define
7573 both this macro and @code{FINI_SECTION_ASM_OP}.
7576 @defmac MACH_DEP_SECTION_ASM_FLAG
7577 If defined, a C expression whose value is a character constant
7578 containing the flag used to mark a machine-dependent section. This
7579 corresponds to the @code{SECTION_MACH_DEP} section flag.
7582 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
7583 If defined, an ASM statement that switches to a different section
7584 via @var{section_op}, calls @var{function}, and switches back to
7585 the text section. This is used in @file{crtstuff.c} if
7586 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
7587 to initialization and finalization functions from the init and fini
7588 sections. By default, this macro uses a simple function call. Some
7589 ports need hand-crafted assembly code to avoid dependencies on
7590 registers initialized in the function prologue or to ensure that
7591 constant pools don't end up too far way in the text section.
7594 @defmac TARGET_LIBGCC_SDATA_SECTION
7595 If defined, a string which names the section into which small
7596 variables defined in crtstuff and libgcc should go. This is useful
7597 when the target has options for optimizing access to small data, and
7598 you want the crtstuff and libgcc routines to be conservative in what
7599 they expect of your application yet liberal in what your application
7600 expects. For example, for targets with a @code{.sdata} section (like
7601 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
7602 require small data support from your application, but use this macro
7603 to put small data into @code{.sdata} so that your application can
7604 access these variables whether it uses small data or not.
7607 @defmac FORCE_CODE_SECTION_ALIGN
7608 If defined, an ASM statement that aligns a code section to some
7609 arbitrary boundary. This is used to force all fragments of the
7610 @code{.init} and @code{.fini} sections to have to same alignment
7611 and thus prevent the linker from having to add any padding.
7614 @defmac JUMP_TABLES_IN_TEXT_SECTION
7615 Define this macro to be an expression with a nonzero value if jump
7616 tables (for @code{tablejump} insns) should be output in the text
7617 section, along with the assembler instructions. Otherwise, the
7618 readonly data section is used.
7620 This macro is irrelevant if there is no separate readonly data section.
7623 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
7624 Define this hook if you need to do something special to set up the
7625 @file{varasm.c} sections, or if your target has some special sections
7626 of its own that you need to create.
7628 GCC calls this hook after processing the command line, but before writing
7629 any assembly code, and before calling any of the section-returning hooks
7633 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
7634 Return a mask describing how relocations should be treated when
7635 selecting sections. Bit 1 should be set if global relocations
7636 should be placed in a read-write section; bit 0 should be set if
7637 local relocations should be placed in a read-write section.
7639 The default version of this function returns 3 when @option{-fpic}
7640 is in effect, and 0 otherwise. The hook is typically redefined
7641 when the target cannot support (some kinds of) dynamic relocations
7642 in read-only sections even in executables.
7645 @deftypefn {Target Hook} bool TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC (void)
7646 Return true to generate ADDR_DIF_VEC table
7647 or false to generate ADDR_VEC table for jumps in case of -fPIC.
7649 The default version of this function returns true if flag_pic
7650 equals true and false otherwise
7653 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
7654 Return the section into which @var{exp} should be placed. You can
7655 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
7656 some sort. @var{reloc} indicates whether the initial value of @var{exp}
7657 requires link-time relocations. Bit 0 is set when variable contains
7658 local relocations only, while bit 1 is set for global relocations.
7659 @var{align} is the constant alignment in bits.
7661 The default version of this function takes care of putting read-only
7662 variables in @code{readonly_data_section}.
7664 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7667 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7668 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7669 for @code{FUNCTION_DECL}s as well as for variables and constants.
7671 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7672 function has been determined to be likely to be called, and nonzero if
7673 it is unlikely to be called.
7676 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7677 Build up a unique section name, expressed as a @code{STRING_CST} node,
7678 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7679 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7680 the initial value of @var{exp} requires link-time relocations.
7682 The default version of this function appends the symbol name to the
7683 ELF section name that would normally be used for the symbol. For
7684 example, the function @code{foo} would be placed in @code{.text.foo}.
7685 Whatever the actual target object format, this is often good enough.
7688 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7689 Return the readonly data section associated with
7690 @samp{DECL_SECTION_NAME (@var{decl})}.
7691 The default version of this function selects @code{.gnu.linkonce.r.name} if
7692 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7693 if function is in @code{.text.name}, and the normal readonly-data section
7697 @deftypevr {Target Hook} {const char *} TARGET_ASM_MERGEABLE_RODATA_PREFIX
7698 Usually, the compiler uses the prefix @code{".rodata"} to construct
7699 section names for mergeable constant data. Define this macro to override
7700 the string if a different section name should be used.
7703 @deftypefn {Target Hook} {section *} TARGET_ASM_TM_CLONE_TABLE_SECTION (void)
7704 Return the section that should be used for transactional memory clone tables.
7707 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7708 Return the section into which a constant @var{x}, of mode @var{mode},
7709 should be placed. You can assume that @var{x} is some kind of
7710 constant in RTL@. The argument @var{mode} is redundant except in the
7711 case of a @code{const_int} rtx. @var{align} is the constant alignment
7714 The default version of this function takes care of putting symbolic
7715 constants in @code{flag_pic} mode in @code{data_section} and everything
7716 else in @code{readonly_data_section}.
7719 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7720 Define this hook if you need to postprocess the assembler name generated
7721 by target-independent code. The @var{id} provided to this hook will be
7722 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7723 or the mangled name of the @var{decl} in C++). The return value of the
7724 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7725 your target system. The default implementation of this hook just
7726 returns the @var{id} provided.
7729 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7730 Define this hook if references to a symbol or a constant must be
7731 treated differently depending on something about the variable or
7732 function named by the symbol (such as what section it is in).
7734 The hook is executed immediately after rtl has been created for
7735 @var{decl}, which may be a variable or function declaration or
7736 an entry in the constant pool. In either case, @var{rtl} is the
7737 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7738 in this hook; that field may not have been initialized yet.
7740 In the case of a constant, it is safe to assume that the rtl is
7741 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7742 will also have this form, but that is not guaranteed. Global
7743 register variables, for instance, will have a @code{reg} for their
7744 rtl. (Normally the right thing to do with such unusual rtl is
7747 The @var{new_decl_p} argument will be true if this is the first time
7748 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7749 be false for subsequent invocations, which will happen for duplicate
7750 declarations. Whether or not anything must be done for the duplicate
7751 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7752 @var{new_decl_p} is always true when the hook is called for a constant.
7754 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7755 The usual thing for this hook to do is to record flags in the
7756 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7757 Historically, the name string was modified if it was necessary to
7758 encode more than one bit of information, but this practice is now
7759 discouraged; use @code{SYMBOL_REF_FLAGS}.
7761 The default definition of this hook, @code{default_encode_section_info}
7762 in @file{varasm.c}, sets a number of commonly-useful bits in
7763 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7764 before overriding it.
7767 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7768 Decode @var{name} and return the real name part, sans
7769 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7773 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7774 Returns true if @var{exp} should be placed into a ``small data'' section.
7775 The default version of this hook always returns false.
7778 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7779 Contains the value true if the target places read-only
7780 ``small data'' into a separate section. The default value is false.
7783 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7784 It returns true if target wants profile code emitted before prologue.
7786 The default version of this hook use the target macro
7787 @code{PROFILE_BEFORE_PROLOGUE}.
7790 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7791 Returns true if @var{exp} names an object for which name resolution
7792 rules must resolve to the current ``module'' (dynamic shared library
7793 or executable image).
7795 The default version of this hook implements the name resolution rules
7796 for ELF, which has a looser model of global name binding than other
7797 currently supported object file formats.
7800 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7801 Contains the value true if the target supports thread-local storage.
7802 The default value is false.
7807 @section Position Independent Code
7808 @cindex position independent code
7811 This section describes macros that help implement generation of position
7812 independent code. Simply defining these macros is not enough to
7813 generate valid PIC; you must also add support to the hook
7814 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7815 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7816 must modify the definition of @samp{movsi} to do something appropriate
7817 when the source operand contains a symbolic address. You may also
7818 need to alter the handling of switch statements so that they use
7820 @c i rearranged the order of the macros above to try to force one of
7821 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7823 @defmac PIC_OFFSET_TABLE_REGNUM
7824 The register number of the register used to address a table of static
7825 data addresses in memory. In some cases this register is defined by a
7826 processor's ``application binary interface'' (ABI)@. When this macro
7827 is defined, RTL is generated for this register once, as with the stack
7828 pointer and frame pointer registers. If this macro is not defined, it
7829 is up to the machine-dependent files to allocate such a register (if
7830 necessary). Note that this register must be fixed when in use (e.g.@:
7831 when @code{flag_pic} is true).
7834 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7835 A C expression that is nonzero if the register defined by
7836 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7837 the default is zero. Do not define
7838 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7841 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7842 A C expression that is nonzero if @var{x} is a legitimate immediate
7843 operand on the target machine when generating position independent code.
7844 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7845 check this. You can also assume @var{flag_pic} is true, so you need not
7846 check it either. You need not define this macro if all constants
7847 (including @code{SYMBOL_REF}) can be immediate operands when generating
7848 position independent code.
7851 @node Assembler Format
7852 @section Defining the Output Assembler Language
7854 This section describes macros whose principal purpose is to describe how
7855 to write instructions in assembler language---rather than what the
7859 * File Framework:: Structural information for the assembler file.
7860 * Data Output:: Output of constants (numbers, strings, addresses).
7861 * Uninitialized Data:: Output of uninitialized variables.
7862 * Label Output:: Output and generation of labels.
7863 * Initialization:: General principles of initialization
7864 and termination routines.
7865 * Macros for Initialization::
7866 Specific macros that control the handling of
7867 initialization and termination routines.
7868 * Instruction Output:: Output of actual instructions.
7869 * Dispatch Tables:: Output of jump tables.
7870 * Exception Region Output:: Output of exception region code.
7871 * Alignment Output:: Pseudo ops for alignment and skipping data.
7874 @node File Framework
7875 @subsection The Overall Framework of an Assembler File
7876 @cindex assembler format
7877 @cindex output of assembler code
7879 @c prevent bad page break with this line
7880 This describes the overall framework of an assembly file.
7882 @findex default_file_start
7883 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7884 Output to @code{asm_out_file} any text which the assembler expects to
7885 find at the beginning of a file. The default behavior is controlled
7886 by two flags, documented below. Unless your target's assembler is
7887 quite unusual, if you override the default, you should call
7888 @code{default_file_start} at some point in your target hook. This
7889 lets other target files rely on these variables.
7892 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7893 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7894 printed as the very first line in the assembly file, unless
7895 @option{-fverbose-asm} is in effect. (If that macro has been defined
7896 to the empty string, this variable has no effect.) With the normal
7897 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7898 assembler that it need not bother stripping comments or extra
7899 whitespace from its input. This allows it to work a bit faster.
7901 The default is false. You should not set it to true unless you have
7902 verified that your port does not generate any extra whitespace or
7903 comments that will cause GAS to issue errors in NO_APP mode.
7906 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7907 If this flag is true, @code{output_file_directive} will be called
7908 for the primary source file, immediately after printing
7909 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7910 this to be done. The default is false.
7913 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7914 Output to @code{asm_out_file} any text which the assembler expects
7915 to find at the end of a file. The default is to output nothing.
7918 @deftypefun void file_end_indicate_exec_stack ()
7919 Some systems use a common convention, the @samp{.note.GNU-stack}
7920 special section, to indicate whether or not an object file relies on
7921 the stack being executable. If your system uses this convention, you
7922 should define @code{TARGET_ASM_FILE_END} to this function. If you
7923 need to do other things in that hook, have your hook function call
7927 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7928 Output to @code{asm_out_file} any text which the assembler expects
7929 to find at the start of an LTO section. The default is to output
7933 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7934 Output to @code{asm_out_file} any text which the assembler expects
7935 to find at the end of an LTO section. The default is to output
7939 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7940 Output to @code{asm_out_file} any text which is needed before emitting
7941 unwind info and debug info at the end of a file. Some targets emit
7942 here PIC setup thunks that cannot be emitted at the end of file,
7943 because they couldn't have unwind info then. The default is to output
7947 @defmac ASM_COMMENT_START
7948 A C string constant describing how to begin a comment in the target
7949 assembler language. The compiler assumes that the comment will end at
7950 the end of the line.
7954 A C string constant for text to be output before each @code{asm}
7955 statement or group of consecutive ones. Normally this is
7956 @code{"#APP"}, which is a comment that has no effect on most
7957 assemblers but tells the GNU assembler that it must check the lines
7958 that follow for all valid assembler constructs.
7962 A C string constant for text to be output after each @code{asm}
7963 statement or group of consecutive ones. Normally this is
7964 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7965 time-saving assumptions that are valid for ordinary compiler output.
7968 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7969 A C statement to output COFF information or DWARF debugging information
7970 which indicates that filename @var{name} is the current source file to
7971 the stdio stream @var{stream}.
7973 This macro need not be defined if the standard form of output
7974 for the file format in use is appropriate.
7977 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7978 Output DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7980 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7983 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_IDENT (const char *@var{name})
7984 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.
7987 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7988 A C statement to output the string @var{string} to the stdio stream
7989 @var{stream}. If you do not call the function @code{output_quoted_string}
7990 in your config files, GCC will only call it to output filenames to
7991 the assembler source. So you can use it to canonicalize the format
7992 of the filename using this macro.
7995 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7996 Output assembly directives to switch to section @var{name}. The section
7997 should have attributes as specified by @var{flags}, which is a bit mask
7998 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7999 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
8000 this section is associated.
8003 @deftypefn {Target Hook} bool TARGET_ASM_ELF_FLAGS_NUMERIC (unsigned int @var{flags}, unsigned int *@var{num})
8004 This hook can be used to encode ELF section flags for which no letter
8005 code has been defined in the assembler. It is called by
8006 @code{default_asm_named_section} whenever the section flags need to be
8007 emitted in the assembler output. If the hook returns true, then the
8008 numerical value for ELF section flags should be calculated from
8009 @var{flags} and saved in @var{*num}; the value is printed out instead of the
8010 normal sequence of letter codes. If the hook is not defined, or if it
8011 returns false, then @var{num} is ignored and the traditional letter sequence
8015 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
8016 Return preferred text (sub)section for function @var{decl}.
8017 Main purpose of this function is to separate cold, normal and hot
8018 functions. @var{startup} is true when function is known to be used only
8019 at startup (from static constructors or it is @code{main()}).
8020 @var{exit} is true when function is known to be used only at exit
8021 (from static destructors).
8022 Return NULL if function should go to default text section.
8025 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
8026 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}.
8029 @deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
8030 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
8031 It must not be modified by command-line option processing.
8034 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
8035 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
8036 This flag is true if we can create zeroed data by switching to a BSS
8037 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
8038 This is true on most ELF targets.
8041 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
8042 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
8043 based on a variable or function decl, a section name, and whether or not the
8044 declaration's initializer may contain runtime relocations. @var{decl} may be
8045 null, in which case read-write data should be assumed.
8047 The default version of this function handles choosing code vs data,
8048 read-only vs read-write data, and @code{flag_pic}. You should only
8049 need to override this if your target has special flags that might be
8050 set via @code{__attribute__}.
8053 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
8054 Provides the target with the ability to record the gcc command line
8055 switches that have been passed to the compiler, and options that are
8056 enabled. The @var{type} argument specifies what is being recorded.
8057 It can take the following values:
8060 @item SWITCH_TYPE_PASSED
8061 @var{text} is a command line switch that has been set by the user.
8063 @item SWITCH_TYPE_ENABLED
8064 @var{text} is an option which has been enabled. This might be as a
8065 direct result of a command line switch, or because it is enabled by
8066 default or because it has been enabled as a side effect of a different
8067 command line switch. For example, the @option{-O2} switch enables
8068 various different individual optimization passes.
8070 @item SWITCH_TYPE_DESCRIPTIVE
8071 @var{text} is either NULL or some descriptive text which should be
8072 ignored. If @var{text} is NULL then it is being used to warn the
8073 target hook that either recording is starting or ending. The first
8074 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
8075 warning is for start up and the second time the warning is for
8076 wind down. This feature is to allow the target hook to make any
8077 necessary preparations before it starts to record switches and to
8078 perform any necessary tidying up after it has finished recording
8081 @item SWITCH_TYPE_LINE_START
8082 This option can be ignored by this target hook.
8084 @item SWITCH_TYPE_LINE_END
8085 This option can be ignored by this target hook.
8088 The hook's return value must be zero. Other return values may be
8089 supported in the future.
8091 By default this hook is set to NULL, but an example implementation is
8092 provided for ELF based targets. Called @var{elf_record_gcc_switches},
8093 it records the switches as ASCII text inside a new, string mergeable
8094 section in the assembler output file. The name of the new section is
8095 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
8099 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
8100 This is the name of the section that will be created by the example
8101 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
8107 @subsection Output of Data
8110 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
8111 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
8112 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_PSI_OP
8113 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
8114 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_PDI_OP
8115 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
8116 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_PTI_OP
8117 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
8118 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
8119 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_PSI_OP
8120 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
8121 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_PDI_OP
8122 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
8123 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_PTI_OP
8124 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
8125 These hooks specify assembly directives for creating certain kinds
8126 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
8127 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
8128 aligned two-byte object, and so on. Any of the hooks may be
8129 @code{NULL}, indicating that no suitable directive is available.
8131 The compiler will print these strings at the start of a new line,
8132 followed immediately by the object's initial value. In most cases,
8133 the string should contain a tab, a pseudo-op, and then another tab.
8136 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
8137 The @code{assemble_integer} function uses this hook to output an
8138 integer object. @var{x} is the object's value, @var{size} is its size
8139 in bytes and @var{aligned_p} indicates whether it is aligned. The
8140 function should return @code{true} if it was able to output the
8141 object. If it returns false, @code{assemble_integer} will try to
8142 split the object into smaller parts.
8144 The default implementation of this hook will use the
8145 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
8146 when the relevant string is @code{NULL}.
8149 @deftypefn {Target Hook} void TARGET_ASM_DECL_END (void)
8150 Define this hook if the target assembler requires a special marker to
8151 terminate an initialized variable declaration.
8154 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
8155 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
8156 can't deal with, and output assembly code to @var{file} corresponding to
8157 the pattern @var{x}. This may be used to allow machine-dependent
8158 @code{UNSPEC}s to appear within constants.
8160 If target hook fails to recognize a pattern, it must return @code{false},
8161 so that a standard error message is printed. If it prints an error message
8162 itself, by calling, for example, @code{output_operand_lossage}, it may just
8166 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
8167 A C statement to output to the stdio stream @var{stream} an assembler
8168 instruction to assemble a string constant containing the @var{len}
8169 bytes at @var{ptr}. @var{ptr} will be a C expression of type
8170 @code{char *} and @var{len} a C expression of type @code{int}.
8172 If the assembler has a @code{.ascii} pseudo-op as found in the
8173 Berkeley Unix assembler, do not define the macro
8174 @code{ASM_OUTPUT_ASCII}.
8177 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
8178 A C statement to output word @var{n} of a function descriptor for
8179 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
8180 is defined, and is otherwise unused.
8183 @defmac CONSTANT_POOL_BEFORE_FUNCTION
8184 You may define this macro as a C expression. You should define the
8185 expression to have a nonzero value if GCC should output the constant
8186 pool for a function before the code for the function, or a zero value if
8187 GCC should output the constant pool after the function. If you do
8188 not define this macro, the usual case, GCC will output the constant
8189 pool before the function.
8192 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
8193 A C statement to output assembler commands to define the start of the
8194 constant pool for a function. @var{funname} is a string giving
8195 the name of the function. Should the return type of the function
8196 be required, it can be obtained via @var{fundecl}. @var{size}
8197 is the size, in bytes, of the constant pool that will be written
8198 immediately after this call.
8200 If no constant-pool prefix is required, the usual case, this macro need
8204 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
8205 A C statement (with or without semicolon) to output a constant in the
8206 constant pool, if it needs special treatment. (This macro need not do
8207 anything for RTL expressions that can be output normally.)
8209 The argument @var{file} is the standard I/O stream to output the
8210 assembler code on. @var{x} is the RTL expression for the constant to
8211 output, and @var{mode} is the machine mode (in case @var{x} is a
8212 @samp{const_int}). @var{align} is the required alignment for the value
8213 @var{x}; you should output an assembler directive to force this much
8216 The argument @var{labelno} is a number to use in an internal label for
8217 the address of this pool entry. The definition of this macro is
8218 responsible for outputting the label definition at the proper place.
8219 Here is how to do this:
8222 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
8225 When you output a pool entry specially, you should end with a
8226 @code{goto} to the label @var{jumpto}. This will prevent the same pool
8227 entry from being output a second time in the usual manner.
8229 You need not define this macro if it would do nothing.
8232 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
8233 A C statement to output assembler commands to at the end of the constant
8234 pool for a function. @var{funname} is a string giving the name of the
8235 function. Should the return type of the function be required, you can
8236 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
8237 constant pool that GCC wrote immediately before this call.
8239 If no constant-pool epilogue is required, the usual case, you need not
8243 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
8244 Define this macro as a C expression which is nonzero if @var{C} is
8245 used as a logical line separator by the assembler. @var{STR} points
8246 to the position in the string where @var{C} was found; this can be used if
8247 a line separator uses multiple characters.
8249 If you do not define this macro, the default is that only
8250 the character @samp{;} is treated as a logical line separator.
8253 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
8254 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
8255 These target hooks are C string constants, describing the syntax in the
8256 assembler for grouping arithmetic expressions. If not overridden, they
8257 default to normal parentheses, which is correct for most assemblers.
8260 These macros are provided by @file{real.h} for writing the definitions
8261 of @code{ASM_OUTPUT_DOUBLE} and the like:
8263 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
8264 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
8265 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
8266 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
8267 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
8268 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
8269 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
8270 target's floating point representation, and store its bit pattern in
8271 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
8272 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
8273 simple @code{long int}. For the others, it should be an array of
8274 @code{long int}. The number of elements in this array is determined
8275 by the size of the desired target floating point data type: 32 bits of
8276 it go in each @code{long int} array element. Each array element holds
8277 32 bits of the result, even if @code{long int} is wider than 32 bits
8278 on the host machine.
8280 The array element values are designed so that you can print them out
8281 using @code{fprintf} in the order they should appear in the target
8285 @node Uninitialized Data
8286 @subsection Output of Uninitialized Variables
8288 Each of the macros in this section is used to do the whole job of
8289 outputting a single uninitialized variable.
8291 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
8292 A C statement (sans semicolon) to output to the stdio stream
8293 @var{stream} the assembler definition of a common-label named
8294 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
8295 is the size rounded up to whatever alignment the caller wants. It is
8296 possible that @var{size} may be zero, for instance if a struct with no
8297 other member than a zero-length array is defined. In this case, the
8298 backend must output a symbol definition that allocates at least one
8299 byte, both so that the address of the resulting object does not compare
8300 equal to any other, and because some object formats cannot even express
8301 the concept of a zero-sized common symbol, as that is how they represent
8302 an ordinary undefined external.
8304 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8305 output the name itself; before and after that, output the additional
8306 assembler syntax for defining the name, and a newline.
8308 This macro controls how the assembler definitions of uninitialized
8309 common global variables are output.
8312 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
8313 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
8314 separate, explicit argument. If you define this macro, it is used in
8315 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
8316 handling the required alignment of the variable. The alignment is specified
8317 as the number of bits.
8320 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
8321 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
8322 variable to be output, if there is one, or @code{NULL_TREE} if there
8323 is no corresponding variable. If you define this macro, GCC will use it
8324 in place of both @code{ASM_OUTPUT_COMMON} and
8325 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
8326 the variable's decl in order to chose what to output.
8329 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
8330 A C statement (sans semicolon) to output to the stdio stream
8331 @var{stream} the assembler definition of uninitialized global @var{decl} named
8332 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
8333 is the alignment specified as the number of bits.
8335 Try to use function @code{asm_output_aligned_bss} defined in file
8336 @file{varasm.c} when defining this macro. If unable, use the expression
8337 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
8338 before and after that, output the additional assembler syntax for defining
8339 the name, and a newline.
8341 There are two ways of handling global BSS@. One is to define this macro.
8342 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
8343 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
8344 You do not need to do both.
8346 Some languages do not have @code{common} data, and require a
8347 non-common form of global BSS in order to handle uninitialized globals
8348 efficiently. C++ is one example of this. However, if the target does
8349 not support global BSS, the front end may choose to make globals
8350 common in order to save space in the object file.
8353 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
8354 A C statement (sans semicolon) to output to the stdio stream
8355 @var{stream} the assembler definition of a local-common-label named
8356 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
8357 is the size rounded up to whatever alignment the caller wants.
8359 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8360 output the name itself; before and after that, output the additional
8361 assembler syntax for defining the name, and a newline.
8363 This macro controls how the assembler definitions of uninitialized
8364 static variables are output.
8367 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
8368 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
8369 separate, explicit argument. If you define this macro, it is used in
8370 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
8371 handling the required alignment of the variable. The alignment is specified
8372 as the number of bits.
8375 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
8376 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
8377 variable to be output, if there is one, or @code{NULL_TREE} if there
8378 is no corresponding variable. If you define this macro, GCC will use it
8379 in place of both @code{ASM_OUTPUT_DECL} and
8380 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
8381 the variable's decl in order to chose what to output.
8385 @subsection Output and Generation of Labels
8387 @c prevent bad page break with this line
8388 This is about outputting labels.
8390 @findex assemble_name
8391 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
8392 A C statement (sans semicolon) to output to the stdio stream
8393 @var{stream} the assembler definition of a label named @var{name}.
8394 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8395 output the name itself; before and after that, output the additional
8396 assembler syntax for defining the name, and a newline. A default
8397 definition of this macro is provided which is correct for most systems.
8400 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
8401 A C statement (sans semicolon) to output to the stdio stream
8402 @var{stream} the assembler definition of a label named @var{name} of
8404 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
8405 output the name itself; before and after that, output the additional
8406 assembler syntax for defining the name, and a newline. A default
8407 definition of this macro is provided which is correct for most systems.
8409 If this macro is not defined, then the function name is defined in the
8410 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8413 @findex assemble_name_raw
8414 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
8415 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
8416 to refer to a compiler-generated label. The default definition uses
8417 @code{assemble_name_raw}, which is like @code{assemble_name} except
8418 that it is more efficient.
8422 A C string containing the appropriate assembler directive to specify the
8423 size of a symbol, without any arguments. On systems that use ELF, the
8424 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
8425 systems, the default is not to define this macro.
8427 Define this macro only if it is correct to use the default definitions
8428 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
8429 for your system. If you need your own custom definitions of those
8430 macros, or if you do not need explicit symbol sizes at all, do not
8434 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
8435 A C statement (sans semicolon) to output to the stdio stream
8436 @var{stream} a directive telling the assembler that the size of the
8437 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
8438 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
8442 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
8443 A C statement (sans semicolon) to output to the stdio stream
8444 @var{stream} a directive telling the assembler to calculate the size of
8445 the symbol @var{name} by subtracting its address from the current
8448 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
8449 provided. The default assumes that the assembler recognizes a special
8450 @samp{.} symbol as referring to the current address, and can calculate
8451 the difference between this and another symbol. If your assembler does
8452 not recognize @samp{.} or cannot do calculations with it, you will need
8453 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
8456 @defmac NO_DOLLAR_IN_LABEL
8457 Define this macro if the assembler does not accept the character
8458 @samp{$} in label names. By default constructors and destructors in
8459 G++ have @samp{$} in the identifiers. If this macro is defined,
8460 @samp{.} is used instead.
8463 @defmac NO_DOT_IN_LABEL
8464 Define this macro if the assembler does not accept the character
8465 @samp{.} in label names. By default constructors and destructors in G++
8466 have names that use @samp{.}. If this macro is defined, these names
8467 are rewritten to avoid @samp{.}.
8471 A C string containing the appropriate assembler directive to specify the
8472 type of a symbol, without any arguments. On systems that use ELF, the
8473 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
8474 systems, the default is not to define this macro.
8476 Define this macro only if it is correct to use the default definition of
8477 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
8478 custom definition of this macro, or if you do not need explicit symbol
8479 types at all, do not define this macro.
8482 @defmac TYPE_OPERAND_FMT
8483 A C string which specifies (using @code{printf} syntax) the format of
8484 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
8485 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
8486 the default is not to define this macro.
8488 Define this macro only if it is correct to use the default definition of
8489 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
8490 custom definition of this macro, or if you do not need explicit symbol
8491 types at all, do not define this macro.
8494 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
8495 A C statement (sans semicolon) to output to the stdio stream
8496 @var{stream} a directive telling the assembler that the type of the
8497 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
8498 that string is always either @samp{"function"} or @samp{"object"}, but
8499 you should not count on this.
8501 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
8502 definition of this macro is provided.
8505 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
8506 A C statement (sans semicolon) to output to the stdio stream
8507 @var{stream} any text necessary for declaring the name @var{name} of a
8508 function which is being defined. This macro is responsible for
8509 outputting the label definition (perhaps using
8510 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
8511 @code{FUNCTION_DECL} tree node representing the function.
8513 If this macro is not defined, then the function name is defined in the
8514 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
8516 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
8520 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
8521 A C statement (sans semicolon) to output to the stdio stream
8522 @var{stream} any text necessary for declaring the size of a function
8523 which is being defined. The argument @var{name} is the name of the
8524 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
8525 representing the function.
8527 If this macro is not defined, then the function size is not defined.
8529 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
8533 @defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
8534 A C statement (sans semicolon) to output to the stdio stream
8535 @var{stream} any text necessary for declaring the name @var{name} of a
8536 cold function partition which is being defined. This macro is responsible
8537 for outputting the label definition (perhaps using
8538 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
8539 @code{FUNCTION_DECL} tree node representing the function.
8541 If this macro is not defined, then the cold partition name is defined in the
8542 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8544 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
8548 @defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
8549 A C statement (sans semicolon) to output to the stdio stream
8550 @var{stream} any text necessary for declaring the size of a cold function
8551 partition which is being defined. The argument @var{name} is the name of the
8552 cold partition of the function. The argument @var{decl} is the
8553 @code{FUNCTION_DECL} tree node representing the function.
8555 If this macro is not defined, then the partition size is not defined.
8557 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
8561 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
8562 A C statement (sans semicolon) to output to the stdio stream
8563 @var{stream} any text necessary for declaring the name @var{name} of an
8564 initialized variable which is being defined. This macro must output the
8565 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
8566 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
8568 If this macro is not defined, then the variable name is defined in the
8569 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8571 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
8572 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
8575 @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})
8576 A target hook to output to the stdio stream @var{file} any text necessary
8577 for declaring the name @var{name} of a constant which is being defined. This
8578 target hook is responsible for outputting the label definition (perhaps using
8579 @code{assemble_label}). The argument @var{exp} is the value of the constant,
8580 and @var{size} is the size of the constant in bytes. The @var{name}
8581 will be an internal label.
8583 The default version of this target hook, define the @var{name} in the
8584 usual manner as a label (by means of @code{assemble_label}).
8586 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
8589 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
8590 A C statement (sans semicolon) to output to the stdio stream
8591 @var{stream} any text necessary for claiming a register @var{regno}
8592 for a global variable @var{decl} with name @var{name}.
8594 If you don't define this macro, that is equivalent to defining it to do
8598 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
8599 A C statement (sans semicolon) to finish up declaring a variable name
8600 once the compiler has processed its initializer fully and thus has had a
8601 chance to determine the size of an array when controlled by an
8602 initializer. This is used on systems where it's necessary to declare
8603 something about the size of the object.
8605 If you don't define this macro, that is equivalent to defining it to do
8608 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
8609 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
8612 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
8613 This target hook is a function to output to the stdio stream
8614 @var{stream} some commands that will make the label @var{name} global;
8615 that is, available for reference from other files.
8617 The default implementation relies on a proper definition of
8618 @code{GLOBAL_ASM_OP}.
8621 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
8622 This target hook is a function to output to the stdio stream
8623 @var{stream} some commands that will make the name associated with @var{decl}
8624 global; that is, available for reference from other files.
8626 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
8629 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_UNDEFINED_DECL (FILE *@var{stream}, const char *@var{name}, const_tree @var{decl})
8630 This target hook is a function to output to the stdio stream
8631 @var{stream} some commands that will declare the name associated with
8632 @var{decl} which is not defined in the current translation unit. Most
8633 assemblers do not require anything to be output in this case.
8636 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
8637 A C statement (sans semicolon) to output to the stdio stream
8638 @var{stream} some commands that will make the label @var{name} weak;
8639 that is, available for reference from other files but only used if
8640 no other definition is available. Use the expression
8641 @code{assemble_name (@var{stream}, @var{name})} to output the name
8642 itself; before and after that, output the additional assembler syntax
8643 for making that name weak, and a newline.
8645 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
8646 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
8650 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
8651 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
8652 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
8653 or variable decl. If @var{value} is not @code{NULL}, this C statement
8654 should output to the stdio stream @var{stream} assembler code which
8655 defines (equates) the weak symbol @var{name} to have the value
8656 @var{value}. If @var{value} is @code{NULL}, it should output commands
8657 to make @var{name} weak.
8660 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
8661 Outputs a directive that enables @var{name} to be used to refer to
8662 symbol @var{value} with weak-symbol semantics. @code{decl} is the
8663 declaration of @code{name}.
8666 @defmac SUPPORTS_WEAK
8667 A preprocessor constant expression which evaluates to true if the target
8668 supports weak symbols.
8670 If you don't define this macro, @file{defaults.h} provides a default
8671 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
8672 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
8675 @defmac TARGET_SUPPORTS_WEAK
8676 A C expression which evaluates to true if the target supports weak symbols.
8678 If you don't define this macro, @file{defaults.h} provides a default
8679 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
8680 this macro if you want to control weak symbol support with a compiler
8681 flag such as @option{-melf}.
8684 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
8685 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
8686 public symbol such that extra copies in multiple translation units will
8687 be discarded by the linker. Define this macro if your object file
8688 format provides support for this concept, such as the @samp{COMDAT}
8689 section flags in the Microsoft Windows PE/COFF format, and this support
8690 requires changes to @var{decl}, such as putting it in a separate section.
8693 @defmac SUPPORTS_ONE_ONLY
8694 A C expression which evaluates to true if the target supports one-only
8697 If you don't define this macro, @file{varasm.c} provides a default
8698 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
8699 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
8700 you want to control one-only symbol support with a compiler flag, or if
8701 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
8702 be emitted as one-only.
8705 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
8706 This target hook is a function to output to @var{asm_out_file} some
8707 commands that will make the symbol(s) associated with @var{decl} have
8708 hidden, protected or internal visibility as specified by @var{visibility}.
8711 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
8712 A C expression that evaluates to true if the target's linker expects
8713 that weak symbols do not appear in a static archive's table of contents.
8714 The default is @code{0}.
8716 Leaving weak symbols out of an archive's table of contents means that,
8717 if a symbol will only have a definition in one translation unit and
8718 will have undefined references from other translation units, that
8719 symbol should not be weak. Defining this macro to be nonzero will
8720 thus have the effect that certain symbols that would normally be weak
8721 (explicit template instantiations, and vtables for polymorphic classes
8722 with noninline key methods) will instead be nonweak.
8724 The C++ ABI requires this macro to be zero. Define this macro for
8725 targets where full C++ ABI compliance is impossible and where linker
8726 restrictions require weak symbols to be left out of a static archive's
8730 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8731 A C statement (sans semicolon) to output to the stdio stream
8732 @var{stream} any text necessary for declaring the name of an external
8733 symbol named @var{name} which is referenced in this compilation but
8734 not defined. The value of @var{decl} is the tree node for the
8737 This macro need not be defined if it does not need to output anything.
8738 The GNU assembler and most Unix assemblers don't require anything.
8741 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8742 This target hook is a function to output to @var{asm_out_file} an assembler
8743 pseudo-op to declare a library function name external. The name of the
8744 library function is given by @var{symref}, which is a @code{symbol_ref}.
8747 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8748 This target hook is a function to output to @var{asm_out_file} an assembler
8749 directive to annotate @var{symbol} as used. The Darwin target uses the
8750 .no_dead_code_strip directive.
8753 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8754 A C statement (sans semicolon) to output to the stdio stream
8755 @var{stream} a reference in assembler syntax to a label named
8756 @var{name}. This should add @samp{_} to the front of the name, if that
8757 is customary on your operating system, as it is in most Berkeley Unix
8758 systems. This macro is used in @code{assemble_name}.
8761 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8762 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.
8765 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8766 A C statement (sans semicolon) to output a reference to
8767 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8768 will be used to output the name of the symbol. This macro may be used
8769 to modify the way a symbol is referenced depending on information
8770 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8773 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8774 A C statement (sans semicolon) to output a reference to @var{buf}, the
8775 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8776 @code{assemble_name} will be used to output the name of the symbol.
8777 This macro is not used by @code{output_asm_label}, or the @code{%l}
8778 specifier that calls it; the intention is that this macro should be set
8779 when it is necessary to output a label differently when its address is
8783 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8784 A function to output to the stdio stream @var{stream} a label whose
8785 name is made from the string @var{prefix} and the number @var{labelno}.
8787 It is absolutely essential that these labels be distinct from the labels
8788 used for user-level functions and variables. Otherwise, certain programs
8789 will have name conflicts with internal labels.
8791 It is desirable to exclude internal labels from the symbol table of the
8792 object file. Most assemblers have a naming convention for labels that
8793 should be excluded; on many systems, the letter @samp{L} at the
8794 beginning of a label has this effect. You should find out what
8795 convention your system uses, and follow it.
8797 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8800 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8801 A C statement to output to the stdio stream @var{stream} a debug info
8802 label whose name is made from the string @var{prefix} and the number
8803 @var{num}. This is useful for VLIW targets, where debug info labels
8804 may need to be treated differently than branch target labels. On some
8805 systems, branch target labels must be at the beginning of instruction
8806 bundles, but debug info labels can occur in the middle of instruction
8809 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8813 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8814 A C statement to store into the string @var{string} a label whose name
8815 is made from the string @var{prefix} and the number @var{num}.
8817 This string, when output subsequently by @code{assemble_name}, should
8818 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8819 with the same @var{prefix} and @var{num}.
8821 If the string begins with @samp{*}, then @code{assemble_name} will
8822 output the rest of the string unchanged. It is often convenient for
8823 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8824 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8825 to output the string, and may change it. (Of course,
8826 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8827 you should know what it does on your machine.)
8830 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8831 A C expression to assign to @var{outvar} (which is a variable of type
8832 @code{char *}) a newly allocated string made from the string
8833 @var{name} and the number @var{number}, with some suitable punctuation
8834 added. Use @code{alloca} to get space for the string.
8836 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8837 produce an assembler label for an internal static variable whose name is
8838 @var{name}. Therefore, the string must be such as to result in valid
8839 assembler code. The argument @var{number} is different each time this
8840 macro is executed; it prevents conflicts between similarly-named
8841 internal static variables in different scopes.
8843 Ideally this string should not be a valid C identifier, to prevent any
8844 conflict with the user's own symbols. Most assemblers allow periods
8845 or percent signs in assembler symbols; putting at least one of these
8846 between the name and the number will suffice.
8848 If this macro is not defined, a default definition will be provided
8849 which is correct for most systems.
8852 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8853 A C statement to output to the stdio stream @var{stream} assembler code
8854 which defines (equates) the symbol @var{name} to have the value @var{value}.
8857 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8858 correct for most systems.
8861 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8862 A C statement to output to the stdio stream @var{stream} assembler code
8863 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8864 to have the value of the tree node @var{decl_of_value}. This macro will
8865 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8866 the tree nodes are available.
8869 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8870 correct for most systems.
8873 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8874 A C statement that evaluates to true if the assembler code which defines
8875 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8876 of the tree node @var{decl_of_value} should be emitted near the end of the
8877 current compilation unit. The default is to not defer output of defines.
8878 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8879 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8882 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8883 A C statement to output to the stdio stream @var{stream} assembler code
8884 which defines (equates) the weak symbol @var{name} to have the value
8885 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8886 an undefined weak symbol.
8888 Define this macro if the target only supports weak aliases; define
8889 @code{ASM_OUTPUT_DEF} instead if possible.
8892 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8893 Define this macro to override the default assembler names used for
8894 Objective-C methods.
8896 The default name is a unique method number followed by the name of the
8897 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8898 the category is also included in the assembler name (e.g.@:
8901 These names are safe on most systems, but make debugging difficult since
8902 the method's selector is not present in the name. Therefore, particular
8903 systems define other ways of computing names.
8905 @var{buf} is an expression of type @code{char *} which gives you a
8906 buffer in which to store the name; its length is as long as
8907 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8908 50 characters extra.
8910 The argument @var{is_inst} specifies whether the method is an instance
8911 method or a class method; @var{class_name} is the name of the class;
8912 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8913 in a category); and @var{sel_name} is the name of the selector.
8915 On systems where the assembler can handle quoted names, you can use this
8916 macro to provide more human-readable names.
8919 @node Initialization
8920 @subsection How Initialization Functions Are Handled
8921 @cindex initialization routines
8922 @cindex termination routines
8923 @cindex constructors, output of
8924 @cindex destructors, output of
8926 The compiled code for certain languages includes @dfn{constructors}
8927 (also called @dfn{initialization routines})---functions to initialize
8928 data in the program when the program is started. These functions need
8929 to be called before the program is ``started''---that is to say, before
8930 @code{main} is called.
8932 Compiling some languages generates @dfn{destructors} (also called
8933 @dfn{termination routines}) that should be called when the program
8936 To make the initialization and termination functions work, the compiler
8937 must output something in the assembler code to cause those functions to
8938 be called at the appropriate time. When you port the compiler to a new
8939 system, you need to specify how to do this.
8941 There are two major ways that GCC currently supports the execution of
8942 initialization and termination functions. Each way has two variants.
8943 Much of the structure is common to all four variations.
8945 @findex __CTOR_LIST__
8946 @findex __DTOR_LIST__
8947 The linker must build two lists of these functions---a list of
8948 initialization functions, called @code{__CTOR_LIST__}, and a list of
8949 termination functions, called @code{__DTOR_LIST__}.
8951 Each list always begins with an ignored function pointer (which may hold
8952 0, @minus{}1, or a count of the function pointers after it, depending on
8953 the environment). This is followed by a series of zero or more function
8954 pointers to constructors (or destructors), followed by a function
8955 pointer containing zero.
8957 Depending on the operating system and its executable file format, either
8958 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8959 time and exit time. Constructors are called in reverse order of the
8960 list; destructors in forward order.
8962 The best way to handle static constructors works only for object file
8963 formats which provide arbitrarily-named sections. A section is set
8964 aside for a list of constructors, and another for a list of destructors.
8965 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8966 object file that defines an initialization function also puts a word in
8967 the constructor section to point to that function. The linker
8968 accumulates all these words into one contiguous @samp{.ctors} section.
8969 Termination functions are handled similarly.
8971 This method will be chosen as the default by @file{target-def.h} if
8972 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8973 support arbitrary sections, but does support special designated
8974 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8975 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8977 When arbitrary sections are available, there are two variants, depending
8978 upon how the code in @file{crtstuff.c} is called. On systems that
8979 support a @dfn{.init} section which is executed at program startup,
8980 parts of @file{crtstuff.c} are compiled into that section. The
8981 program is linked by the @command{gcc} driver like this:
8984 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8987 The prologue of a function (@code{__init}) appears in the @code{.init}
8988 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8989 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8990 files are provided by the operating system or by the GNU C library, but
8991 are provided by GCC for a few targets.
8993 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8994 compiled from @file{crtstuff.c}. They contain, among other things, code
8995 fragments within the @code{.init} and @code{.fini} sections that branch
8996 to routines in the @code{.text} section. The linker will pull all parts
8997 of a section together, which results in a complete @code{__init} function
8998 that invokes the routines we need at startup.
9000 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
9003 If no init section is available, when GCC compiles any function called
9004 @code{main} (or more accurately, any function designated as a program
9005 entry point by the language front end calling @code{expand_main_function}),
9006 it inserts a procedure call to @code{__main} as the first executable code
9007 after the function prologue. The @code{__main} function is defined
9008 in @file{libgcc2.c} and runs the global constructors.
9010 In file formats that don't support arbitrary sections, there are again
9011 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
9012 and an `a.out' format must be used. In this case,
9013 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
9014 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
9015 and with the address of the void function containing the initialization
9016 code as its value. The GNU linker recognizes this as a request to add
9017 the value to a @dfn{set}; the values are accumulated, and are eventually
9018 placed in the executable as a vector in the format described above, with
9019 a leading (ignored) count and a trailing zero element.
9020 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
9021 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
9022 the compilation of @code{main} to call @code{__main} as above, starting
9023 the initialization process.
9025 The last variant uses neither arbitrary sections nor the GNU linker.
9026 This is preferable when you want to do dynamic linking and when using
9027 file formats which the GNU linker does not support, such as `ECOFF'@. In
9028 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
9029 termination functions are recognized simply by their names. This requires
9030 an extra program in the linkage step, called @command{collect2}. This program
9031 pretends to be the linker, for use with GCC; it does its job by running
9032 the ordinary linker, but also arranges to include the vectors of
9033 initialization and termination functions. These functions are called
9034 via @code{__main} as described above. In order to use this method,
9035 @code{use_collect2} must be defined in the target in @file{config.gcc}.
9038 The following section describes the specific macros that control and
9039 customize the handling of initialization and termination functions.
9042 @node Macros for Initialization
9043 @subsection Macros Controlling Initialization Routines
9045 Here are the macros that control how the compiler handles initialization
9046 and termination functions:
9048 @defmac INIT_SECTION_ASM_OP
9049 If defined, a C string constant, including spacing, for the assembler
9050 operation to identify the following data as initialization code. If not
9051 defined, GCC will assume such a section does not exist. When you are
9052 using special sections for initialization and termination functions, this
9053 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
9054 run the initialization functions.
9057 @defmac HAS_INIT_SECTION
9058 If defined, @code{main} will not call @code{__main} as described above.
9059 This macro should be defined for systems that control start-up code
9060 on a symbol-by-symbol basis, such as OSF/1, and should not
9061 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
9064 @defmac LD_INIT_SWITCH
9065 If defined, a C string constant for a switch that tells the linker that
9066 the following symbol is an initialization routine.
9069 @defmac LD_FINI_SWITCH
9070 If defined, a C string constant for a switch that tells the linker that
9071 the following symbol is a finalization routine.
9074 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
9075 If defined, a C statement that will write a function that can be
9076 automatically called when a shared library is loaded. The function
9077 should call @var{func}, which takes no arguments. If not defined, and
9078 the object format requires an explicit initialization function, then a
9079 function called @code{_GLOBAL__DI} will be generated.
9081 This function and the following one are used by collect2 when linking a
9082 shared library that needs constructors or destructors, or has DWARF2
9083 exception tables embedded in the code.
9086 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
9087 If defined, a C statement that will write a function that can be
9088 automatically called when a shared library is unloaded. The function
9089 should call @var{func}, which takes no arguments. If not defined, and
9090 the object format requires an explicit finalization function, then a
9091 function called @code{_GLOBAL__DD} will be generated.
9094 @defmac INVOKE__main
9095 If defined, @code{main} will call @code{__main} despite the presence of
9096 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
9097 where the init section is not actually run automatically, but is still
9098 useful for collecting the lists of constructors and destructors.
9101 @defmac SUPPORTS_INIT_PRIORITY
9102 If nonzero, the C++ @code{init_priority} attribute is supported and the
9103 compiler should emit instructions to control the order of initialization
9104 of objects. If zero, the compiler will issue an error message upon
9105 encountering an @code{init_priority} attribute.
9108 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
9109 This value is true if the target supports some ``native'' method of
9110 collecting constructors and destructors to be run at startup and exit.
9111 It is false if we must use @command{collect2}.
9114 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
9115 If defined, a function that outputs assembler code to arrange to call
9116 the function referenced by @var{symbol} at initialization time.
9118 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
9119 no arguments and with no return value. If the target supports initialization
9120 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
9121 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
9123 If this macro is not defined by the target, a suitable default will
9124 be chosen if (1) the target supports arbitrary section names, (2) the
9125 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
9129 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
9130 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
9131 functions rather than initialization functions.
9134 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
9135 generated for the generated object file will have static linkage.
9137 If your system uses @command{collect2} as the means of processing
9138 constructors, then that program normally uses @command{nm} to scan
9139 an object file for constructor functions to be called.
9141 On certain kinds of systems, you can define this macro to make
9142 @command{collect2} work faster (and, in some cases, make it work at all):
9144 @defmac OBJECT_FORMAT_COFF
9145 Define this macro if the system uses COFF (Common Object File Format)
9146 object files, so that @command{collect2} can assume this format and scan
9147 object files directly for dynamic constructor/destructor functions.
9149 This macro is effective only in a native compiler; @command{collect2} as
9150 part of a cross compiler always uses @command{nm} for the target machine.
9153 @defmac REAL_NM_FILE_NAME
9154 Define this macro as a C string constant containing the file name to use
9155 to execute @command{nm}. The default is to search the path normally for
9160 @command{collect2} calls @command{nm} to scan object files for static
9161 constructors and destructors and LTO info. By default, @option{-n} is
9162 passed. Define @code{NM_FLAGS} to a C string constant if other options
9163 are needed to get the same output format as GNU @command{nm -n}
9167 If your system supports shared libraries and has a program to list the
9168 dynamic dependencies of a given library or executable, you can define
9169 these macros to enable support for running initialization and
9170 termination functions in shared libraries:
9173 Define this macro to a C string constant containing the name of the program
9174 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
9177 @defmac PARSE_LDD_OUTPUT (@var{ptr})
9178 Define this macro to be C code that extracts filenames from the output
9179 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
9180 of type @code{char *} that points to the beginning of a line of output
9181 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
9182 code must advance @var{ptr} to the beginning of the filename on that
9183 line. Otherwise, it must set @var{ptr} to @code{NULL}.
9186 @defmac SHLIB_SUFFIX
9187 Define this macro to a C string constant containing the default shared
9188 library extension of the target (e.g., @samp{".so"}). @command{collect2}
9189 strips version information after this suffix when generating global
9190 constructor and destructor names. This define is only needed on targets
9191 that use @command{collect2} to process constructors and destructors.
9194 @node Instruction Output
9195 @subsection Output of Assembler Instructions
9197 @c prevent bad page break with this line
9198 This describes assembler instruction output.
9200 @defmac REGISTER_NAMES
9201 A C initializer containing the assembler's names for the machine
9202 registers, each one as a C string constant. This is what translates
9203 register numbers in the compiler into assembler language.
9206 @defmac ADDITIONAL_REGISTER_NAMES
9207 If defined, a C initializer for an array of structures containing a name
9208 and a register number. This macro defines additional names for hard
9209 registers, thus allowing the @code{asm} option in declarations to refer
9210 to registers using alternate names.
9213 @defmac OVERLAPPING_REGISTER_NAMES
9214 If defined, a C initializer for an array of structures containing a
9215 name, a register number and a count of the number of consecutive
9216 machine registers the name overlaps. This macro defines additional
9217 names for hard registers, thus allowing the @code{asm} option in
9218 declarations to refer to registers using alternate names. Unlike
9219 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
9220 register name implies multiple underlying registers.
9222 This macro should be used when it is important that a clobber in an
9223 @code{asm} statement clobbers all the underlying values implied by the
9224 register name. For example, on ARM, clobbering the double-precision
9225 VFP register ``d0'' implies clobbering both single-precision registers
9229 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
9230 Define this macro if you are using an unusual assembler that
9231 requires different names for the machine instructions.
9233 The definition is a C statement or statements which output an
9234 assembler instruction opcode to the stdio stream @var{stream}. The
9235 macro-operand @var{ptr} is a variable of type @code{char *} which
9236 points to the opcode name in its ``internal'' form---the form that is
9237 written in the machine description. The definition should output the
9238 opcode name to @var{stream}, performing any translation you desire, and
9239 increment the variable @var{ptr} to point at the end of the opcode
9240 so that it will not be output twice.
9242 In fact, your macro definition may process less than the entire opcode
9243 name, or more than the opcode name; but if you want to process text
9244 that includes @samp{%}-sequences to substitute operands, you must take
9245 care of the substitution yourself. Just be sure to increment
9246 @var{ptr} over whatever text should not be output normally.
9248 @findex recog_data.operand
9249 If you need to look at the operand values, they can be found as the
9250 elements of @code{recog_data.operand}.
9252 If the macro definition does nothing, the instruction is output
9256 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
9257 If defined, a C statement to be executed just prior to the output of
9258 assembler code for @var{insn}, to modify the extracted operands so
9259 they will be output differently.
9261 Here the argument @var{opvec} is the vector containing the operands
9262 extracted from @var{insn}, and @var{noperands} is the number of
9263 elements of the vector which contain meaningful data for this insn.
9264 The contents of this vector are what will be used to convert the insn
9265 template into assembler code, so you can change the assembler output
9266 by changing the contents of the vector.
9268 This macro is useful when various assembler syntaxes share a single
9269 file of instruction patterns; by defining this macro differently, you
9270 can cause a large class of instructions to be output differently (such
9271 as with rearranged operands). Naturally, variations in assembler
9272 syntax affecting individual insn patterns ought to be handled by
9273 writing conditional output routines in those patterns.
9275 If this macro is not defined, it is equivalent to a null statement.
9278 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx_insn *@var{insn}, rtx *@var{opvec}, int @var{noperands})
9279 If defined, this target hook is a function which is executed just after the
9280 output of assembler code for @var{insn}, to change the mode of the assembler
9283 Here the argument @var{opvec} is the vector containing the operands
9284 extracted from @var{insn}, and @var{noperands} is the number of
9285 elements of the vector which contain meaningful data for this insn.
9286 The contents of this vector are what was used to convert the insn
9287 template into assembler code, so you can change the assembler mode
9288 by checking the contents of the vector.
9291 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
9292 A C compound statement to output to stdio stream @var{stream} the
9293 assembler syntax for an instruction operand @var{x}. @var{x} is an
9296 @var{code} is a value that can be used to specify one of several ways
9297 of printing the operand. It is used when identical operands must be
9298 printed differently depending on the context. @var{code} comes from
9299 the @samp{%} specification that was used to request printing of the
9300 operand. If the specification was just @samp{%@var{digit}} then
9301 @var{code} is 0; if the specification was @samp{%@var{ltr}
9302 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
9305 If @var{x} is a register, this macro should print the register's name.
9306 The names can be found in an array @code{reg_names} whose type is
9307 @code{char *[]}. @code{reg_names} is initialized from
9308 @code{REGISTER_NAMES}.
9310 When the machine description has a specification @samp{%@var{punct}}
9311 (a @samp{%} followed by a punctuation character), this macro is called
9312 with a null pointer for @var{x} and the punctuation character for
9316 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
9317 A C expression which evaluates to true if @var{code} is a valid
9318 punctuation character for use in the @code{PRINT_OPERAND} macro. If
9319 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
9320 punctuation characters (except for the standard one, @samp{%}) are used
9324 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
9325 A C compound statement to output to stdio stream @var{stream} the
9326 assembler syntax for an instruction operand that is a memory reference
9327 whose address is @var{x}. @var{x} is an RTL expression.
9329 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
9330 On some machines, the syntax for a symbolic address depends on the
9331 section that the address refers to. On these machines, define the hook
9332 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
9333 @code{symbol_ref}, and then check for it here. @xref{Assembler
9337 @findex dbr_sequence_length
9338 @defmac DBR_OUTPUT_SEQEND (@var{file})
9339 A C statement, to be executed after all slot-filler instructions have
9340 been output. If necessary, call @code{dbr_sequence_length} to
9341 determine the number of slots filled in a sequence (zero if not
9342 currently outputting a sequence), to decide how many no-ops to output,
9345 Don't define this macro if it has nothing to do, but it is helpful in
9346 reading assembly output if the extent of the delay sequence is made
9347 explicit (e.g.@: with white space).
9350 @findex final_sequence
9351 Note that output routines for instructions with delay slots must be
9352 prepared to deal with not being output as part of a sequence
9353 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
9354 found.) The variable @code{final_sequence} is null when not
9355 processing a sequence, otherwise it contains the @code{sequence} rtx
9359 @defmac REGISTER_PREFIX
9360 @defmacx LOCAL_LABEL_PREFIX
9361 @defmacx USER_LABEL_PREFIX
9362 @defmacx IMMEDIATE_PREFIX
9363 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
9364 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
9365 @file{final.c}). These are useful when a single @file{md} file must
9366 support multiple assembler formats. In that case, the various @file{tm.h}
9367 files can define these macros differently.
9370 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
9371 If defined this macro should expand to a series of @code{case}
9372 statements which will be parsed inside the @code{switch} statement of
9373 the @code{asm_fprintf} function. This allows targets to define extra
9374 printf formats which may useful when generating their assembler
9375 statements. Note that uppercase letters are reserved for future
9376 generic extensions to asm_fprintf, and so are not available to target
9377 specific code. The output file is given by the parameter @var{file}.
9378 The varargs input pointer is @var{argptr} and the rest of the format
9379 string, starting the character after the one that is being switched
9380 upon, is pointed to by @var{format}.
9383 @defmac ASSEMBLER_DIALECT
9384 If your target supports multiple dialects of assembler language (such as
9385 different opcodes), define this macro as a C expression that gives the
9386 numeric index of the assembler language dialect to use, with zero as the
9389 If this macro is defined, you may use constructs of the form
9391 @samp{@{option0|option1|option2@dots{}@}}
9394 in the output templates of patterns (@pxref{Output Template}) or in the
9395 first argument of @code{asm_fprintf}. This construct outputs
9396 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
9397 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
9398 within these strings retain their usual meaning. If there are fewer
9399 alternatives within the braces than the value of
9400 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
9401 to print curly braces or @samp{|} character in assembler output directly,
9402 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
9404 If you do not define this macro, the characters @samp{@{}, @samp{|} and
9405 @samp{@}} do not have any special meaning when used in templates or
9406 operands to @code{asm_fprintf}.
9408 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
9409 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
9410 the variations in assembler language syntax with that mechanism. Define
9411 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
9412 if the syntax variant are larger and involve such things as different
9413 opcodes or operand order.
9416 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
9417 A C expression to output to @var{stream} some assembler code
9418 which will push hard register number @var{regno} onto the stack.
9419 The code need not be optimal, since this macro is used only when
9423 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
9424 A C expression to output to @var{stream} some assembler code
9425 which will pop hard register number @var{regno} off of the stack.
9426 The code need not be optimal, since this macro is used only when
9430 @node Dispatch Tables
9431 @subsection Output of Dispatch Tables
9433 @c prevent bad page break with this line
9434 This concerns dispatch tables.
9436 @cindex dispatch table
9437 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
9438 A C statement to output to the stdio stream @var{stream} an assembler
9439 pseudo-instruction to generate a difference between two labels.
9440 @var{value} and @var{rel} are the numbers of two internal labels. The
9441 definitions of these labels are output using
9442 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
9443 way here. For example,
9446 fprintf (@var{stream}, "\t.word L%d-L%d\n",
9447 @var{value}, @var{rel})
9450 You must provide this macro on machines where the addresses in a
9451 dispatch table are relative to the table's own address. If defined, GCC
9452 will also use this macro on all machines when producing PIC@.
9453 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
9454 mode and flags can be read.
9457 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
9458 This macro should be provided on machines where the addresses
9459 in a dispatch table are absolute.
9461 The definition should be a C statement to output to the stdio stream
9462 @var{stream} an assembler pseudo-instruction to generate a reference to
9463 a label. @var{value} is the number of an internal label whose
9464 definition is output using @code{(*targetm.asm_out.internal_label)}.
9468 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
9472 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
9473 Define this if the label before a jump-table needs to be output
9474 specially. The first three arguments are the same as for
9475 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
9476 jump-table which follows (a @code{jump_table_data} containing an
9477 @code{addr_vec} or @code{addr_diff_vec}).
9479 This feature is used on system V to output a @code{swbeg} statement
9482 If this macro is not defined, these labels are output with
9483 @code{(*targetm.asm_out.internal_label)}.
9486 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
9487 Define this if something special must be output at the end of a
9488 jump-table. The definition should be a C statement to be executed
9489 after the assembler code for the table is written. It should write
9490 the appropriate code to stdio stream @var{stream}. The argument
9491 @var{table} is the jump-table insn, and @var{num} is the label-number
9492 of the preceding label.
9494 If this macro is not defined, nothing special is output at the end of
9498 @deftypefn {Target Hook} void TARGET_ASM_POST_CFI_STARTPROC (FILE *@var{}, @var{tree})
9499 This target hook is used to emit assembly strings required by the target
9500 after the .cfi_startproc directive. The first argument is the file stream to
9501 write the strings to and the second argument is the function's declaration. The
9502 expected use is to add more .cfi_* directives.
9504 The default is to not output any assembly strings.
9507 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
9508 This target hook emits a label at the beginning of each FDE@. It
9509 should be defined on targets where FDEs need special labels, and it
9510 should write the appropriate label, for the FDE associated with the
9511 function declaration @var{decl}, to the stdio stream @var{stream}.
9512 The third argument, @var{for_eh}, is a boolean: true if this is for an
9513 exception table. The fourth argument, @var{empty}, is a boolean:
9514 true if this is a placeholder label for an omitted FDE@.
9516 The default is that FDEs are not given nonlocal labels.
9519 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
9520 This target hook emits a label at the beginning of the exception table.
9521 It should be defined on targets where it is desirable for the table
9522 to be broken up according to function.
9524 The default is that no label is emitted.
9527 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
9528 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.
9531 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx_insn *@var{insn})
9532 This target hook emits assembly directives required to unwind the
9533 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
9534 returns @code{UI_TARGET}.
9537 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
9538 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.
9541 @node Exception Region Output
9542 @subsection Assembler Commands for Exception Regions
9544 @c prevent bad page break with this line
9546 This describes commands marking the start and the end of an exception
9549 @defmac EH_FRAME_SECTION_NAME
9550 If defined, a C string constant for the name of the section containing
9551 exception handling frame unwind information. If not defined, GCC will
9552 provide a default definition if the target supports named sections.
9553 @file{crtstuff.c} uses this macro to switch to the appropriate section.
9555 You should define this symbol if your target supports DWARF 2 frame
9556 unwind information and the default definition does not work.
9559 @defmac EH_FRAME_THROUGH_COLLECT2
9560 If defined, DWARF 2 frame unwind information will identified by
9561 specially named labels. The collect2 process will locate these
9562 labels and generate code to register the frames.
9564 This might be necessary, for instance, if the system linker will not
9565 place the eh_frames in-between the sentinals from @file{crtstuff.c},
9566 or if the system linker does garbage collection and sections cannot
9567 be marked as not to be collected.
9570 @defmac EH_TABLES_CAN_BE_READ_ONLY
9571 Define this macro to 1 if your target is such that no frame unwind
9572 information encoding used with non-PIC code will ever require a
9573 runtime relocation, but the linker may not support merging read-only
9574 and read-write sections into a single read-write section.
9577 @defmac MASK_RETURN_ADDR
9578 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
9579 that it does not contain any extraneous set bits in it.
9582 @defmac DWARF2_UNWIND_INFO
9583 Define this macro to 0 if your target supports DWARF 2 frame unwind
9584 information, but it does not yet work with exception handling.
9585 Otherwise, if your target supports this information (if it defines
9586 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
9587 GCC will provide a default definition of 1.
9590 @deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
9591 This hook defines the mechanism that will be used for exception handling
9592 by the target. If the target has ABI specified unwind tables, the hook
9593 should return @code{UI_TARGET}. If the target is to use the
9594 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
9595 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
9596 information, the hook should return @code{UI_DWARF2}.
9598 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
9599 This may end up simplifying other parts of target-specific code. The
9600 default implementation of this hook never returns @code{UI_NONE}.
9602 Note that the value returned by this hook should be constant. It should
9603 not depend on anything except the command-line switches described by
9604 @var{opts}. In particular, the
9605 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
9606 macros and builtin functions related to exception handling are set up
9607 depending on this setting.
9609 The default implementation of the hook first honors the
9610 @option{--enable-sjlj-exceptions} configure option, then
9611 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
9612 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
9613 must define this hook so that @var{opts} is used correctly.
9616 @deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
9617 This variable should be set to @code{true} if the target ABI requires unwinding
9618 tables even when exceptions are not used. It must not be modified by
9619 command-line option processing.
9622 @defmac DONT_USE_BUILTIN_SETJMP
9623 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
9624 should use the @code{setjmp}/@code{longjmp} functions from the C library
9625 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
9628 @defmac JMP_BUF_SIZE
9629 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
9630 defined. Define this macro if the default size of @code{jmp_buf} buffer
9631 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
9632 is not large enough, or if it is much too large.
9633 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
9636 @defmac DWARF_CIE_DATA_ALIGNMENT
9637 This macro need only be defined if the target might save registers in the
9638 function prologue at an offset to the stack pointer that is not aligned to
9639 @code{UNITS_PER_WORD}. The definition should be the negative minimum
9640 alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive
9641 minimum alignment otherwise. @xref{DWARF}. Only applicable if
9642 the target supports DWARF 2 frame unwind information.
9645 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
9646 Contains the value true if the target should add a zero word onto the
9647 end of a Dwarf-2 frame info section when used for exception handling.
9648 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
9652 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
9653 Given a register, this hook should return a parallel of registers to
9654 represent where to find the register pieces. Define this hook if the
9655 register and its mode are represented in Dwarf in non-contiguous
9656 locations, or if the register should be represented in more than one
9657 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
9658 If not defined, the default is to return @code{NULL_RTX}.
9661 @deftypefn {Target Hook} machine_mode TARGET_DWARF_FRAME_REG_MODE (int @var{regno})
9662 Given a register, this hook should return the mode which the
9663 corresponding Dwarf frame register should have. This is normally
9664 used to return a smaller mode than the raw mode to prevent call
9665 clobbered parts of a register altering the frame register size
9668 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
9669 If some registers are represented in Dwarf-2 unwind information in
9670 multiple pieces, define this hook to fill in information about the
9671 sizes of those pieces in the table used by the unwinder at runtime.
9672 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
9673 filling in a single size corresponding to each hard register;
9674 @var{address} is the address of the table.
9677 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
9678 This hook is used to output a reference from a frame unwinding table to
9679 the type_info object identified by @var{sym}. It should return @code{true}
9680 if the reference was output. Returning @code{false} will cause the
9681 reference to be output using the normal Dwarf2 routines.
9684 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
9685 This flag should be set to @code{true} on targets that use an ARM EABI
9686 based unwinding library, and @code{false} on other targets. This effects
9687 the format of unwinding tables, and how the unwinder in entered after
9688 running a cleanup. The default is @code{false}.
9691 @node Alignment Output
9692 @subsection Assembler Commands for Alignment
9694 @c prevent bad page break with this line
9695 This describes commands for alignment.
9697 @defmac JUMP_ALIGN (@var{label})
9698 The alignment (log base 2) to put in front of @var{label}, which is
9699 a common destination of jumps and has no fallthru incoming edge.
9701 This macro need not be defined if you don't want any special alignment
9702 to be done at such a time. Most machine descriptions do not currently
9705 Unless it's necessary to inspect the @var{label} parameter, it is better
9706 to set the variable @var{align_jumps} in the target's
9707 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9708 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
9711 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
9712 The alignment (log base 2) to put in front of @var{label}, which follows
9715 This macro need not be defined if you don't want any special alignment
9716 to be done at such a time. Most machine descriptions do not currently
9720 @defmac LOOP_ALIGN (@var{label})
9721 The alignment (log base 2) to put in front of @var{label} that heads
9722 a frequently executed basic block (usually the header of a loop).
9724 This macro need not be defined if you don't want any special alignment
9725 to be done at such a time. Most machine descriptions do not currently
9728 Unless it's necessary to inspect the @var{label} parameter, it is better
9729 to set the variable @code{align_loops} in the target's
9730 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9731 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
9734 @defmac LABEL_ALIGN (@var{label})
9735 The alignment (log base 2) to put in front of @var{label}.
9736 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
9737 the maximum of the specified values is used.
9739 Unless it's necessary to inspect the @var{label} parameter, it is better
9740 to set the variable @code{align_labels} in the target's
9741 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9742 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9745 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9746 A C statement to output to the stdio stream @var{stream} an assembler
9747 instruction to advance the location counter by @var{nbytes} bytes.
9748 Those bytes should be zero when loaded. @var{nbytes} will be a C
9749 expression of type @code{unsigned HOST_WIDE_INT}.
9752 @defmac ASM_NO_SKIP_IN_TEXT
9753 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9754 text section because it fails to put zeros in the bytes that are skipped.
9755 This is true on many Unix systems, where the pseudo--op to skip bytes
9756 produces no-op instructions rather than zeros when used in the text
9760 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9761 A C statement to output to the stdio stream @var{stream} an assembler
9762 command to advance the location counter to a multiple of 2 to the
9763 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9766 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9767 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9768 for padding, if necessary.
9771 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9772 A C statement to output to the stdio stream @var{stream} an assembler
9773 command to advance the location counter to a multiple of 2 to the
9774 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9775 satisfy the alignment request. @var{power} and @var{max_skip} will be
9776 a C expression of type @code{int}.
9780 @node Debugging Info
9781 @section Controlling Debugging Information Format
9783 @c prevent bad page break with this line
9784 This describes how to specify debugging information.
9787 * All Debuggers:: Macros that affect all debugging formats uniformly.
9788 * DBX Options:: Macros enabling specific options in DBX format.
9789 * DBX Hooks:: Hook macros for varying DBX format.
9790 * File Names and DBX:: Macros controlling output of file names in DBX format.
9791 * DWARF:: Macros for DWARF format.
9792 * VMS Debug:: Macros for VMS debug format.
9796 @subsection Macros Affecting All Debugging Formats
9798 @c prevent bad page break with this line
9799 These macros affect all debugging formats.
9801 @defmac DBX_REGISTER_NUMBER (@var{regno})
9802 A C expression that returns the DBX register number for the compiler
9803 register number @var{regno}. In the default macro provided, the value
9804 of this expression will be @var{regno} itself. But sometimes there are
9805 some registers that the compiler knows about and DBX does not, or vice
9806 versa. In such cases, some register may need to have one number in the
9807 compiler and another for DBX@.
9809 If two registers have consecutive numbers inside GCC, and they can be
9810 used as a pair to hold a multiword value, then they @emph{must} have
9811 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9812 Otherwise, debuggers will be unable to access such a pair, because they
9813 expect register pairs to be consecutive in their own numbering scheme.
9815 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9816 does not preserve register pairs, then what you must do instead is
9817 redefine the actual register numbering scheme.
9820 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9821 A C expression that returns the integer offset value for an automatic
9822 variable having address @var{x} (an RTL expression). The default
9823 computation assumes that @var{x} is based on the frame-pointer and
9824 gives the offset from the frame-pointer. This is required for targets
9825 that produce debugging output for DBX and allow the frame-pointer to be
9826 eliminated when the @option{-g} option is used.
9829 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9830 A C expression that returns the integer offset value for an argument
9831 having address @var{x} (an RTL expression). The nominal offset is
9835 @defmac PREFERRED_DEBUGGING_TYPE
9836 A C expression that returns the type of debugging output GCC should
9837 produce when the user specifies just @option{-g}. Define
9838 this if you have arranged for GCC to support more than one format of
9839 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9840 @code{DWARF2_DEBUG}, @code{XCOFF_DEBUG}, @code{VMS_DEBUG},
9841 and @code{VMS_AND_DWARF2_DEBUG}.
9843 When the user specifies @option{-ggdb}, GCC normally also uses the
9844 value of this macro to select the debugging output format, but with two
9845 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9846 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9847 defined, GCC uses @code{DBX_DEBUG}.
9849 The value of this macro only affects the default debugging output; the
9850 user can always get a specific type of output by using @option{-gstabs},
9851 @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9855 @subsection Specific Options for DBX Output
9857 @c prevent bad page break with this line
9858 These are specific options for DBX output.
9860 @defmac DBX_DEBUGGING_INFO
9861 Define this macro if GCC should produce debugging output for DBX
9862 in response to the @option{-g} option.
9865 @defmac XCOFF_DEBUGGING_INFO
9866 Define this macro if GCC should produce XCOFF format debugging output
9867 in response to the @option{-g} option. This is a variant of DBX format.
9870 @defmac DEFAULT_GDB_EXTENSIONS
9871 Define this macro to control whether GCC should by default generate
9872 GDB's extended version of DBX debugging information (assuming DBX-format
9873 debugging information is enabled at all). If you don't define the
9874 macro, the default is 1: always generate the extended information
9875 if there is any occasion to.
9878 @defmac DEBUG_SYMS_TEXT
9879 Define this macro if all @code{.stabs} commands should be output while
9880 in the text section.
9883 @defmac ASM_STABS_OP
9884 A C string constant, including spacing, naming the assembler pseudo op to
9885 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9886 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9887 applies only to DBX debugging information format.
9890 @defmac ASM_STABD_OP
9891 A C string constant, including spacing, naming the assembler pseudo op to
9892 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9893 value is the current location. If you don't define this macro,
9894 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9898 @defmac ASM_STABN_OP
9899 A C string constant, including spacing, naming the assembler pseudo op to
9900 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9901 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9902 macro applies only to DBX debugging information format.
9905 @defmac DBX_NO_XREFS
9906 Define this macro if DBX on your system does not support the construct
9907 @samp{xs@var{tagname}}. On some systems, this construct is used to
9908 describe a forward reference to a structure named @var{tagname}.
9909 On other systems, this construct is not supported at all.
9912 @defmac DBX_CONTIN_LENGTH
9913 A symbol name in DBX-format debugging information is normally
9914 continued (split into two separate @code{.stabs} directives) when it
9915 exceeds a certain length (by default, 80 characters). On some
9916 operating systems, DBX requires this splitting; on others, splitting
9917 must not be done. You can inhibit splitting by defining this macro
9918 with the value zero. You can override the default splitting-length by
9919 defining this macro as an expression for the length you desire.
9922 @defmac DBX_CONTIN_CHAR
9923 Normally continuation is indicated by adding a @samp{\} character to
9924 the end of a @code{.stabs} string when a continuation follows. To use
9925 a different character instead, define this macro as a character
9926 constant for the character you want to use. Do not define this macro
9927 if backslash is correct for your system.
9930 @defmac DBX_STATIC_STAB_DATA_SECTION
9931 Define this macro if it is necessary to go to the data section before
9932 outputting the @samp{.stabs} pseudo-op for a non-global static
9936 @defmac DBX_TYPE_DECL_STABS_CODE
9937 The value to use in the ``code'' field of the @code{.stabs} directive
9938 for a typedef. The default is @code{N_LSYM}.
9941 @defmac DBX_STATIC_CONST_VAR_CODE
9942 The value to use in the ``code'' field of the @code{.stabs} directive
9943 for a static variable located in the text section. DBX format does not
9944 provide any ``right'' way to do this. The default is @code{N_FUN}.
9947 @defmac DBX_REGPARM_STABS_CODE
9948 The value to use in the ``code'' field of the @code{.stabs} directive
9949 for a parameter passed in registers. DBX format does not provide any
9950 ``right'' way to do this. The default is @code{N_RSYM}.
9953 @defmac DBX_REGPARM_STABS_LETTER
9954 The letter to use in DBX symbol data to identify a symbol as a parameter
9955 passed in registers. DBX format does not customarily provide any way to
9956 do this. The default is @code{'P'}.
9959 @defmac DBX_FUNCTION_FIRST
9960 Define this macro if the DBX information for a function and its
9961 arguments should precede the assembler code for the function. Normally,
9962 in DBX format, the debugging information entirely follows the assembler
9966 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9967 Define this macro, with value 1, if the value of a symbol describing
9968 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9969 relative to the start of the enclosing function. Normally, GCC uses
9970 an absolute address.
9973 @defmac DBX_LINES_FUNCTION_RELATIVE
9974 Define this macro, with value 1, if the value of a symbol indicating
9975 the current line number (@code{N_SLINE}) should be relative to the
9976 start of the enclosing function. Normally, GCC uses an absolute address.
9979 @defmac DBX_USE_BINCL
9980 Define this macro if GCC should generate @code{N_BINCL} and
9981 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9982 macro also directs GCC to output a type number as a pair of a file
9983 number and a type number within the file. Normally, GCC does not
9984 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9985 number for a type number.
9989 @subsection Open-Ended Hooks for DBX Format
9991 @c prevent bad page break with this line
9992 These are hooks for DBX format.
9994 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9995 A C statement to output DBX debugging information before code for line
9996 number @var{line} of the current source file to the stdio stream
9997 @var{stream}. @var{counter} is the number of time the macro was
9998 invoked, including the current invocation; it is intended to generate
9999 unique labels in the assembly output.
10001 This macro should not be defined if the default output is correct, or
10002 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
10005 @defmac NO_DBX_FUNCTION_END
10006 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
10007 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
10008 On those machines, define this macro to turn this feature off without
10009 disturbing the rest of the gdb extensions.
10012 @defmac NO_DBX_BNSYM_ENSYM
10013 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
10014 extension construct. On those machines, define this macro to turn this
10015 feature off without disturbing the rest of the gdb extensions.
10018 @node File Names and DBX
10019 @subsection File Names in DBX Format
10021 @c prevent bad page break with this line
10022 This describes file names in DBX format.
10024 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
10025 A C statement to output DBX debugging information to the stdio stream
10026 @var{stream}, which indicates that file @var{name} is the main source
10027 file---the file specified as the input file for compilation.
10028 This macro is called only once, at the beginning of compilation.
10030 This macro need not be defined if the standard form of output
10031 for DBX debugging information is appropriate.
10033 It may be necessary to refer to a label equal to the beginning of the
10034 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
10035 to do so. If you do this, you must also set the variable
10036 @var{used_ltext_label_name} to @code{true}.
10039 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
10040 Define this macro, with value 1, if GCC should not emit an indication
10041 of the current directory for compilation and current source language at
10042 the beginning of the file.
10045 @defmac NO_DBX_GCC_MARKER
10046 Define this macro, with value 1, if GCC should not emit an indication
10047 that this object file was compiled by GCC@. The default is to emit
10048 an @code{N_OPT} stab at the beginning of every source file, with
10049 @samp{gcc2_compiled.} for the string and value 0.
10052 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
10053 A C statement to output DBX debugging information at the end of
10054 compilation of the main source file @var{name}. Output should be
10055 written to the stdio stream @var{stream}.
10057 If you don't define this macro, nothing special is output at the end
10058 of compilation, which is correct for most machines.
10061 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
10062 Define this macro @emph{instead of} defining
10063 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
10064 the end of compilation is an @code{N_SO} stab with an empty string,
10065 whose value is the highest absolute text address in the file.
10070 @subsection Macros for DWARF Output
10072 @c prevent bad page break with this line
10073 Here are macros for DWARF output.
10075 @defmac DWARF2_DEBUGGING_INFO
10076 Define this macro if GCC should produce dwarf version 2 format
10077 debugging output in response to the @option{-g} option.
10079 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
10080 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
10081 be emitted for each function. Instead of an integer return the enum
10082 value for the @code{DW_CC_} tag.
10085 To support optional call frame debugging information, you must also
10086 define @code{INCOMING_RETURN_ADDR_RTX} and either set
10087 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
10088 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
10089 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
10092 @defmac DWARF2_FRAME_INFO
10093 Define this macro to a nonzero value if GCC should always output
10094 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
10095 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
10096 exceptions are enabled, GCC will output this information not matter
10097 how you define @code{DWARF2_FRAME_INFO}.
10100 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
10101 This hook defines the mechanism that will be used for describing frame
10102 unwind information to the debugger. Normally the hook will return
10103 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
10104 return @code{UI_NONE} otherwise.
10106 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
10107 is disabled in order to always output DWARF 2 frame information.
10109 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
10110 This will suppress generation of the normal debug frame unwind information.
10113 @defmac DWARF2_ASM_LINE_DEBUG_INFO
10114 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
10115 line debug info sections. This will result in much more compact line number
10116 tables, and hence is desirable if it works.
10119 @defmac DWARF2_ASM_VIEW_DEBUG_INFO
10120 Define this macro to be a nonzero value if the assembler supports view
10121 assignment and verification in @code{.loc}. If it does not, but the
10122 user enables location views, the compiler may have to fallback to
10123 internal line number tables.
10126 @deftypefn {Target Hook} int TARGET_RESET_LOCATION_VIEW (rtx_insn *@var{})
10127 This hook, if defined, enables -ginternal-reset-location-views, and
10128 uses its result to override cases in which the estimated min insn
10129 length might be nonzero even when a PC advance (i.e., a view reset)
10130 cannot be taken for granted.
10132 If the hook is defined, it must return a positive value to indicate
10133 the insn definitely advances the PC, and so the view number can be
10134 safely assumed to be reset; a negative value to mean the insn
10135 definitely does not advance the PC, and os the view number must not
10136 be reset; or zero to decide based on the estimated insn length.
10138 If insn length is to be regarded as reliable, set the hook to
10139 @code{hook_int_rtx_insn_0}.
10142 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
10143 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.
10146 @deftypevr {Target Hook} bool TARGET_DELAY_SCHED2
10147 True if sched2 is not to be run at its normal place.
10148 This usually means it will be run as part of machine-specific reorg.
10151 @deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK
10152 True if vartrack is not to be run at its normal place.
10153 This usually means it will be run as part of machine-specific reorg.
10156 @deftypevr {Target Hook} bool TARGET_NO_REGISTER_ALLOCATION
10157 True if register allocation and the passes
10158 following it should not be run. Usually true only for virtual assembler
10162 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
10163 A C statement to issue assembly directives that create a difference
10164 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
10167 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
10168 A C statement to issue assembly directives that create a difference
10169 between the two given labels in system defined units, e.g.@: instruction
10170 slots on IA64 VMS, using an integer of the given size.
10173 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{offset}, @var{section})
10174 A C statement to issue assembly directives that create a
10175 section-relative reference to the given @var{label} plus @var{offset}, using
10176 an integer of the given @var{size}. The label is known to be defined in the
10177 given @var{section}.
10180 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
10181 A C statement to issue assembly directives that create a self-relative
10182 reference to the given @var{label}, using an integer of the given @var{size}.
10185 @defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label})
10186 A C statement to issue assembly directives that create a reference to the
10187 given @var{label} relative to the dbase, using an integer of the given @var{size}.
10190 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
10191 A C statement to issue assembly directives that create a reference to
10192 the DWARF table identifier @var{label} from the current section. This
10193 is used on some systems to avoid garbage collecting a DWARF table which
10194 is referenced by a function.
10197 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
10198 If defined, this target hook is a function which outputs a DTP-relative
10199 reference to the given TLS symbol of the specified size.
10204 @subsection Macros for VMS Debug Format
10206 @c prevent bad page break with this line
10207 Here are macros for VMS debug format.
10209 @defmac VMS_DEBUGGING_INFO
10210 Define this macro if GCC should produce debugging output for VMS
10211 in response to the @option{-g} option. The default behavior for VMS
10212 is to generate minimal debug info for a traceback in the absence of
10213 @option{-g} unless explicitly overridden with @option{-g0}. This
10214 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
10215 @code{TARGET_OPTION_OVERRIDE}.
10218 @node Floating Point
10219 @section Cross Compilation and Floating Point
10220 @cindex cross compilation and floating point
10221 @cindex floating point and cross compilation
10223 While all modern machines use twos-complement representation for integers,
10224 there are a variety of representations for floating point numbers. This
10225 means that in a cross-compiler the representation of floating point numbers
10226 in the compiled program may be different from that used in the machine
10227 doing the compilation.
10229 Because different representation systems may offer different amounts of
10230 range and precision, all floating point constants must be represented in
10231 the target machine's format. Therefore, the cross compiler cannot
10232 safely use the host machine's floating point arithmetic; it must emulate
10233 the target's arithmetic. To ensure consistency, GCC always uses
10234 emulation to work with floating point values, even when the host and
10235 target floating point formats are identical.
10237 The following macros are provided by @file{real.h} for the compiler to
10238 use. All parts of the compiler which generate or optimize
10239 floating-point calculations must use these macros. They may evaluate
10240 their operands more than once, so operands must not have side effects.
10242 @defmac REAL_VALUE_TYPE
10243 The C data type to be used to hold a floating point value in the target
10244 machine's format. Typically this is a @code{struct} containing an
10245 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
10249 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
10250 Truncates @var{x} to a signed integer, rounding toward zero.
10253 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
10254 Truncates @var{x} to an unsigned integer, rounding toward zero. If
10255 @var{x} is negative, returns zero.
10258 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode})
10259 Converts @var{string} into a floating point number in the target machine's
10260 representation for mode @var{mode}. This routine can handle both
10261 decimal and hexadecimal floating point constants, using the syntax
10262 defined by the C language for both.
10265 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
10266 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
10269 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
10270 Determines whether @var{x} represents infinity (positive or negative).
10273 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
10274 Determines whether @var{x} represents a ``NaN'' (not-a-number).
10277 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
10278 Returns the negative of the floating point value @var{x}.
10281 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
10282 Returns the absolute value of @var{x}.
10285 @node Mode Switching
10286 @section Mode Switching Instructions
10287 @cindex mode switching
10288 The following macros control mode switching optimizations:
10290 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
10291 Define this macro if the port needs extra instructions inserted for mode
10292 switching in an optimizing compilation.
10294 For an example, the SH4 can perform both single and double precision
10295 floating point operations, but to perform a single precision operation,
10296 the FPSCR PR bit has to be cleared, while for a double precision
10297 operation, this bit has to be set. Changing the PR bit requires a general
10298 purpose register as a scratch register, hence these FPSCR sets have to
10299 be inserted before reload, i.e.@: you cannot put this into instruction emitting
10300 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
10302 You can have multiple entities that are mode-switched, and select at run time
10303 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
10304 return nonzero for any @var{entity} that needs mode-switching.
10305 If you define this macro, you also have to define
10306 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
10307 @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
10308 @code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
10312 @defmac NUM_MODES_FOR_MODE_SWITCHING
10313 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
10314 initializer for an array of integers. Each initializer element
10315 N refers to an entity that needs mode switching, and specifies the number
10316 of different modes that might need to be set for this entity.
10317 The position of the initializer in the initializer---starting counting at
10318 zero---determines the integer that is used to refer to the mode-switched
10319 entity in question.
10320 In macros that take mode arguments / yield a mode result, modes are
10321 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
10322 switch is needed / supplied.
10325 @deftypefn {Target Hook} void TARGET_MODE_EMIT (int @var{entity}, int @var{mode}, int @var{prev_mode}, HARD_REG_SET @var{regs_live})
10326 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.
10329 @deftypefn {Target Hook} int TARGET_MODE_NEEDED (int @var{entity}, rtx_insn *@var{insn})
10330 @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}.
10333 @deftypefn {Target Hook} int TARGET_MODE_AFTER (int @var{entity}, int @var{mode}, rtx_insn *@var{insn})
10334 @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).
10337 @deftypefn {Target Hook} int TARGET_MODE_ENTRY (int @var{entity})
10338 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.
10341 @deftypefn {Target Hook} int TARGET_MODE_EXIT (int @var{entity})
10342 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.
10345 @deftypefn {Target Hook} int TARGET_MODE_PRIORITY (int @var{entity}, int @var{n})
10346 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}.
10349 @node Target Attributes
10350 @section Defining target-specific uses of @code{__attribute__}
10351 @cindex target attributes
10352 @cindex machine attributes
10353 @cindex attributes, target-specific
10355 Target-specific attributes may be defined for functions, data and types.
10356 These are described using the following target hooks; they also need to
10357 be documented in @file{extend.texi}.
10359 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
10360 If defined, this target hook points to an array of @samp{struct
10361 attribute_spec} (defined in @file{tree-core.h}) specifying the machine
10362 specific attributes for this target and some of the restrictions on the
10363 entities to which these attributes are applied and the arguments they
10367 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
10368 If defined, this target hook is a function which returns true if the
10369 machine-specific attribute named @var{name} expects an identifier
10370 given as its first argument to be passed on as a plain identifier, not
10371 subjected to name lookup. If this is not defined, the default is
10372 false for all machine-specific attributes.
10375 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
10376 If defined, this target hook is a function which returns zero if the attributes on
10377 @var{type1} and @var{type2} are incompatible, one if they are compatible,
10378 and two if they are nearly compatible (which causes a warning to be
10379 generated). If this is not defined, machine-specific attributes are
10380 supposed always to be compatible.
10383 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
10384 If defined, this target hook is a function which assigns default attributes to
10385 the newly defined @var{type}.
10388 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
10389 Define this target hook if the merging of type attributes needs special
10390 handling. If defined, the result is a list of the combined
10391 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
10392 that @code{comptypes} has already been called and returned 1. This
10393 function may call @code{merge_attributes} to handle machine-independent
10397 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
10398 Define this target hook if the merging of decl attributes needs special
10399 handling. If defined, the result is a list of the combined
10400 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
10401 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
10402 when this is needed are when one attribute overrides another, or when an
10403 attribute is nullified by a subsequent definition. This function may
10404 call @code{merge_attributes} to handle machine-independent merging.
10406 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
10407 If the only target-specific handling you require is @samp{dllimport}
10408 for Microsoft Windows targets, you should define the macro
10409 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
10410 will then define a function called
10411 @code{merge_dllimport_decl_attributes} which can then be defined as
10412 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
10413 add @code{handle_dll_attribute} in the attribute table for your port
10414 to perform initial processing of the @samp{dllimport} and
10415 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
10416 @file{i386/i386.c}, for example.
10419 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
10420 @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}.
10423 @defmac TARGET_DECLSPEC
10424 Define this macro to a nonzero value if you want to treat
10425 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
10426 default, this behavior is enabled only for targets that define
10427 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
10428 of @code{__declspec} is via a built-in macro, but you should not rely
10429 on this implementation detail.
10432 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
10433 Define this target hook if you want to be able to add attributes to a decl
10434 when it is being created. This is normally useful for back ends which
10435 wish to implement a pragma by using the attributes which correspond to
10436 the pragma's effect. The @var{node} argument is the decl which is being
10437 created. The @var{attr_ptr} argument is a pointer to the attribute list
10438 for this decl. The list itself should not be modified, since it may be
10439 shared with other decls, but attributes may be chained on the head of
10440 the list and @code{*@var{attr_ptr}} modified to point to the new
10441 attributes, or a copy of the list may be made if further changes are
10445 @deftypefn {Target Hook} tree TARGET_HANDLE_GENERIC_ATTRIBUTE (tree *@var{node}, tree @var{name}, tree @var{args}, int @var{flags}, bool *@var{no_add_attrs})
10446 Define this target hook if you want to be able to perform additional
10447 target-specific processing of an attribute which is handled generically
10448 by a front end. The arguments are the same as those which are passed to
10449 attribute handlers. So far this only affects the @var{noinit} and
10450 @var{section} attribute.
10453 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
10455 This target hook returns @code{true} if it is OK to inline @var{fndecl}
10456 into the current function, despite its having target-specific
10457 attributes, @code{false} otherwise. By default, if a function has a
10458 target specific attribute attached to it, it will not be inlined.
10461 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
10462 This hook is called to parse @code{attribute(target("..."))}, which
10463 allows setting target-specific options on individual functions.
10464 These function-specific options may differ
10465 from the options specified on the command line. The hook should return
10466 @code{true} if the options are valid.
10468 The hook should set the @code{DECL_FUNCTION_SPECIFIC_TARGET} field in
10469 the function declaration to hold a pointer to a target-specific
10470 @code{struct cl_target_option} structure.
10473 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr}, struct gcc_options *@var{opts})
10474 This hook is called to save any additional target-specific information
10475 in the @code{struct cl_target_option} structure for function-specific
10476 options from the @code{struct gcc_options} structure.
10477 @xref{Option file format}.
10480 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct gcc_options *@var{opts}, struct cl_target_option *@var{ptr})
10481 This hook is called to restore any additional target-specific
10482 information in the @code{struct cl_target_option} structure for
10483 function-specific options to the @code{struct gcc_options} structure.
10486 @deftypefn {Target Hook} void TARGET_OPTION_POST_STREAM_IN (struct cl_target_option *@var{ptr})
10487 This hook is called to update target-specific information in the
10488 @code{struct cl_target_option} structure after it is streamed in from
10492 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
10493 This hook is called to print any additional target-specific
10494 information in the @code{struct cl_target_option} structure for
10495 function-specific options.
10498 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
10499 This target hook parses the options for @code{#pragma GCC target}, which
10500 sets the target-specific options for functions that occur later in the
10501 input stream. The options accepted should be the same as those handled by the
10502 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
10505 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
10506 Sometimes certain combinations of command options do not make sense on
10507 a particular target machine. You can override the hook
10508 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
10509 once just after all the command options have been parsed.
10511 Don't use this hook to turn on various extra optimizations for
10512 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
10514 If you need to do something whenever the optimization level is
10515 changed via the optimize attribute or pragma, see
10516 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
10519 @deftypefn {Target Hook} bool TARGET_OPTION_FUNCTION_VERSIONS (tree @var{decl1}, tree @var{decl2})
10520 This target hook returns @code{true} if @var{DECL1} and @var{DECL2} are
10521 versions of the same function. @var{DECL1} and @var{DECL2} are function
10522 versions if and only if they have the same function signature and
10523 different target specific attributes, that is, they are compiled for
10524 different target machines.
10527 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
10528 This target hook returns @code{false} if the @var{caller} function
10529 cannot inline @var{callee}, based on target specific information. By
10530 default, inlining is not allowed if the callee function has function
10531 specific target options and the caller does not use the same options.
10534 @deftypefn {Target Hook} void TARGET_RELAYOUT_FUNCTION (tree @var{fndecl})
10535 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.
10539 @section Emulating TLS
10540 @cindex Emulated TLS
10542 For targets whose psABI does not provide Thread Local Storage via
10543 specific relocations and instruction sequences, an emulation layer is
10544 used. A set of target hooks allows this emulation layer to be
10545 configured for the requirements of a particular target. For instance
10546 the psABI may in fact specify TLS support in terms of an emulation
10549 The emulation layer works by creating a control object for every TLS
10550 object. To access the TLS object, a lookup function is provided
10551 which, when given the address of the control object, will return the
10552 address of the current thread's instance of the TLS object.
10554 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
10555 Contains the name of the helper function that uses a TLS control
10556 object to locate a TLS instance. The default causes libgcc's
10557 emulated TLS helper function to be used.
10560 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
10561 Contains the name of the helper function that should be used at
10562 program startup to register TLS objects that are implicitly
10563 initialized to zero. If this is @code{NULL}, all TLS objects will
10564 have explicit initializers. The default causes libgcc's emulated TLS
10565 registration function to be used.
10568 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
10569 Contains the name of the section in which TLS control variables should
10570 be placed. The default of @code{NULL} allows these to be placed in
10574 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
10575 Contains the name of the section in which TLS initializers should be
10576 placed. The default of @code{NULL} allows these to be placed in any
10580 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
10581 Contains the prefix to be prepended to TLS control variable names.
10582 The default of @code{NULL} uses a target-specific prefix.
10585 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
10586 Contains the prefix to be prepended to TLS initializer objects. The
10587 default of @code{NULL} uses a target-specific prefix.
10590 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
10591 Specifies a function that generates the FIELD_DECLs for a TLS control
10592 object type. @var{type} is the RECORD_TYPE the fields are for and
10593 @var{name} should be filled with the structure tag, if the default of
10594 @code{__emutls_object} is unsuitable. The default creates a type suitable
10595 for libgcc's emulated TLS function.
10598 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
10599 Specifies a function that generates the CONSTRUCTOR to initialize a
10600 TLS control object. @var{var} is the TLS control object, @var{decl}
10601 is the TLS object and @var{tmpl_addr} is the address of the
10602 initializer. The default initializes libgcc's emulated TLS control object.
10605 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
10606 Specifies whether the alignment of TLS control variable objects is
10607 fixed and should not be increased as some backends may do to optimize
10608 single objects. The default is false.
10611 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
10612 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
10613 may be used to describe emulated TLS control objects.
10616 @node MIPS Coprocessors
10617 @section Defining coprocessor specifics for MIPS targets.
10618 @cindex MIPS coprocessor-definition macros
10620 The MIPS specification allows MIPS implementations to have as many as 4
10621 coprocessors, each with as many as 32 private registers. GCC supports
10622 accessing these registers and transferring values between the registers
10623 and memory using asm-ized variables. For example:
10626 register unsigned int cp0count asm ("c0r1");
10632 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
10633 names may be added as described below, or the default names may be
10634 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
10636 Coprocessor registers are assumed to be epilogue-used; sets to them will
10637 be preserved even if it does not appear that the register is used again
10638 later in the function.
10640 Another note: according to the MIPS spec, coprocessor 1 (if present) is
10641 the FPU@. One accesses COP1 registers through standard mips
10642 floating-point support; they are not included in this mechanism.
10645 @section Parameters for Precompiled Header Validity Checking
10646 @cindex parameters, precompiled headers
10648 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
10649 This hook returns a pointer to the data needed by
10650 @code{TARGET_PCH_VALID_P} and sets
10651 @samp{*@var{sz}} to the size of the data in bytes.
10654 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
10655 This hook checks whether the options used to create a PCH file are
10656 compatible with the current settings. It returns @code{NULL}
10657 if so and a suitable error message if not. Error messages will
10658 be presented to the user and must be localized using @samp{_(@var{msg})}.
10660 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
10661 when the PCH file was created and @var{sz} is the size of that data in bytes.
10662 It's safe to assume that the data was created by the same version of the
10663 compiler, so no format checking is needed.
10665 The default definition of @code{default_pch_valid_p} should be
10666 suitable for most targets.
10669 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
10670 If this hook is nonnull, the default implementation of
10671 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
10672 of @code{target_flags}. @var{pch_flags} specifies the value that
10673 @code{target_flags} had when the PCH file was created. The return
10674 value is the same as for @code{TARGET_PCH_VALID_P}.
10677 @deftypefn {Target Hook} void TARGET_PREPARE_PCH_SAVE (void)
10678 Called before writing out a PCH file. If the target has some
10679 garbage-collected data that needs to be in a particular state on PCH loads,
10680 it can use this hook to enforce that state. Very few targets need
10681 to do anything here.
10685 @section C++ ABI parameters
10686 @cindex parameters, c++ abi
10688 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10689 Define this hook to override the integer type used for guard variables.
10690 These are used to implement one-time construction of static objects. The
10691 default is long_long_integer_type_node.
10694 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10695 This hook determines how guard variables are used. It should return
10696 @code{false} (the default) if the first byte should be used. A return value of
10697 @code{true} indicates that only the least significant bit should be used.
10700 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10701 This hook returns the size of the cookie to use when allocating an array
10702 whose elements have the indicated @var{type}. Assumes that it is already
10703 known that a cookie is needed. The default is
10704 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10705 IA64/Generic C++ ABI@.
10708 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10709 This hook should return @code{true} if the element size should be stored in
10710 array cookies. The default is to return @code{false}.
10713 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10714 If defined by a backend this hook allows the decision made to export
10715 class @var{type} to be overruled. Upon entry @var{import_export}
10716 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10717 to be imported and 0 otherwise. This function should return the
10718 modified value and perform any other actions necessary to support the
10719 backend's targeted operating system.
10722 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10723 This hook should return @code{true} if constructors and destructors return
10724 the address of the object created/destroyed. The default is to return
10728 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10729 This hook returns true if the key method for a class (i.e., the method
10730 which, if defined in the current translation unit, causes the virtual
10731 table to be emitted) may be an inline function. Under the standard
10732 Itanium C++ ABI the key method may be an inline function so long as
10733 the function is not declared inline in the class definition. Under
10734 some variants of the ABI, an inline function can never be the key
10735 method. The default is to return @code{true}.
10738 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10739 @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}.
10742 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10743 This hook returns true (the default) if virtual tables and other
10744 similar implicit class data objects are always COMDAT if they have
10745 external linkage. If this hook returns false, then class data for
10746 classes whose virtual table will be emitted in only one translation
10747 unit will not be COMDAT.
10750 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10751 This hook returns true (the default) if the RTTI information for
10752 the basic types which is defined in the C++ runtime should always
10753 be COMDAT, false if it should not be COMDAT.
10756 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10757 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10758 should be used to register static destructors when @option{-fuse-cxa-atexit}
10759 is in effect. The default is to return false to use @code{__cxa_atexit}.
10762 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10763 This hook returns true if the target @code{atexit} function can be used
10764 in the same manner as @code{__cxa_atexit} to register C++ static
10765 destructors. This requires that @code{atexit}-registered functions in
10766 shared libraries are run in the correct order when the libraries are
10767 unloaded. The default is to return false.
10770 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10771 @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).
10774 @deftypefn {Target Hook} tree TARGET_CXX_DECL_MANGLING_CONTEXT (const_tree @var{decl})
10775 Return target-specific mangling context of @var{decl} or @code{NULL_TREE}.
10778 @node D Language and ABI
10779 @section D ABI parameters
10780 @cindex parameters, d abi
10782 @deftypefn {D Target Hook} void TARGET_D_CPU_VERSIONS (void)
10783 Declare all environmental version identifiers relating to the target CPU
10784 using the function @code{builtin_version}, which takes a string representing
10785 the name of the version. Version identifiers predefined by this hook apply
10786 to all modules that are being compiled and imported.
10789 @deftypefn {D Target Hook} void TARGET_D_OS_VERSIONS (void)
10790 Similarly to @code{TARGET_D_CPU_VERSIONS}, but is used for versions
10791 relating to the target operating system.
10794 @deftypefn {D Target Hook} unsigned TARGET_D_CRITSEC_SIZE (void)
10795 Returns the size of the data structure used by the target operating system
10796 for critical sections and monitors. For example, on Microsoft Windows this
10797 would return the @code{sizeof(CRITICAL_SECTION)}, while other platforms that
10798 implement pthreads would return @code{sizeof(pthread_mutex_t)}.
10801 @node Named Address Spaces
10802 @section Adding support for named address spaces
10803 @cindex named address spaces
10805 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10806 standards committee, @cite{Programming Languages - C - Extensions to
10807 support embedded processors}, specifies a syntax for embedded
10808 processors to specify alternate address spaces. You can configure a
10809 GCC port to support section 5.1 of the draft report to add support for
10810 address spaces other than the default address space. These address
10811 spaces are new keywords that are similar to the @code{volatile} and
10812 @code{const} type attributes.
10814 Pointers to named address spaces can have a different size than
10815 pointers to the generic address space.
10817 For example, the SPU port uses the @code{__ea} address space to refer
10818 to memory in the host processor, rather than memory local to the SPU
10819 processor. Access to memory in the @code{__ea} address space involves
10820 issuing DMA operations to move data between the host processor and the
10821 local processor memory address space. Pointers in the @code{__ea}
10822 address space are either 32 bits or 64 bits based on the
10823 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10826 Internally, address spaces are represented as a small integer in the
10827 range 0 to 15 with address space 0 being reserved for the generic
10830 To register a named address space qualifier keyword with the C front end,
10831 the target may call the @code{c_register_addr_space} routine. For example,
10832 the SPU port uses the following to declare @code{__ea} as the keyword for
10833 named address space #1:
10835 #define ADDR_SPACE_EA 1
10836 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10839 @deftypefn {Target Hook} scalar_int_mode TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10840 Define this to return the machine mode to use for pointers to
10841 @var{address_space} if the target supports named address spaces.
10842 The default version of this hook returns @code{ptr_mode}.
10845 @deftypefn {Target Hook} scalar_int_mode TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10846 Define this to return the machine mode to use for addresses in
10847 @var{address_space} if the target supports named address spaces.
10848 The default version of this hook returns @code{Pmode}.
10851 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (scalar_int_mode @var{mode}, addr_space_t @var{as})
10852 Define this to return nonzero if the port can handle pointers
10853 with machine mode @var{mode} to address space @var{as}. This target
10854 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10855 except that it includes explicit named address space support. The default
10856 version of this hook returns true for the modes returned by either the
10857 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10858 target hooks for the given address space.
10861 @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})
10862 Define this to return true if @var{exp} is a valid address for mode
10863 @var{mode} in the named address space @var{as}. The @var{strict}
10864 parameter says whether strict addressing is in effect after reload has
10865 finished. This target hook is the same as the
10866 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10867 explicit named address space support.
10870 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode}, addr_space_t @var{as})
10871 Define this to modify an invalid address @var{x} to be a valid address
10872 with mode @var{mode} in the named address space @var{as}. This target
10873 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10874 except that it includes explicit named address space support.
10877 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{subset}, addr_space_t @var{superset})
10878 Define this to return whether the @var{subset} named address space is
10879 contained within the @var{superset} named address space. Pointers to
10880 a named address space that is a subset of another named address space
10881 will be converted automatically without a cast if used together in
10882 arithmetic operations. Pointers to a superset address space can be
10883 converted to pointers to a subset address space via explicit casts.
10886 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID (addr_space_t @var{as})
10887 Define this to modify the default handling of address 0 for the
10888 address space. Return true if 0 should be considered a valid address.
10891 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10892 Define this to convert the pointer expression represented by the RTL
10893 @var{op} with type @var{from_type} that points to a named address
10894 space to a new pointer expression with type @var{to_type} that points
10895 to a different named address space. When this hook it called, it is
10896 guaranteed that one of the two address spaces is a subset of the other,
10897 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10900 @deftypefn {Target Hook} int TARGET_ADDR_SPACE_DEBUG (addr_space_t @var{as})
10901 Define this to define how the address space is encoded in dwarf.
10902 The result is the value to be used with @code{DW_AT_address_class}.
10905 @deftypefn {Target Hook} void TARGET_ADDR_SPACE_DIAGNOSE_USAGE (addr_space_t @var{as}, location_t @var{loc})
10906 Define this hook if the availability of an address space depends on
10907 command line options and some diagnostics should be printed when the
10908 address space is used. This hook is called during parsing and allows
10909 to emit a better diagnostic compared to the case where the address space
10910 was not registered with @code{c_register_addr_space}. @var{as} is
10911 the address space as registered with @code{c_register_addr_space}.
10912 @var{loc} is the location of the address space qualifier token.
10913 The default implementation does nothing.
10917 @section Miscellaneous Parameters
10918 @cindex parameters, miscellaneous
10920 @c prevent bad page break with this line
10921 Here are several miscellaneous parameters.
10923 @defmac HAS_LONG_COND_BRANCH
10924 Define this boolean macro to indicate whether or not your architecture
10925 has conditional branches that can span all of memory. It is used in
10926 conjunction with an optimization that partitions hot and cold basic
10927 blocks into separate sections of the executable. If this macro is
10928 set to false, gcc will convert any conditional branches that attempt
10929 to cross between sections into unconditional branches or indirect jumps.
10932 @defmac HAS_LONG_UNCOND_BRANCH
10933 Define this boolean macro to indicate whether or not your architecture
10934 has unconditional branches that can span all of memory. It is used in
10935 conjunction with an optimization that partitions hot and cold basic
10936 blocks into separate sections of the executable. If this macro is
10937 set to false, gcc will convert any unconditional branches that attempt
10938 to cross between sections into indirect jumps.
10941 @defmac CASE_VECTOR_MODE
10942 An alias for a machine mode name. This is the machine mode that
10943 elements of a jump-table should have.
10946 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10947 Optional: return the preferred mode for an @code{addr_diff_vec}
10948 when the minimum and maximum offset are known. If you define this,
10949 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10950 To make this work, you also have to define @code{INSN_ALIGN} and
10951 make the alignment for @code{addr_diff_vec} explicit.
10952 The @var{body} argument is provided so that the offset_unsigned and scale
10953 flags can be updated.
10956 @defmac CASE_VECTOR_PC_RELATIVE
10957 Define this macro to be a C expression to indicate when jump-tables
10958 should contain relative addresses. You need not define this macro if
10959 jump-tables never contain relative addresses, or jump-tables should
10960 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10964 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10965 This function return the smallest number of different values for which it
10966 is best to use a jump-table instead of a tree of conditional branches.
10967 The default is four for machines with a @code{casesi} instruction and
10968 five otherwise. This is best for most machines.
10971 @defmac WORD_REGISTER_OPERATIONS
10972 Define this macro to 1 if operations between registers with integral mode
10973 smaller than a word are always performed on the entire register. To be
10974 more explicit, if you start with a pair of @code{word_mode} registers with
10975 known values and you do a subword, for example @code{QImode}, addition on
10976 the low part of the registers, then the compiler may consider that the
10977 result has a known value in @code{word_mode} too if the macro is defined
10978 to 1. Most RISC machines have this property and most CISC machines do not.
10981 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_ARITHMETIC_PRECISION (void)
10982 On some RISC architectures with 64-bit registers, the processor also
10983 maintains 32-bit condition codes that make it possible to do real 32-bit
10984 arithmetic, although the operations are performed on the full registers.
10986 On such architectures, defining this hook to 32 tells the compiler to try
10987 using 32-bit arithmetical operations setting the condition codes instead
10988 of doing full 64-bit arithmetic.
10990 More generally, define this hook on RISC architectures if you want the
10991 compiler to try using arithmetical operations setting the condition codes
10992 with a precision lower than the word precision.
10994 You need not define this hook if @code{WORD_REGISTER_OPERATIONS} is not
10998 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10999 Define this macro to be a C expression indicating when insns that read
11000 memory in @var{mem_mode}, an integral mode narrower than a word, set the
11001 bits outside of @var{mem_mode} to be either the sign-extension or the
11002 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
11003 of @var{mem_mode} for which the
11004 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
11005 @code{UNKNOWN} for other modes.
11007 This macro is not called with @var{mem_mode} non-integral or with a width
11008 greater than or equal to @code{BITS_PER_WORD}, so you may return any
11009 value in this case. Do not define this macro if it would always return
11010 @code{UNKNOWN}. On machines where this macro is defined, you will normally
11011 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
11013 You may return a non-@code{UNKNOWN} value even if for some hard registers
11014 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
11015 of these hard registers @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false
11016 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
11017 integral mode larger than this but not larger than @code{word_mode}.
11019 You must return @code{UNKNOWN} if for some hard registers that allow this
11020 mode, @code{TARGET_CAN_CHANGE_MODE_CLASS} says that they cannot change to
11021 @code{word_mode}, but that they can change to another integral mode that
11022 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
11025 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
11026 Define this macro to 1 if loading short immediate values into registers sign
11030 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (machine_mode @var{mode})
11031 When @option{-ffast-math} is in effect, GCC tries to optimize
11032 divisions by the same divisor, by turning them into multiplications by
11033 the reciprocal. This target hook specifies the minimum number of divisions
11034 that should be there for GCC to perform the optimization for a variable
11035 of mode @var{mode}. The default implementation returns 3 if the machine
11036 has an instruction for the division, and 2 if it does not.
11040 The maximum number of bytes that a single instruction can move quickly
11041 between memory and registers or between two memory locations.
11044 @defmac MAX_MOVE_MAX
11045 The maximum number of bytes that a single instruction can move quickly
11046 between memory and registers or between two memory locations. If this
11047 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
11048 constant value that is the largest value that @code{MOVE_MAX} can have
11052 @defmac SHIFT_COUNT_TRUNCATED
11053 A C expression that is nonzero if on this machine the number of bits
11054 actually used for the count of a shift operation is equal to the number
11055 of bits needed to represent the size of the object being shifted. When
11056 this macro is nonzero, the compiler will assume that it is safe to omit
11057 a sign-extend, zero-extend, and certain bitwise `and' instructions that
11058 truncates the count of a shift operation. On machines that have
11059 instructions that act on bit-fields at variable positions, which may
11060 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
11061 also enables deletion of truncations of the values that serve as
11062 arguments to bit-field instructions.
11064 If both types of instructions truncate the count (for shifts) and
11065 position (for bit-field operations), or if no variable-position bit-field
11066 instructions exist, you should define this macro.
11068 However, on some machines, such as the 80386 and the 680x0, truncation
11069 only applies to shift operations and not the (real or pretended)
11070 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
11071 such machines. Instead, add patterns to the @file{md} file that include
11072 the implied truncation of the shift instructions.
11074 You need not define this macro if it would always have the value of zero.
11077 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
11078 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (machine_mode @var{mode})
11079 This function describes how the standard shift patterns for @var{mode}
11080 deal with shifts by negative amounts or by more than the width of the mode.
11081 @xref{shift patterns}.
11083 On many machines, the shift patterns will apply a mask @var{m} to the
11084 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
11085 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
11086 this is true for mode @var{mode}, the function should return @var{m},
11087 otherwise it should return 0. A return value of 0 indicates that no
11088 particular behavior is guaranteed.
11090 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
11091 @emph{not} apply to general shift rtxes; it applies only to instructions
11092 that are generated by the named shift patterns.
11094 The default implementation of this function returns
11095 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
11096 and 0 otherwise. This definition is always safe, but if
11097 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
11098 nevertheless truncate the shift count, you may get better code
11102 @deftypefn {Target Hook} bool TARGET_TRULY_NOOP_TRUNCATION (poly_uint64 @var{outprec}, poly_uint64 @var{inprec})
11103 This hook returns true if it is safe to ``convert'' a value of
11104 @var{inprec} bits to one of @var{outprec} bits (where @var{outprec} is
11105 smaller than @var{inprec}) by merely operating on it as if it had only
11106 @var{outprec} bits. The default returns true unconditionally, which
11107 is correct for most machines.
11109 If @code{TARGET_MODES_TIEABLE_P} returns false for a pair of modes,
11110 suboptimal code can result if this hook returns true for the corresponding
11111 mode sizes. Making this hook return false in such cases may improve things.
11114 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (scalar_int_mode @var{mode}, scalar_int_mode @var{rep_mode})
11115 The representation of an integral mode can be such that the values
11116 are always extended to a wider integral mode. Return
11117 @code{SIGN_EXTEND} if values of @var{mode} are represented in
11118 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
11119 otherwise. (Currently, none of the targets use zero-extended
11120 representation this way so unlike @code{LOAD_EXTEND_OP},
11121 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
11122 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
11123 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
11124 widest integral mode and currently we take advantage of this fact.)
11126 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
11127 value even if the extension is not performed on certain hard registers
11128 as long as for the @code{REGNO_REG_CLASS} of these hard registers
11129 @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false.
11131 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
11132 describe two related properties. If you define
11133 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
11134 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
11137 In order to enforce the representation of @code{mode},
11138 @code{TARGET_TRULY_NOOP_TRUNCATION} should return false when truncating to
11142 @deftypefn {Target Hook} bool TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P (void)
11143 On some targets, it is assumed that the compiler will spill all pseudos
11144 that are live across a call to @code{setjmp}, while other targets treat
11145 @code{setjmp} calls as normal function calls.
11147 This hook returns false if @code{setjmp} calls do not preserve all
11148 non-volatile registers so that gcc that must spill all pseudos that are
11149 live across @code{setjmp} calls. Define this to return true if the
11150 target does not need to spill all pseudos live across @code{setjmp} calls.
11151 The default implementation conservatively assumes all pseudos must be
11152 spilled across @code{setjmp} calls.
11155 @defmac STORE_FLAG_VALUE
11156 A C expression describing the value returned by a comparison operator
11157 with an integral mode and stored by a store-flag instruction
11158 (@samp{cstore@var{mode}4}) when the condition is true. This description must
11159 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
11160 comparison operators whose results have a @code{MODE_INT} mode.
11162 A value of 1 or @minus{}1 means that the instruction implementing the
11163 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
11164 and 0 when the comparison is false. Otherwise, the value indicates
11165 which bits of the result are guaranteed to be 1 when the comparison is
11166 true. This value is interpreted in the mode of the comparison
11167 operation, which is given by the mode of the first operand in the
11168 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
11169 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
11172 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
11173 generate code that depends only on the specified bits. It can also
11174 replace comparison operators with equivalent operations if they cause
11175 the required bits to be set, even if the remaining bits are undefined.
11176 For example, on a machine whose comparison operators return an
11177 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
11178 @samp{0x80000000}, saying that just the sign bit is relevant, the
11182 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
11186 can be converted to
11189 (ashift:SI @var{x} (const_int @var{n}))
11193 where @var{n} is the appropriate shift count to move the bit being
11194 tested into the sign bit.
11196 There is no way to describe a machine that always sets the low-order bit
11197 for a true value, but does not guarantee the value of any other bits,
11198 but we do not know of any machine that has such an instruction. If you
11199 are trying to port GCC to such a machine, include an instruction to
11200 perform a logical-and of the result with 1 in the pattern for the
11201 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
11203 Often, a machine will have multiple instructions that obtain a value
11204 from a comparison (or the condition codes). Here are rules to guide the
11205 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
11210 Use the shortest sequence that yields a valid definition for
11211 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
11212 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
11213 comparison operators to do so because there may be opportunities to
11214 combine the normalization with other operations.
11217 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
11218 slightly preferred on machines with expensive jumps and 1 preferred on
11222 As a second choice, choose a value of @samp{0x80000001} if instructions
11223 exist that set both the sign and low-order bits but do not define the
11227 Otherwise, use a value of @samp{0x80000000}.
11230 Many machines can produce both the value chosen for
11231 @code{STORE_FLAG_VALUE} and its negation in the same number of
11232 instructions. On those machines, you should also define a pattern for
11233 those cases, e.g., one matching
11236 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
11239 Some machines can also perform @code{and} or @code{plus} operations on
11240 condition code values with less instructions than the corresponding
11241 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
11242 machines, define the appropriate patterns. Use the names @code{incscc}
11243 and @code{decscc}, respectively, for the patterns which perform
11244 @code{plus} or @code{minus} operations on condition code values. See
11245 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
11246 find such instruction sequences on other machines.
11248 If this macro is not defined, the default value, 1, is used. You need
11249 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
11250 instructions, or if the value generated by these instructions is 1.
11253 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
11254 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
11255 returned when comparison operators with floating-point results are true.
11256 Define this macro on machines that have comparison operations that return
11257 floating-point values. If there are no such operations, do not define
11261 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
11262 A C expression that gives a rtx representing the nonzero true element
11263 for vector comparisons. The returned rtx should be valid for the inner
11264 mode of @var{mode} which is guaranteed to be a vector mode. Define
11265 this macro on machines that have vector comparison operations that
11266 return a vector result. If there are no such operations, do not define
11267 this macro. Typically, this macro is defined as @code{const1_rtx} or
11268 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
11269 the compiler optimizing such vector comparison operations for the
11273 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
11274 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
11275 A C expression that indicates whether the architecture defines a value
11276 for @code{clz} or @code{ctz} with a zero operand.
11277 A result of @code{0} indicates the value is undefined.
11278 If the value is defined for only the RTL expression, the macro should
11279 evaluate to @code{1}; if the value applies also to the corresponding optab
11280 entry (which is normally the case if it expands directly into
11281 the corresponding RTL), then the macro should evaluate to @code{2}.
11282 In the cases where the value is defined, @var{value} should be set to
11285 If this macro is not defined, the value of @code{clz} or
11286 @code{ctz} at zero is assumed to be undefined.
11288 This macro must be defined if the target's expansion for @code{ffs}
11289 relies on a particular value to get correct results. Otherwise it
11290 is not necessary, though it may be used to optimize some corner cases, and
11291 to provide a default expansion for the @code{ffs} optab.
11293 Note that regardless of this macro the ``definedness'' of @code{clz}
11294 and @code{ctz} at zero do @emph{not} extend to the builtin functions
11295 visible to the user. Thus one may be free to adjust the value at will
11296 to match the target expansion of these operations without fear of
11301 An alias for the machine mode for pointers. On most machines, define
11302 this to be the integer mode corresponding to the width of a hardware
11303 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
11304 On some machines you must define this to be one of the partial integer
11305 modes, such as @code{PSImode}.
11307 The width of @code{Pmode} must be at least as large as the value of
11308 @code{POINTER_SIZE}. If it is not equal, you must define the macro
11309 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
11313 @defmac FUNCTION_MODE
11314 An alias for the machine mode used for memory references to functions
11315 being called, in @code{call} RTL expressions. On most CISC machines,
11316 where an instruction can begin at any byte address, this should be
11317 @code{QImode}. On most RISC machines, where all instructions have fixed
11318 size and alignment, this should be a mode with the same size and alignment
11319 as the machine instruction words - typically @code{SImode} or @code{HImode}.
11322 @defmac STDC_0_IN_SYSTEM_HEADERS
11323 In normal operation, the preprocessor expands @code{__STDC__} to the
11324 constant 1, to signify that GCC conforms to ISO Standard C@. On some
11325 hosts, like Solaris, the system compiler uses a different convention,
11326 where @code{__STDC__} is normally 0, but is 1 if the user specifies
11327 strict conformance to the C Standard.
11329 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
11330 convention when processing system header files, but when processing user
11331 files @code{__STDC__} will always expand to 1.
11334 @deftypefn {C Target Hook} {const char *} TARGET_C_PREINCLUDE (void)
11335 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.
11337 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.
11340 @deftypefn {C Target Hook} bool TARGET_CXX_IMPLICIT_EXTERN_C (const char*@var{})
11341 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.
11344 @defmac SYSTEM_IMPLICIT_EXTERN_C
11345 Define this macro if the system header files do not support C++@.
11346 This macro handles system header files by pretending that system
11347 header files are enclosed in @samp{extern "C" @{@dots{}@}}.
11352 @defmac REGISTER_TARGET_PRAGMAS ()
11353 Define this macro if you want to implement any target-specific pragmas.
11354 If defined, it is a C expression which makes a series of calls to
11355 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
11356 for each pragma. The macro may also do any
11357 setup required for the pragmas.
11359 The primary reason to define this macro is to provide compatibility with
11360 other compilers for the same target. In general, we discourage
11361 definition of target-specific pragmas for GCC@.
11363 If the pragma can be implemented by attributes then you should consider
11364 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
11366 Preprocessor macros that appear on pragma lines are not expanded. All
11367 @samp{#pragma} directives that do not match any registered pragma are
11368 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
11371 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
11372 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
11374 Each call to @code{c_register_pragma} or
11375 @code{c_register_pragma_with_expansion} establishes one pragma. The
11376 @var{callback} routine will be called when the preprocessor encounters a
11380 #pragma [@var{space}] @var{name} @dots{}
11383 @var{space} is the case-sensitive namespace of the pragma, or
11384 @code{NULL} to put the pragma in the global namespace. The callback
11385 routine receives @var{pfile} as its first argument, which can be passed
11386 on to cpplib's functions if necessary. You can lex tokens after the
11387 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
11388 callback will be silently ignored. The end of the line is indicated by
11389 a token of type @code{CPP_EOF}. Macro expansion occurs on the
11390 arguments of pragmas registered with
11391 @code{c_register_pragma_with_expansion} but not on the arguments of
11392 pragmas registered with @code{c_register_pragma}.
11394 Note that the use of @code{pragma_lex} is specific to the C and C++
11395 compilers. It will not work in the Java or Fortran compilers, or any
11396 other language compilers for that matter. Thus if @code{pragma_lex} is going
11397 to be called from target-specific code, it must only be done so when
11398 building the C and C++ compilers. This can be done by defining the
11399 variables @code{c_target_objs} and @code{cxx_target_objs} in the
11400 target entry in the @file{config.gcc} file. These variables should name
11401 the target-specific, language-specific object file which contains the
11402 code that uses @code{pragma_lex}. Note it will also be necessary to add a
11403 rule to the makefile fragment pointed to by @code{tmake_file} that shows
11404 how to build this object file.
11407 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
11408 Define this macro if macros should be expanded in the
11409 arguments of @samp{#pragma pack}.
11412 @defmac TARGET_DEFAULT_PACK_STRUCT
11413 If your target requires a structure packing default other than 0 (meaning
11414 the machine default), define this macro to the necessary value (in bytes).
11415 This must be a value that would also be valid to use with
11416 @samp{#pragma pack()} (that is, a small power of two).
11419 @defmac DOLLARS_IN_IDENTIFIERS
11420 Define this macro to control use of the character @samp{$} in
11421 identifier names for the C family of languages. 0 means @samp{$} is
11422 not allowed by default; 1 means it is allowed. 1 is the default;
11423 there is no need to define this macro in that case.
11426 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
11427 Define this macro as a C expression that is nonzero if it is safe for the
11428 delay slot scheduler to place instructions in the delay slot of @var{insn},
11429 even if they appear to use a resource set or clobbered in @var{insn}.
11430 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
11431 every @code{call_insn} has this behavior. On machines where some @code{insn}
11432 or @code{jump_insn} is really a function call and hence has this behavior,
11433 you should define this macro.
11435 You need not define this macro if it would always return zero.
11438 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
11439 Define this macro as a C expression that is nonzero if it is safe for the
11440 delay slot scheduler to place instructions in the delay slot of @var{insn},
11441 even if they appear to set or clobber a resource referenced in @var{insn}.
11442 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
11443 some @code{insn} or @code{jump_insn} is really a function call and its operands
11444 are registers whose use is actually in the subroutine it calls, you should
11445 define this macro. Doing so allows the delay slot scheduler to move
11446 instructions which copy arguments into the argument registers into the delay
11447 slot of @var{insn}.
11449 You need not define this macro if it would always return zero.
11452 @defmac MULTIPLE_SYMBOL_SPACES
11453 Define this macro as a C expression that is nonzero if, in some cases,
11454 global symbols from one translation unit may not be bound to undefined
11455 symbols in another translation unit without user intervention. For
11456 instance, under Microsoft Windows symbols must be explicitly imported
11457 from shared libraries (DLLs).
11459 You need not define this macro if it would always evaluate to zero.
11462 @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})
11463 This target hook may add @dfn{clobbers} to @var{clobbers} and
11464 @var{clobbered_regs} for any hard regs the port wishes to automatically
11465 clobber for an asm. The @var{outputs} and @var{inputs} may be inspected
11466 to avoid clobbering a register that is already used by the asm.
11468 It may modify the @var{outputs}, @var{inputs}, and @var{constraints}
11469 as necessary for other pre-processing. In this case the return value is
11470 a sequence of insns to emit after the asm.
11473 @defmac MATH_LIBRARY
11474 Define this macro as a C string constant for the linker argument to link
11475 in the system math library, minus the initial @samp{"-l"}, or
11476 @samp{""} if the target does not have a
11477 separate math library.
11479 You need only define this macro if the default of @samp{"m"} is wrong.
11482 @defmac LIBRARY_PATH_ENV
11483 Define this macro as a C string constant for the environment variable that
11484 specifies where the linker should look for libraries.
11486 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
11490 @defmac TARGET_POSIX_IO
11491 Define this macro if the target supports the following POSIX@ file
11492 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
11493 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
11494 to use file locking when exiting a program, which avoids race conditions
11495 if the program has forked. It will also create directories at run-time
11496 for cross-profiling.
11499 @defmac MAX_CONDITIONAL_EXECUTE
11501 A C expression for the maximum number of instructions to execute via
11502 conditional execution instructions instead of a branch. A value of
11503 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
11504 1 if it does use cc0.
11507 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
11508 Used if the target needs to perform machine-dependent modifications on the
11509 conditionals used for turning basic blocks into conditionally executed code.
11510 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
11511 contains information about the currently processed blocks. @var{true_expr}
11512 and @var{false_expr} are the tests that are used for converting the
11513 then-block and the else-block, respectively. Set either @var{true_expr} or
11514 @var{false_expr} to a null pointer if the tests cannot be converted.
11517 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
11518 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
11519 if-statements into conditions combined by @code{and} and @code{or} operations.
11520 @var{bb} contains the basic block that contains the test that is currently
11521 being processed and about to be turned into a condition.
11524 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
11525 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
11526 be converted to conditional execution format. @var{ce_info} points to
11527 a data structure, @code{struct ce_if_block}, which contains information
11528 about the currently processed blocks.
11531 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
11532 A C expression to perform any final machine dependent modifications in
11533 converting code to conditional execution. The involved basic blocks
11534 can be found in the @code{struct ce_if_block} structure that is pointed
11535 to by @var{ce_info}.
11538 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
11539 A C expression to cancel any machine dependent modifications in
11540 converting code to conditional execution. The involved basic blocks
11541 can be found in the @code{struct ce_if_block} structure that is pointed
11542 to by @var{ce_info}.
11545 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
11546 A C expression to initialize any machine specific data for if-conversion
11547 of the if-block in the @code{struct ce_if_block} structure that is pointed
11548 to by @var{ce_info}.
11551 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
11552 If non-null, this hook performs a target-specific pass over the
11553 instruction stream. The compiler will run it at all optimization levels,
11554 just before the point at which it normally does delayed-branch scheduling.
11556 The exact purpose of the hook varies from target to target. Some use
11557 it to do transformations that are necessary for correctness, such as
11558 laying out in-function constant pools or avoiding hardware hazards.
11559 Others use it as an opportunity to do some machine-dependent optimizations.
11561 You need not implement the hook if it has nothing to do. The default
11562 definition is null.
11565 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
11566 Define this hook if you have any machine-specific built-in functions
11567 that need to be defined. It should be a function that performs the
11570 Machine specific built-in functions can be useful to expand special machine
11571 instructions that would otherwise not normally be generated because
11572 they have no equivalent in the source language (for example, SIMD vector
11573 instructions or prefetch instructions).
11575 To create a built-in function, call the function
11576 @code{lang_hooks.builtin_function}
11577 which is defined by the language front end. You can use any type nodes set
11578 up by @code{build_common_tree_nodes};
11579 only language front ends that use those two functions will call
11580 @samp{TARGET_INIT_BUILTINS}.
11583 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
11584 Define this hook if you have any machine-specific built-in functions
11585 that need to be defined. It should be a function that returns the
11586 builtin function declaration for the builtin function code @var{code}.
11587 If there is no such builtin and it cannot be initialized at this time
11588 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
11589 If @var{code} is out of range the function should return
11590 @code{error_mark_node}.
11593 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, machine_mode @var{mode}, int @var{ignore})
11595 Expand a call to a machine specific built-in function that was set up by
11596 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
11597 function call; the result should go to @var{target} if that is
11598 convenient, and have mode @var{mode} if that is convenient.
11599 @var{subtarget} may be used as the target for computing one of
11600 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
11601 ignored. This function should return the result of the call to the
11605 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
11606 Select a replacement for a machine specific built-in function that
11607 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
11608 @emph{before} regular type checking, and so allows the target to
11609 implement a crude form of function overloading. @var{fndecl} is the
11610 declaration of the built-in function. @var{arglist} is the list of
11611 arguments passed to the built-in function. The result is a
11612 complete expression that implements the operation, usually
11613 another @code{CALL_EXPR}.
11614 @var{arglist} really has type @samp{VEC(tree,gc)*}
11617 @deftypefn {Target Hook} bool TARGET_CHECK_BUILTIN_CALL (location_t @var{loc}, vec<location_t> @var{arg_loc}, tree @var{fndecl}, tree @var{orig_fndecl}, unsigned int @var{nargs}, tree *@var{args})
11618 Perform semantic checking on a call to a machine-specific built-in
11619 function after its arguments have been constrained to the function
11620 signature. Return true if the call is valid, otherwise report an error
11623 This hook is called after @code{TARGET_RESOLVE_OVERLOADED_BUILTIN}.
11624 The call was originally to built-in function @var{orig_fndecl},
11625 but after the optional @code{TARGET_RESOLVE_OVERLOADED_BUILTIN}
11626 step is now to built-in function @var{fndecl}. @var{loc} is the
11627 location of the call and @var{args} is an array of function arguments,
11628 of which there are @var{nargs}. @var{arg_loc} specifies the location
11632 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
11633 Fold a call to a machine specific built-in function that was set up by
11634 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
11635 built-in function. @var{n_args} is the number of arguments passed to
11636 the function; the arguments themselves are pointed to by @var{argp}.
11637 The result is another tree, valid for both GIMPLE and GENERIC,
11638 containing a simplified expression for the call's result. If
11639 @var{ignore} is true the value will be ignored.
11642 @deftypefn {Target Hook} bool TARGET_GIMPLE_FOLD_BUILTIN (gimple_stmt_iterator *@var{gsi})
11643 Fold a call to a machine specific built-in function that was set up
11644 by @samp{TARGET_INIT_BUILTINS}. @var{gsi} points to the gimple
11645 statement holding the function call. Returns true if any change
11646 was made to the GIMPLE stream.
11649 @deftypefn {Target Hook} int TARGET_COMPARE_VERSION_PRIORITY (tree @var{decl1}, tree @var{decl2})
11650 This hook is used to compare the target attributes in two functions to
11651 determine which function's features get higher priority. This is used
11652 during function multi-versioning to figure out the order in which two
11653 versions must be dispatched. A function version with a higher priority
11654 is checked for dispatching earlier. @var{decl1} and @var{decl2} are
11655 the two function decls that will be compared.
11658 @deftypefn {Target Hook} tree TARGET_GET_FUNCTION_VERSIONS_DISPATCHER (void *@var{decl})
11659 This hook is used to get the dispatcher function for a set of function
11660 versions. The dispatcher function is called to invoke the right function
11661 version at run-time. @var{decl} is one version from a set of semantically
11662 identical versions.
11665 @deftypefn {Target Hook} tree TARGET_GENERATE_VERSION_DISPATCHER_BODY (void *@var{arg})
11666 This hook is used to generate the dispatcher logic to invoke the right
11667 function version at run-time for a given set of function versions.
11668 @var{arg} points to the callgraph node of the dispatcher function whose
11669 body must be generated.
11672 @deftypefn {Target Hook} bool TARGET_PREDICT_DOLOOP_P (class loop *@var{loop})
11673 Return true if we can predict it is possible to use a low-overhead loop
11674 for a particular loop. The parameter @var{loop} is a pointer to the loop.
11675 This target hook is required only when the target supports low-overhead
11676 loops, and will help ivopts to make some decisions.
11677 The default version of this hook returns false.
11680 @deftypevr {Target Hook} bool TARGET_HAVE_COUNT_REG_DECR_P
11681 Return true if the target supports hardware count register for decrement
11683 The default value is false.
11686 @deftypevr {Target Hook} int64_t TARGET_DOLOOP_COST_FOR_GENERIC
11687 One IV candidate dedicated for doloop is introduced in IVOPTs, we can
11688 calculate the computation cost of adopting it to any generic IV use by
11689 function get_computation_cost as before. But for targets which have
11690 hardware count register support for decrement and branch, it may have to
11691 move IV value from hardware count register to general purpose register
11692 while doloop IV candidate is used for generic IV uses. It probably takes
11693 expensive penalty. This hook allows target owners to define the cost for
11694 this especially for generic IV uses.
11695 The default value is zero.
11698 @deftypevr {Target Hook} int64_t TARGET_DOLOOP_COST_FOR_ADDRESS
11699 One IV candidate dedicated for doloop is introduced in IVOPTs, we can
11700 calculate the computation cost of adopting it to any address IV use by
11701 function get_computation_cost as before. But for targets which have
11702 hardware count register support for decrement and branch, it may have to
11703 move IV value from hardware count register to general purpose register
11704 while doloop IV candidate is used for address IV uses. It probably takes
11705 expensive penalty. This hook allows target owners to define the cost for
11706 this escpecially for address IV uses.
11707 The default value is zero.
11710 @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})
11711 Return true if it is possible to use low-overhead loops (@code{doloop_end}
11712 and @code{doloop_begin}) for a particular loop. @var{iterations} gives the
11713 exact number of iterations, or 0 if not known. @var{iterations_max} gives
11714 the maximum number of iterations, or 0 if not known. @var{loop_depth} is
11715 the nesting depth of the loop, with 1 for innermost loops, 2 for loops that
11716 contain innermost loops, and so on. @var{entered_at_top} is true if the
11717 loop is only entered from the top.
11719 This hook is only used if @code{doloop_end} is available. The default
11720 implementation returns true. You can use @code{can_use_doloop_if_innermost}
11721 if the loop must be the innermost, and if there are no other restrictions.
11724 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const rtx_insn *@var{insn})
11726 Take an instruction in @var{insn} and return NULL if it is valid within a
11727 low-overhead loop, otherwise return a string explaining why doloop
11728 could not be applied.
11730 Many targets use special registers for low-overhead looping. For any
11731 instruction that clobbers these this function should return a string indicating
11732 the reason why the doloop could not be applied.
11733 By default, the RTL loop optimizer does not use a present doloop pattern for
11734 loops containing function calls or branch on table instructions.
11737 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_COMBINED_INSN (rtx_insn *@var{insn})
11738 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.
11741 @deftypefn {Target Hook} bool TARGET_CAN_FOLLOW_JUMP (const rtx_insn *@var{follower}, const rtx_insn *@var{followee})
11742 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.
11745 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
11746 This target hook returns @code{true} if @var{x} is considered to be commutative.
11747 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
11748 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
11749 of the enclosing rtl, if known, otherwise it is UNKNOWN.
11752 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
11754 When the initial value of a hard register has been copied in a pseudo
11755 register, it is often not necessary to actually allocate another register
11756 to this pseudo register, because the original hard register or a stack slot
11757 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
11758 is called at the start of register allocation once for each hard register
11759 that had its initial value copied by using
11760 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
11761 Possible values are @code{NULL_RTX}, if you don't want
11762 to do any special allocation, a @code{REG} rtx---that would typically be
11763 the hard register itself, if it is known not to be clobbered---or a
11765 If you are returning a @code{MEM}, this is only a hint for the allocator;
11766 it might decide to use another register anyways.
11767 You may use @code{current_function_is_leaf} or
11768 @code{REG_N_SETS} in the hook to determine if the hard
11769 register in question will not be clobbered.
11770 The default value of this hook is @code{NULL}, which disables any special
11774 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
11775 This target hook returns nonzero if @var{x}, an @code{unspec} or
11776 @code{unspec_volatile} operation, might cause a trap. Targets can use
11777 this hook to enhance precision of analysis for @code{unspec} and
11778 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
11779 to analyze inner elements of @var{x} in which case @var{flags} should be
11783 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
11784 The compiler invokes this hook whenever it changes its current function
11785 context (@code{cfun}). You can define this function if
11786 the back end needs to perform any initialization or reset actions on a
11787 per-function basis. For example, it may be used to implement function
11788 attributes that affect register usage or code generation patterns.
11789 The argument @var{decl} is the declaration for the new function context,
11790 and may be null to indicate that the compiler has left a function context
11791 and is returning to processing at the top level.
11792 The default hook function does nothing.
11794 GCC sets @code{cfun} to a dummy function context during initialization of
11795 some parts of the back end. The hook function is not invoked in this
11796 situation; you need not worry about the hook being invoked recursively,
11797 or when the back end is in a partially-initialized state.
11798 @code{cfun} might be @code{NULL} to indicate processing at top level,
11799 outside of any function scope.
11802 @defmac TARGET_OBJECT_SUFFIX
11803 Define this macro to be a C string representing the suffix for object
11804 files on your target machine. If you do not define this macro, GCC will
11805 use @samp{.o} as the suffix for object files.
11808 @defmac TARGET_EXECUTABLE_SUFFIX
11809 Define this macro to be a C string representing the suffix to be
11810 automatically added to executable files on your target machine. If you
11811 do not define this macro, GCC will use the null string as the suffix for
11815 @defmac COLLECT_EXPORT_LIST
11816 If defined, @code{collect2} will scan the individual object files
11817 specified on its command line and create an export list for the linker.
11818 Define this macro for systems like AIX, where the linker discards
11819 object files that are not referenced from @code{main} and uses export
11823 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11824 Define this macro to a C expression representing a variant of the
11825 method call @var{mdecl}, if Java Native Interface (JNI) methods
11826 must be invoked differently from other methods on your target.
11827 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11828 the @code{stdcall} calling convention and this macro is then
11829 defined as this expression:
11832 build_type_attribute_variant (@var{mdecl},
11834 (get_identifier ("stdcall"),
11839 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11840 This target hook returns @code{true} past the point in which new jump
11841 instructions could be created. On machines that require a register for
11842 every jump such as the SHmedia ISA of SH5, this point would typically be
11843 reload, so this target hook should be defined to a function such as:
11847 cannot_modify_jumps_past_reload_p ()
11849 return (reload_completed || reload_in_progress);
11854 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11855 This target hook returns true if the target supports conditional execution.
11856 This target hook is required only when the target has several different
11857 modes and they have different conditional execution capability, such as ARM.
11860 @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})
11861 This function prepares to emit a comparison insn for the first compare in a
11862 sequence of conditional comparisions. It returns an appropriate comparison
11863 with @code{CC} for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11864 The insns to prepare the compare are saved in @var{prep_seq} and the compare
11865 insns are saved in @var{gen_seq}. They will be emitted when all the
11866 compares in the the conditional comparision are generated without error.
11867 @var{code} is the @code{rtx_code} of the compare for @var{op0} and @var{op1}.
11870 @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})
11871 This function prepares to emit a conditional comparison within a sequence
11872 of conditional comparisons. It returns an appropriate comparison with
11873 @code{CC} for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11874 The insns to prepare the compare are saved in @var{prep_seq} and the compare
11875 insns are saved in @var{gen_seq}. They will be emitted when all the
11876 compares in the conditional comparision are generated without error. The
11877 @var{prev} expression is the result of a prior call to @code{gen_ccmp_first}
11878 or @code{gen_ccmp_next}. It may return @code{NULL} if the combination of
11879 @var{prev} and this comparison is not supported, otherwise the result must
11880 be appropriate for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11881 @var{code} is the @code{rtx_code} of the compare for @var{op0} and @var{op1}.
11882 @var{bit_code} is @code{AND} or @code{IOR}, which is the op on the compares.
11885 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, class loop *@var{loop})
11886 This target hook returns a new value for the number of times @var{loop}
11887 should be unrolled. The parameter @var{nunroll} is the number of times
11888 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11889 the loop, which is going to be checked for unrolling. This target hook
11890 is required only when the target has special constraints like maximum
11891 number of memory accesses.
11894 @defmac POWI_MAX_MULTS
11895 If defined, this macro is interpreted as a signed integer C expression
11896 that specifies the maximum number of floating point multiplications
11897 that should be emitted when expanding exponentiation by an integer
11898 constant inline. When this value is defined, exponentiation requiring
11899 more than this number of multiplications is implemented by calling the
11900 system library's @code{pow}, @code{powf} or @code{powl} routines.
11901 The default value places no upper bound on the multiplication count.
11904 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11905 This target hook should register any extra include files for the
11906 target. The parameter @var{stdinc} indicates if normal include files
11907 are present. The parameter @var{sysroot} is the system root directory.
11908 The parameter @var{iprefix} is the prefix for the gcc directory.
11911 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11912 This target hook should register any extra include files for the
11913 target before any standard headers. The parameter @var{stdinc}
11914 indicates if normal include files are present. The parameter
11915 @var{sysroot} is the system root directory. The parameter
11916 @var{iprefix} is the prefix for the gcc directory.
11919 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11920 This target hook should register special include paths for the target.
11921 The parameter @var{path} is the include to register. On Darwin
11922 systems, this is used for Framework includes, which have semantics
11923 that are different from @option{-I}.
11926 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11927 This target macro returns @code{true} if it is safe to use a local alias
11928 for a virtual function @var{fndecl} when constructing thunks,
11929 @code{false} otherwise. By default, the macro returns @code{true} for all
11930 functions, if a target supports aliases (i.e.@: defines
11931 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11934 @defmac TARGET_FORMAT_TYPES
11935 If defined, this macro is the name of a global variable containing
11936 target-specific format checking information for the @option{-Wformat}
11937 option. The default is to have no target-specific format checks.
11940 @defmac TARGET_N_FORMAT_TYPES
11941 If defined, this macro is the number of entries in
11942 @code{TARGET_FORMAT_TYPES}.
11945 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11946 If defined, this macro is the name of a global variable containing
11947 target-specific format overrides for the @option{-Wformat} option. The
11948 default is to have no target-specific format overrides. If defined,
11949 @code{TARGET_FORMAT_TYPES} must be defined, too.
11952 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11953 If defined, this macro specifies the number of entries in
11954 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11957 @defmac TARGET_OVERRIDES_FORMAT_INIT
11958 If defined, this macro specifies the optional initialization
11959 routine for target specific customizations of the system printf
11960 and scanf formatter settings.
11963 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11964 If defined, this macro returns the diagnostic message when it is
11965 illegal to pass argument @var{val} to function @var{funcdecl}
11966 with prototype @var{typelist}.
11969 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11970 If defined, this macro returns the diagnostic message when it is
11971 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11972 if validity should be determined by the front end.
11975 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11976 If defined, this macro returns the diagnostic message when it is
11977 invalid to apply operation @var{op} (where unary plus is denoted by
11978 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11979 if validity should be determined by the front end.
11982 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11983 If defined, this macro returns the diagnostic message when it is
11984 invalid to apply operation @var{op} to operands of types @var{type1}
11985 and @var{type2}, or @code{NULL} if validity should be determined by
11989 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11990 If defined, this target hook returns the type to which values of
11991 @var{type} should be promoted when they appear in expressions,
11992 analogous to the integer promotions, or @code{NULL_TREE} to use the
11993 front end's normal promotion rules. This hook is useful when there are
11994 target-specific types with special promotion rules.
11995 This is currently used only by the C and C++ front ends.
11998 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11999 If defined, this hook returns the result of converting @var{expr} to
12000 @var{type}. It should return the converted expression,
12001 or @code{NULL_TREE} to apply the front end's normal conversion rules.
12002 This hook is useful when there are target-specific types with special
12004 This is currently used only by the C and C++ front ends.
12008 This macro determines the size of the objective C jump buffer for the
12009 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
12012 @defmac LIBGCC2_UNWIND_ATTRIBUTE
12013 Define this macro if any target-specific attributes need to be attached
12014 to the functions in @file{libgcc} that provide low-level support for
12015 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
12016 and the associated definitions of those functions.
12019 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
12020 Define this macro to update the current function stack boundary if
12024 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
12025 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
12026 different argument pointer register is needed to access the function's
12027 argument list due to stack realignment. Return @code{NULL} if no DRAP
12031 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
12032 When optimization is disabled, this hook indicates whether or not
12033 arguments should be allocated to stack slots. Normally, GCC allocates
12034 stacks slots for arguments when not optimizing in order to make
12035 debugging easier. However, when a function is declared with
12036 @code{__attribute__((naked))}, there is no stack frame, and the compiler
12037 cannot safely move arguments from the registers in which they are passed
12038 to the stack. Therefore, this hook should return true in general, but
12039 false for naked functions. The default implementation always returns true.
12042 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
12043 On some architectures it can take multiple instructions to synthesize
12044 a constant. If there is another constant already in a register that
12045 is close enough in value then it is preferable that the new constant
12046 is computed from this register using immediate addition or
12047 subtraction. We accomplish this through CSE. Besides the value of
12048 the constant we also add a lower and an upper constant anchor to the
12049 available expressions. These are then queried when encountering new
12050 constants. The anchors are computed by rounding the constant up and
12051 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
12052 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
12053 accepted by immediate-add plus one. We currently assume that the
12054 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
12055 MIPS, where add-immediate takes a 16-bit signed value,
12056 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
12057 is zero, which disables this optimization.
12060 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_ASAN_SHADOW_OFFSET (void)
12061 Return the offset bitwise ored into shifted address to get corresponding
12062 Address Sanitizer shadow memory address. NULL if Address Sanitizer is not
12063 supported by the target.
12066 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_MEMMODEL_CHECK (unsigned HOST_WIDE_INT @var{val})
12067 Validate target specific memory model mask bits. When NULL no target specific
12068 memory model bits are allowed.
12071 @deftypevr {Target Hook} {unsigned char} TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
12072 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}.
12075 @deftypefn {Target Hook} bool TARGET_HAS_IFUNC_P (void)
12076 It returns true if the target supports GNU indirect functions.
12077 The support includes the assembler, linker and dynamic linker.
12078 The default value of this hook is based on target's libc.
12081 @deftypefn {Target Hook} {unsigned int} TARGET_ATOMIC_ALIGN_FOR_MODE (machine_mode @var{mode})
12082 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.
12085 @deftypefn {Target Hook} void TARGET_ATOMIC_ASSIGN_EXPAND_FENV (tree *@var{hold}, tree *@var{clear}, tree *@var{update})
12086 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}}.
12089 @deftypefn {Target Hook} void TARGET_RECORD_OFFLOAD_SYMBOL (tree)
12090 Used when offloaded functions are seen in the compilation unit and no named
12091 sections are available. It is called once for each symbol that must be
12092 recorded in the offload function and variable table.
12095 @deftypefn {Target Hook} {char *} TARGET_OFFLOAD_OPTIONS (void)
12096 Used when writing out the list of options into an LTO file. It should
12097 translate any relevant target-specific options (such as the ABI in use)
12098 into one of the @option{-foffload} options that exist as a common interface
12099 to express such options. It should return a string containing these options,
12100 separated by spaces, which the caller will free.
12104 @defmac TARGET_SUPPORTS_WIDE_INT
12106 On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
12107 objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
12108 to indicate that large integers are stored in
12109 @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows
12110 very large integer constants to be represented. @code{CONST_DOUBLE}
12111 is limited to twice the size of the host's @code{HOST_WIDE_INT}
12114 Converting a port mostly requires looking for the places where
12115 @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
12116 code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i
12117 const_double"} at the port level gets you to 95% of the changes that
12118 need to be made. There are a few places that require a deeper look.
12122 There is no equivalent to @code{hval} and @code{lval} for
12123 @code{CONST_WIDE_INT}s. This would be difficult to express in the md
12124 language since there are a variable number of elements.
12126 Most ports only check that @code{hval} is either 0 or -1 to see if the
12127 value is small. As mentioned above, this will no longer be necessary
12128 since small constants are always @code{CONST_INT}. Of course there
12129 are still a few exceptions, the alpha's constraint used by the zap
12130 instruction certainly requires careful examination by C code.
12131 However, all the current code does is pass the hval and lval to C
12132 code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
12133 not really a large change.
12136 Because there is no standard template that ports use to materialize
12137 constants, there is likely to be some futzing that is unique to each
12141 The rtx costs may have to be adjusted to properly account for larger
12142 constants that are represented as @code{CONST_WIDE_INT}.
12145 All and all it does not take long to convert ports that the
12146 maintainer is familiar with.
12150 @deftypefn {Target Hook} bool TARGET_HAVE_SPECULATION_SAFE_VALUE (bool @var{active})
12151 This hook is used to determine the level of target support for
12152 @code{__builtin_speculation_safe_value}. If called with an argument
12153 of false, it returns true if the target has been modified to support
12154 this builtin. If called with an argument of true, it returns true
12155 if the target requires active mitigation execution might be speculative.
12157 The default implementation returns false if the target does not define
12158 a pattern named @code{speculation_barrier}. Else it returns true
12159 for the first case and whether the pattern is enabled for the current
12160 compilation for the second case.
12162 For targets that have no processors that can execute instructions
12163 speculatively an alternative implemenation of this hook is available:
12164 simply redefine this hook to @code{speculation_safe_value_not_needed}
12165 along with your other target hooks.
12168 @deftypefn {Target Hook} rtx TARGET_SPECULATION_SAFE_VALUE (machine_mode @var{mode}, rtx @var{result}, rtx @var{val}, rtx @var{failval})
12169 This target hook can be used to generate a target-specific code
12170 sequence that implements the @code{__builtin_speculation_safe_value}
12171 built-in function. The function must always return @var{val} in
12172 @var{result} in mode @var{mode} when the cpu is not executing
12173 speculatively, but must never return that when speculating until it
12174 is known that the speculation will not be unwound. The hook supports
12175 two primary mechanisms for implementing the requirements. The first
12176 is to emit a speculation barrier which forces the processor to wait
12177 until all prior speculative operations have been resolved; the second
12178 is to use a target-specific mechanism that can track the speculation
12179 state and to return @var{failval} if it can determine that
12180 speculation must be unwound at a later time.
12182 The default implementation simply copies @var{val} to @var{result} and
12183 emits a @code{speculation_barrier} instruction if that is defined.
12186 @deftypefn {Target Hook} void TARGET_RUN_TARGET_SELFTESTS (void)
12187 If selftests are enabled, run any selftests for this target.