1 @c Copyright (C) 1988-2016 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
6 @chapter Target Description Macros and Functions
7 @cindex machine description macros
8 @cindex target description macros
9 @cindex macros, target description
10 @cindex @file{tm.h} macros
12 In addition to the file @file{@var{machine}.md}, a machine description
13 includes a C header file conventionally given the name
14 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
15 The header file defines numerous macros that convey the information
16 about the target machine that does not fit into the scheme of the
17 @file{.md} file. The file @file{tm.h} should be a link to
18 @file{@var{machine}.h}. The header file @file{config.h} includes
19 @file{tm.h} and most compiler source files include @file{config.h}. The
20 source file defines a variable @code{targetm}, which is a structure
21 containing pointers to functions and data relating to the target
22 machine. @file{@var{machine}.c} should also contain their definitions,
23 if they are not defined elsewhere in GCC, and other functions called
24 through the macros defined in the @file{.h} file.
27 * Target Structure:: The @code{targetm} variable.
28 * Driver:: Controlling how the driver runs the compilation passes.
29 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
30 * Per-Function Data:: Defining data structures for per-function information.
31 * Storage Layout:: Defining sizes and alignments of data.
32 * Type Layout:: Defining sizes and properties of basic user data types.
33 * Registers:: Naming and describing the hardware registers.
34 * Register Classes:: Defining the classes of hardware registers.
35 * Stack and Calling:: Defining which way the stack grows and by how much.
36 * Varargs:: Defining the varargs macros.
37 * Trampolines:: Code set up at run time to enter a nested function.
38 * Library Calls:: Controlling how library routines are implicitly called.
39 * Addressing Modes:: Defining addressing modes valid for memory operands.
40 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
41 * Condition Code:: Defining how insns update the condition code.
42 * Costs:: Defining relative costs of different operations.
43 * Scheduling:: Adjusting the behavior of the instruction scheduler.
44 * Sections:: Dividing storage into text, data, and other sections.
45 * PIC:: Macros for position independent code.
46 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
47 * Debugging Info:: Defining the format of debugging output.
48 * Floating Point:: Handling floating point for cross-compilers.
49 * Mode Switching:: Insertion of mode-switching instructions.
50 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
51 * Emulated TLS:: Emulated TLS support.
52 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
53 * PCH Target:: Validity checking for precompiled headers.
54 * C++ ABI:: Controlling C++ ABI changes.
55 * Named Address Spaces:: Adding support for named address spaces
56 * Misc:: Everything else.
59 @node Target Structure
60 @section The Global @code{targetm} Variable
62 @cindex target functions
64 @deftypevar {struct gcc_target} targetm
65 The target @file{.c} file must define the global @code{targetm} variable
66 which contains pointers to functions and data relating to the target
67 machine. The variable is declared in @file{target.h};
68 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
69 used to initialize the variable, and macros for the default initializers
70 for elements of the structure. The @file{.c} file should override those
71 macros for which the default definition is inappropriate. For example:
74 #include "target-def.h"
76 /* @r{Initialize the GCC target structure.} */
78 #undef TARGET_COMP_TYPE_ATTRIBUTES
79 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
81 struct gcc_target targetm = TARGET_INITIALIZER;
85 Where a macro should be defined in the @file{.c} file in this manner to
86 form part of the @code{targetm} structure, it is documented below as a
87 ``Target Hook'' with a prototype. Many macros will change in future
88 from being defined in the @file{.h} file to being part of the
89 @code{targetm} structure.
91 Similarly, there is a @code{targetcm} variable for hooks that are
92 specific to front ends for C-family languages, documented as ``C
93 Target Hook''. This is declared in @file{c-family/c-target.h}, the
94 initializer @code{TARGETCM_INITIALIZER} in
95 @file{c-family/c-target-def.h}. If targets initialize @code{targetcm}
96 themselves, they should set @code{target_has_targetcm=yes} in
97 @file{config.gcc}; otherwise a default definition is used.
99 Similarly, there is a @code{targetm_common} variable for hooks that
100 are shared between the compiler driver and the compilers proper,
101 documented as ``Common Target Hook''. This is declared in
102 @file{common/common-target.h}, the initializer
103 @code{TARGETM_COMMON_INITIALIZER} in
104 @file{common/common-target-def.h}. If targets initialize
105 @code{targetm_common} themselves, they should set
106 @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
107 default definition is used.
110 @section Controlling the Compilation Driver, @file{gcc}
112 @cindex controlling the compilation driver
114 @c prevent bad page break with this line
115 You can control the compilation driver.
117 @defmac DRIVER_SELF_SPECS
118 A list of specs for the driver itself. It should be a suitable
119 initializer for an array of strings, with no surrounding braces.
121 The driver applies these specs to its own command line between loading
122 default @file{specs} files (but not command-line specified ones) and
123 choosing the multilib directory or running any subcommands. It
124 applies them in the order given, so each spec can depend on the
125 options added by earlier ones. It is also possible to remove options
126 using @samp{%<@var{option}} in the usual way.
128 This macro can be useful when a port has several interdependent target
129 options. It provides a way of standardizing the command line so
130 that the other specs are easier to write.
132 Do not define this macro if it does not need to do anything.
135 @defmac OPTION_DEFAULT_SPECS
136 A list of specs used to support configure-time default options (i.e.@:
137 @option{--with} options) in the driver. It should be a suitable initializer
138 for an array of structures, each containing two strings, without the
139 outermost pair of surrounding braces.
141 The first item in the pair is the name of the default. This must match
142 the code in @file{config.gcc} for the target. The second item is a spec
143 to apply if a default with this name was specified. The string
144 @samp{%(VALUE)} in the spec will be replaced by the value of the default
145 everywhere it occurs.
147 The driver will apply these specs to its own command line between loading
148 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
149 the same mechanism as @code{DRIVER_SELF_SPECS}.
151 Do not define this macro if it does not need to do anything.
155 A C string constant that tells the GCC driver program options to
156 pass to CPP@. It can also specify how to translate options you
157 give to GCC into options for GCC to pass to the CPP@.
159 Do not define this macro if it does not need to do anything.
162 @defmac CPLUSPLUS_CPP_SPEC
163 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
164 than C@. If you do not define this macro, then the value of
165 @code{CPP_SPEC} (if any) will be used instead.
169 A C string constant that tells the GCC driver program options to
170 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
172 It can also specify how to translate options you give to GCC into options
173 for GCC to pass to front ends.
175 Do not define this macro if it does not need to do anything.
179 A C string constant that tells the GCC driver program options to
180 pass to @code{cc1plus}. It can also specify how to translate options you
181 give to GCC into options for GCC to pass to the @code{cc1plus}.
183 Do not define this macro if it does not need to do anything.
184 Note that everything defined in CC1_SPEC is already passed to
185 @code{cc1plus} so there is no need to duplicate the contents of
186 CC1_SPEC in CC1PLUS_SPEC@.
190 A C string constant that tells the GCC driver program options to
191 pass to the assembler. It can also specify how to translate options
192 you give to GCC into options for GCC to pass to the assembler.
193 See the file @file{sun3.h} for an example of this.
195 Do not define this macro if it does not need to do anything.
198 @defmac ASM_FINAL_SPEC
199 A C string constant that tells the GCC driver program how to
200 run any programs which cleanup after the normal assembler.
201 Normally, this is not needed. See the file @file{mips.h} for
204 Do not define this macro if it does not need to do anything.
207 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
208 Define this macro, with no value, if the driver should give the assembler
209 an argument consisting of a single dash, @option{-}, to instruct it to
210 read from its standard input (which will be a pipe connected to the
211 output of the compiler proper). This argument is given after any
212 @option{-o} option specifying the name of the output file.
214 If you do not define this macro, the assembler is assumed to read its
215 standard input if given no non-option arguments. If your assembler
216 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
217 see @file{mips.h} for instance.
221 A C string constant that tells the GCC driver program options to
222 pass to the linker. It can also specify how to translate options you
223 give to GCC into options for GCC to pass to the linker.
225 Do not define this macro if it does not need to do anything.
229 Another C string constant used much like @code{LINK_SPEC}. The difference
230 between the two is that @code{LIB_SPEC} is used at the end of the
231 command given to the linker.
233 If this macro is not defined, a default is provided that
234 loads the standard C library from the usual place. See @file{gcc.c}.
238 Another C string constant that tells the GCC driver program
239 how and when to place a reference to @file{libgcc.a} into the
240 linker command line. This constant is placed both before and after
241 the value of @code{LIB_SPEC}.
243 If this macro is not defined, the GCC driver provides a default that
244 passes the string @option{-lgcc} to the linker.
247 @defmac REAL_LIBGCC_SPEC
248 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
249 @code{LIBGCC_SPEC} is not directly used by the driver program but is
250 instead modified to refer to different versions of @file{libgcc.a}
251 depending on the values of the command line flags @option{-static},
252 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
253 targets where these modifications are inappropriate, define
254 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
255 driver how to place a reference to @file{libgcc} on the link command
256 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
259 @defmac USE_LD_AS_NEEDED
260 A macro that controls the modifications to @code{LIBGCC_SPEC}
261 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
262 generated that uses @option{--as-needed} or equivalent options and the
263 shared @file{libgcc} in place of the
264 static exception handler library, when linking without any of
265 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
269 If defined, this C string constant is added to @code{LINK_SPEC}.
270 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
271 the modifications to @code{LIBGCC_SPEC} mentioned in
272 @code{REAL_LIBGCC_SPEC}.
275 @defmac STARTFILE_SPEC
276 Another C string constant used much like @code{LINK_SPEC}. The
277 difference between the two is that @code{STARTFILE_SPEC} is used at
278 the very beginning of the command given to the linker.
280 If this macro is not defined, a default is provided that loads the
281 standard C startup file from the usual place. See @file{gcc.c}.
285 Another C string constant used much like @code{LINK_SPEC}. The
286 difference between the two is that @code{ENDFILE_SPEC} is used at
287 the very end of the command given to the linker.
289 Do not define this macro if it does not need to do anything.
292 @defmac THREAD_MODEL_SPEC
293 GCC @code{-v} will print the thread model GCC was configured to use.
294 However, this doesn't work on platforms that are multilibbed on thread
295 models, such as AIX 4.3. On such platforms, define
296 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
297 blanks that names one of the recognized thread models. @code{%*}, the
298 default value of this macro, will expand to the value of
299 @code{thread_file} set in @file{config.gcc}.
302 @defmac SYSROOT_SUFFIX_SPEC
303 Define this macro to add a suffix to the target sysroot when GCC is
304 configured with a sysroot. This will cause GCC to search for usr/lib,
305 et al, within sysroot+suffix.
308 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
309 Define this macro to add a headers_suffix to the target sysroot when
310 GCC is configured with a sysroot. This will cause GCC to pass the
311 updated sysroot+headers_suffix to CPP, causing it to search for
312 usr/include, et al, within sysroot+headers_suffix.
316 Define this macro to provide additional specifications to put in the
317 @file{specs} file that can be used in various specifications like
320 The definition should be an initializer for an array of structures,
321 containing a string constant, that defines the specification name, and a
322 string constant that provides the specification.
324 Do not define this macro if it does not need to do anything.
326 @code{EXTRA_SPECS} is useful when an architecture contains several
327 related targets, which have various @code{@dots{}_SPECS} which are similar
328 to each other, and the maintainer would like one central place to keep
331 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
332 define either @code{_CALL_SYSV} when the System V calling sequence is
333 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
336 The @file{config/rs6000/rs6000.h} target file defines:
339 #define EXTRA_SPECS \
340 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
342 #define CPP_SYS_DEFAULT ""
345 The @file{config/rs6000/sysv.h} target file defines:
349 "%@{posix: -D_POSIX_SOURCE @} \
350 %@{mcall-sysv: -D_CALL_SYSV @} \
351 %@{!mcall-sysv: %(cpp_sysv_default) @} \
352 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
354 #undef CPP_SYSV_DEFAULT
355 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
358 while the @file{config/rs6000/eabiaix.h} target file defines
359 @code{CPP_SYSV_DEFAULT} as:
362 #undef CPP_SYSV_DEFAULT
363 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
367 @defmac LINK_LIBGCC_SPECIAL_1
368 Define this macro if the driver program should find the library
369 @file{libgcc.a}. If you do not define this macro, the driver program will pass
370 the argument @option{-lgcc} to tell the linker to do the search.
373 @defmac LINK_GCC_C_SEQUENCE_SPEC
374 The sequence in which libgcc and libc are specified to the linker.
375 By default this is @code{%G %L %G}.
378 @defmac POST_LINK_SPEC
379 Define this macro to add additional steps to be executed after linker.
380 The default value of this macro is empty string.
383 @defmac LINK_COMMAND_SPEC
384 A C string constant giving the complete command line need to execute the
385 linker. When you do this, you will need to update your port each time a
386 change is made to the link command line within @file{gcc.c}. Therefore,
387 define this macro only if you need to completely redefine the command
388 line for invoking the linker and there is no other way to accomplish
389 the effect you need. Overriding this macro may be avoidable by overriding
390 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
393 @deftypevr {Common Target Hook} bool TARGET_ALWAYS_STRIP_DOTDOT
394 True if @file{..} components should always be removed from directory names computed relative to GCC's internal directories, false (default) if such components should be preserved and directory names containing them passed to other tools such as the linker.
397 @defmac MULTILIB_DEFAULTS
398 Define this macro as a C expression for the initializer of an array of
399 string to tell the driver program which options are defaults for this
400 target and thus do not need to be handled specially when using
401 @code{MULTILIB_OPTIONS}.
403 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
404 the target makefile fragment or if none of the options listed in
405 @code{MULTILIB_OPTIONS} are set by default.
406 @xref{Target Fragment}.
409 @defmac RELATIVE_PREFIX_NOT_LINKDIR
410 Define this macro to tell @command{gcc} that it should only translate
411 a @option{-B} prefix into a @option{-L} linker option if the prefix
412 indicates an absolute file name.
415 @defmac MD_EXEC_PREFIX
416 If defined, this macro is an additional prefix to try after
417 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
418 when the compiler is built as a cross
419 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
420 to the list of directories used to find the assembler in @file{configure.ac}.
423 @defmac STANDARD_STARTFILE_PREFIX
424 Define this macro as a C string constant if you wish to override the
425 standard choice of @code{libdir} as the default prefix to
426 try when searching for startup files such as @file{crt0.o}.
427 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
428 is built as a cross compiler.
431 @defmac STANDARD_STARTFILE_PREFIX_1
432 Define this macro as a C string constant if you wish to override the
433 standard choice of @code{/lib} as a prefix to try after the default prefix
434 when searching for startup files such as @file{crt0.o}.
435 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
436 is built as a cross compiler.
439 @defmac STANDARD_STARTFILE_PREFIX_2
440 Define this macro as a C string constant if you wish to override the
441 standard choice of @code{/lib} as yet another prefix to try after the
442 default prefix when searching for startup files such as @file{crt0.o}.
443 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
444 is built as a cross compiler.
447 @defmac MD_STARTFILE_PREFIX
448 If defined, this macro supplies an additional prefix to try after the
449 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
450 compiler is built as a cross compiler.
453 @defmac MD_STARTFILE_PREFIX_1
454 If defined, this macro supplies yet another prefix to try after the
455 standard prefixes. It is not searched when the compiler is built as a
459 @defmac INIT_ENVIRONMENT
460 Define this macro as a C string constant if you wish to set environment
461 variables for programs called by the driver, such as the assembler and
462 loader. The driver passes the value of this macro to @code{putenv} to
463 initialize the necessary environment variables.
466 @defmac LOCAL_INCLUDE_DIR
467 Define this macro as a C string constant if you wish to override the
468 standard choice of @file{/usr/local/include} as the default prefix to
469 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
470 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
471 @file{config.gcc}, normally @file{/usr/include}) in the search order.
473 Cross compilers do not search either @file{/usr/local/include} or its
477 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
478 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
479 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
480 If you do not define this macro, no component is used.
483 @defmac INCLUDE_DEFAULTS
484 Define this macro if you wish to override the entire default search path
485 for include files. For a native compiler, the default search path
486 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
487 @code{GPLUSPLUS_INCLUDE_DIR}, and
488 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
489 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
490 and specify private search areas for GCC@. The directory
491 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
493 The definition should be an initializer for an array of structures.
494 Each array element should have four elements: the directory name (a
495 string constant), the component name (also a string constant), a flag
496 for C++-only directories,
497 and a flag showing that the includes in the directory don't need to be
498 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
499 the array with a null element.
501 The component name denotes what GNU package the include file is part of,
502 if any, in all uppercase letters. For example, it might be @samp{GCC}
503 or @samp{BINUTILS}. If the package is part of a vendor-supplied
504 operating system, code the component name as @samp{0}.
506 For example, here is the definition used for VAX/VMS:
509 #define INCLUDE_DEFAULTS \
511 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
512 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
513 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
520 Here is the order of prefixes tried for exec files:
524 Any prefixes specified by the user with @option{-B}.
527 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
528 is not set and the compiler has not been installed in the configure-time
529 @var{prefix}, the location in which the compiler has actually been installed.
532 The directories specified by the environment variable @code{COMPILER_PATH}.
535 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
536 in the configured-time @var{prefix}.
539 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
542 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
545 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
549 Here is the order of prefixes tried for startfiles:
553 Any prefixes specified by the user with @option{-B}.
556 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
557 value based on the installed toolchain location.
560 The directories specified by the environment variable @code{LIBRARY_PATH}
561 (or port-specific name; native only, cross compilers do not use this).
564 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
565 in the configured @var{prefix} or this is a native compiler.
568 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
571 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
575 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
576 native compiler, or we have a target system root.
579 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
580 native compiler, or we have a target system root.
583 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
584 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
585 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
588 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
589 compiler, or we have a target system root. The default for this macro is
593 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
594 compiler, or we have a target system root. The default for this macro is
598 @node Run-time Target
599 @section Run-time Target Specification
600 @cindex run-time target specification
601 @cindex predefined macros
602 @cindex target specifications
604 @c prevent bad page break with this line
605 Here are run-time target specifications.
607 @defmac TARGET_CPU_CPP_BUILTINS ()
608 This function-like macro expands to a block of code that defines
609 built-in preprocessor macros and assertions for the target CPU, using
610 the functions @code{builtin_define}, @code{builtin_define_std} and
611 @code{builtin_assert}. When the front end
612 calls this macro it provides a trailing semicolon, and since it has
613 finished command line option processing your code can use those
616 @code{builtin_assert} takes a string in the form you pass to the
617 command-line option @option{-A}, such as @code{cpu=mips}, and creates
618 the assertion. @code{builtin_define} takes a string in the form
619 accepted by option @option{-D} and unconditionally defines the macro.
621 @code{builtin_define_std} takes a string representing the name of an
622 object-like macro. If it doesn't lie in the user's namespace,
623 @code{builtin_define_std} defines it unconditionally. Otherwise, it
624 defines a version with two leading underscores, and another version
625 with two leading and trailing underscores, and defines the original
626 only if an ISO standard was not requested on the command line. For
627 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
628 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
629 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
630 defines only @code{_ABI64}.
632 You can also test for the C dialect being compiled. The variable
633 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
634 or @code{clk_objective_c}. Note that if we are preprocessing
635 assembler, this variable will be @code{clk_c} but the function-like
636 macro @code{preprocessing_asm_p()} will return true, so you might want
637 to check for that first. If you need to check for strict ANSI, the
638 variable @code{flag_iso} can be used. The function-like macro
639 @code{preprocessing_trad_p()} can be used to check for traditional
643 @defmac TARGET_OS_CPP_BUILTINS ()
644 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
645 and is used for the target operating system instead.
648 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
649 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
650 and is used for the target object format. @file{elfos.h} uses this
651 macro to define @code{__ELF__}, so you probably do not need to define
655 @deftypevar {extern int} target_flags
656 This variable is declared in @file{options.h}, which is included before
657 any target-specific headers.
660 @deftypevr {Common Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
661 This variable specifies the initial value of @code{target_flags}.
662 Its default setting is 0.
665 @cindex optional hardware or system features
666 @cindex features, optional, in system conventions
668 @deftypefn {Common Target Hook} bool TARGET_HANDLE_OPTION (struct gcc_options *@var{opts}, struct gcc_options *@var{opts_set}, const struct cl_decoded_option *@var{decoded}, location_t @var{loc})
669 This hook is called whenever the user specifies one of the
670 target-specific options described by the @file{.opt} definition files
671 (@pxref{Options}). It has the opportunity to do some option-specific
672 processing and should return true if the option is valid. The default
673 definition does nothing but return true.
675 @var{decoded} specifies the option and its arguments. @var{opts} and
676 @var{opts_set} are the @code{gcc_options} structures to be used for
677 storing option state, and @var{loc} is the location at which the
678 option was passed (@code{UNKNOWN_LOCATION} except for options passed
682 @deftypefn {C Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
683 This target hook is called whenever the user specifies one of the
684 target-specific C language family options described by the @file{.opt}
685 definition files(@pxref{Options}). It has the opportunity to do some
686 option-specific processing and should return true if the option is
687 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
688 default definition does nothing but return false.
690 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
691 options. However, if processing an option requires routines that are
692 only available in the C (and related language) front ends, then you
693 should use @code{TARGET_HANDLE_C_OPTION} instead.
696 @deftypefn {C Target Hook} tree TARGET_OBJC_CONSTRUCT_STRING_OBJECT (tree @var{string})
697 Targets may provide a string object type that can be used within and between C, C++ and their respective Objective-C dialects. A string object might, for example, embed encoding and length information. These objects are considered opaque to the compiler and handled as references. An ideal implementation makes the composition of the string object match that of the Objective-C @code{NSString} (@code{NXString} for GNUStep), allowing efficient interworking between C-only and Objective-C code. If a target implements string objects then this hook should return a reference to such an object constructed from the normal `C' string representation provided in @var{string}. At present, the hook is used by Objective-C only, to obtain a common-format string object when the target provides one.
700 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE (const char *@var{classname})
701 Declare that Objective C class @var{classname} is referenced by the current TU.
704 @deftypefn {C Target Hook} void TARGET_OBJC_DECLARE_CLASS_DEFINITION (const char *@var{classname})
705 Declare that Objective C class @var{classname} is defined by the current TU.
708 @deftypefn {C Target Hook} bool TARGET_STRING_OBJECT_REF_TYPE_P (const_tree @var{stringref})
709 If a target implements string objects then this hook should return @code{true} if @var{stringref} is a valid reference to such an object.
712 @deftypefn {C Target Hook} void TARGET_CHECK_STRING_OBJECT_FORMAT_ARG (tree @var{format_arg}, tree @var{args_list})
713 If a target implements string objects then this hook should should provide a facility to check the function arguments in @var{args_list} against the format specifiers in @var{format_arg} where the type of @var{format_arg} is one recognized as a valid string reference type.
716 @deftypefn {Target Hook} void TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE (void)
717 This target function is similar to the hook @code{TARGET_OPTION_OVERRIDE}
718 but is called when the optimize level is changed via an attribute or
719 pragma or when it is reset at the end of the code affected by the
720 attribute or pragma. It is not called at the beginning of compilation
721 when @code{TARGET_OPTION_OVERRIDE} is called so if you want to perform these
722 actions then, you should have @code{TARGET_OPTION_OVERRIDE} call
723 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}.
726 @defmac C_COMMON_OVERRIDE_OPTIONS
727 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
728 but is only used in the C
729 language frontends (C, Objective-C, C++, Objective-C++) and so can be
730 used to alter option flag variables which only exist in those
734 @deftypevr {Common Target Hook} {const struct default_options *} TARGET_OPTION_OPTIMIZATION_TABLE
735 Some machines may desire to change what optimizations are performed for
736 various optimization levels. This variable, if defined, describes
737 options to enable at particular sets of optimization levels. These
738 options are processed once
739 just after the optimization level is determined and before the remainder
740 of the command options have been parsed, so may be overridden by other
741 options passed explicitly.
743 This processing is run once at program startup and when the optimization
744 options are changed via @code{#pragma GCC optimize} or by using the
745 @code{optimize} attribute.
748 @deftypefn {Common Target Hook} void TARGET_OPTION_INIT_STRUCT (struct gcc_options *@var{opts})
749 Set target-dependent initial values of fields in @var{opts}.
752 @deftypefn {Common Target Hook} void TARGET_OPTION_DEFAULT_PARAMS (void)
753 Set target-dependent default values for @option{--param} settings, using calls to @code{set_default_param_value}.
756 @defmac SWITCHABLE_TARGET
757 Some targets need to switch between substantially different subtargets
758 during compilation. For example, the MIPS target has one subtarget for
759 the traditional MIPS architecture and another for MIPS16. Source code
760 can switch between these two subarchitectures using the @code{mips16}
761 and @code{nomips16} attributes.
763 Such subtargets can differ in things like the set of available
764 registers, the set of available instructions, the costs of various
765 operations, and so on. GCC caches a lot of this type of information
766 in global variables, and recomputing them for each subtarget takes a
767 significant amount of time. The compiler therefore provides a facility
768 for maintaining several versions of the global variables and quickly
769 switching between them; see @file{target-globals.h} for details.
771 Define this macro to 1 if your target needs this facility. The default
775 @deftypefn {Target Hook} bool TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P (void)
776 Returns true if the target supports IEEE 754 floating-point exceptions and rounding modes, false otherwise. This is intended to relate to the @code{float} and @code{double} types, but not necessarily @code{long double}. By default, returns true if the @code{adddf3} instruction pattern is available and false otherwise, on the assumption that hardware floating point supports exceptions and rounding modes but software floating point does not.
779 @node Per-Function Data
780 @section Defining data structures for per-function information.
781 @cindex per-function data
782 @cindex data structures
784 If the target needs to store information on a per-function basis, GCC
785 provides a macro and a couple of variables to allow this. Note, just
786 using statics to store the information is a bad idea, since GCC supports
787 nested functions, so you can be halfway through encoding one function
788 when another one comes along.
790 GCC defines a data structure called @code{struct function} which
791 contains all of the data specific to an individual function. This
792 structure contains a field called @code{machine} whose type is
793 @code{struct machine_function *}, which can be used by targets to point
794 to their own specific data.
796 If a target needs per-function specific data it should define the type
797 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
798 This macro should be used to initialize the function pointer
799 @code{init_machine_status}. This pointer is explained below.
801 One typical use of per-function, target specific data is to create an
802 RTX to hold the register containing the function's return address. This
803 RTX can then be used to implement the @code{__builtin_return_address}
804 function, for level 0.
806 Note---earlier implementations of GCC used a single data area to hold
807 all of the per-function information. Thus when processing of a nested
808 function began the old per-function data had to be pushed onto a
809 stack, and when the processing was finished, it had to be popped off the
810 stack. GCC used to provide function pointers called
811 @code{save_machine_status} and @code{restore_machine_status} to handle
812 the saving and restoring of the target specific information. Since the
813 single data area approach is no longer used, these pointers are no
816 @defmac INIT_EXPANDERS
817 Macro called to initialize any target specific information. This macro
818 is called once per function, before generation of any RTL has begun.
819 The intention of this macro is to allow the initialization of the
820 function pointer @code{init_machine_status}.
823 @deftypevar {void (*)(struct function *)} init_machine_status
824 If this function pointer is non-@code{NULL} it will be called once per
825 function, before function compilation starts, in order to allow the
826 target to perform any target specific initialization of the
827 @code{struct function} structure. It is intended that this would be
828 used to initialize the @code{machine} of that structure.
830 @code{struct machine_function} structures are expected to be freed by GC@.
831 Generally, any memory that they reference must be allocated by using
832 GC allocation, including the structure itself.
836 @section Storage Layout
837 @cindex storage layout
839 Note that the definitions of the macros in this table which are sizes or
840 alignments measured in bits do not need to be constant. They can be C
841 expressions that refer to static variables, such as the @code{target_flags}.
842 @xref{Run-time Target}.
844 @defmac BITS_BIG_ENDIAN
845 Define this macro to have the value 1 if the most significant bit in a
846 byte has the lowest number; otherwise define it to have the value zero.
847 This means that bit-field instructions count from the most significant
848 bit. If the machine has no bit-field instructions, then this must still
849 be defined, but it doesn't matter which value it is defined to. This
850 macro need not be a constant.
852 This macro does not affect the way structure fields are packed into
853 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
856 @defmac BYTES_BIG_ENDIAN
857 Define this macro to have the value 1 if the most significant byte in a
858 word has the lowest number. This macro need not be a constant.
861 @defmac WORDS_BIG_ENDIAN
862 Define this macro to have the value 1 if, in a multiword object, the
863 most significant word has the lowest number. This applies to both
864 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
865 order of words in memory is not the same as the order in registers. This
866 macro need not be a constant.
869 @defmac REG_WORDS_BIG_ENDIAN
870 On some machines, the order of words in a multiword object differs between
871 registers in memory. In such a situation, define this macro to describe
872 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
873 the order of words in memory.
876 @defmac FLOAT_WORDS_BIG_ENDIAN
877 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
878 @code{TFmode} floating point numbers are stored in memory with the word
879 containing the sign bit at the lowest address; otherwise define it to
880 have the value 0. This macro need not be a constant.
882 You need not define this macro if the ordering is the same as for
886 @defmac BITS_PER_WORD
887 Number of bits in a word. If you do not define this macro, the default
888 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
891 @defmac MAX_BITS_PER_WORD
892 Maximum number of bits in a word. If this is undefined, the default is
893 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
894 largest value that @code{BITS_PER_WORD} can have at run-time.
897 @defmac UNITS_PER_WORD
898 Number of storage units in a word; normally the size of a general-purpose
899 register, a power of two from 1 or 8.
902 @defmac MIN_UNITS_PER_WORD
903 Minimum number of units in a word. If this is undefined, the default is
904 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
905 smallest value that @code{UNITS_PER_WORD} can have at run-time.
909 Width of a pointer, in bits. You must specify a value no wider than the
910 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
911 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
912 a value the default is @code{BITS_PER_WORD}.
915 @defmac POINTERS_EXTEND_UNSIGNED
916 A C expression that determines how pointers should be extended from
917 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
918 greater than zero if pointers should be zero-extended, zero if they
919 should be sign-extended, and negative if some other sort of conversion
920 is needed. In the last case, the extension is done by the target's
921 @code{ptr_extend} instruction.
923 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
924 and @code{word_mode} are all the same width.
927 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
928 A macro to update @var{m} and @var{unsignedp} when an object whose type
929 is @var{type} and which has the specified mode and signedness is to be
930 stored in a register. This macro is only called when @var{type} is a
933 On most RISC machines, which only have operations that operate on a full
934 register, define this macro to set @var{m} to @code{word_mode} if
935 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
936 cases, only integer modes should be widened because wider-precision
937 floating-point operations are usually more expensive than their narrower
940 For most machines, the macro definition does not change @var{unsignedp}.
941 However, some machines, have instructions that preferentially handle
942 either signed or unsigned quantities of certain modes. For example, on
943 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
944 sign-extend the result to 64 bits. On such machines, set
945 @var{unsignedp} according to which kind of extension is more efficient.
947 Do not define this macro if it would never modify @var{m}.
950 @deftypefn {Target Hook} 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})
951 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
952 function return values. The target hook should return the new mode
953 and possibly change @code{*@var{punsignedp}} if the promotion should
954 change signedness. This function is called only for scalar @emph{or
957 @var{for_return} allows to distinguish the promotion of arguments and
958 return values. If it is @code{1}, a return value is being promoted and
959 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
960 If it is @code{2}, the returned mode should be that of the register in
961 which an incoming parameter is copied, or the outgoing result is computed;
962 then the hook should return the same mode as @code{promote_mode}, though
963 the signedness may be different.
965 @var{type} can be NULL when promoting function arguments of libcalls.
967 The default is to not promote arguments and return values. You can
968 also define the hook to @code{default_promote_function_mode_always_promote}
969 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
972 @defmac PARM_BOUNDARY
973 Normal alignment required for function parameters on the stack, in
974 bits. All stack parameters receive at least this much alignment
975 regardless of data type. On most machines, this is the same as the
979 @defmac STACK_BOUNDARY
980 Define this macro to the minimum alignment enforced by hardware for the
981 stack pointer on this machine. The definition is a C expression for the
982 desired alignment (measured in bits). This value is used as a default
983 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
984 this should be the same as @code{PARM_BOUNDARY}.
987 @defmac PREFERRED_STACK_BOUNDARY
988 Define this macro if you wish to preserve a certain alignment for the
989 stack pointer, greater than what the hardware enforces. The definition
990 is a C expression for the desired alignment (measured in bits). This
991 macro must evaluate to a value equal to or larger than
992 @code{STACK_BOUNDARY}.
995 @defmac INCOMING_STACK_BOUNDARY
996 Define this macro if the incoming stack boundary may be different
997 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
998 to a value equal to or larger than @code{STACK_BOUNDARY}.
1001 @defmac FUNCTION_BOUNDARY
1002 Alignment required for a function entry point, in bits.
1005 @defmac BIGGEST_ALIGNMENT
1006 Biggest alignment that any data type can require on this machine, in
1007 bits. Note that this is not the biggest alignment that is supported,
1008 just the biggest alignment that, when violated, may cause a fault.
1011 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_ABSOLUTE_BIGGEST_ALIGNMENT
1012 If defined, this target hook specifies the absolute biggest alignment
1013 that a type or variable can have on this machine, otherwise,
1014 @code{BIGGEST_ALIGNMENT} is used.
1017 @defmac MALLOC_ABI_ALIGNMENT
1018 Alignment, in bits, a C conformant malloc implementation has to
1019 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1022 @defmac ATTRIBUTE_ALIGNED_VALUE
1023 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1024 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1027 @defmac MINIMUM_ATOMIC_ALIGNMENT
1028 If defined, the smallest alignment, in bits, that can be given to an
1029 object that can be referenced in one operation, without disturbing any
1030 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1031 on machines that don't have byte or half-word store operations.
1034 @defmac BIGGEST_FIELD_ALIGNMENT
1035 Biggest alignment that any structure or union field can require on this
1036 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1037 structure and union fields only, unless the field alignment has been set
1038 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1041 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1042 An expression for the alignment of a structure field @var{field} if the
1043 alignment computed in the usual way (including applying of
1044 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1045 alignment) is @var{computed}. It overrides alignment only if the
1046 field alignment has not been set by the
1047 @code{__attribute__ ((aligned (@var{n})))} construct.
1050 @defmac MAX_STACK_ALIGNMENT
1051 Biggest stack alignment guaranteed by the backend. Use this macro
1052 to specify the maximum alignment of a variable on stack.
1054 If not defined, the default value is @code{STACK_BOUNDARY}.
1056 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1057 @c But the fix for PR 32893 indicates that we can only guarantee
1058 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1059 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1062 @defmac MAX_OFILE_ALIGNMENT
1063 Biggest alignment supported by the object file format of this machine.
1064 Use this macro to limit the alignment which can be specified using the
1065 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1066 the default value is @code{BIGGEST_ALIGNMENT}.
1068 On systems that use ELF, the default (in @file{config/elfos.h}) is
1069 the largest supported 32-bit ELF section alignment representable on
1070 a 32-bit host e.g. @samp{(((uint64_t) 1 << 28) * 8)}.
1071 On 32-bit ELF the largest supported section alignment in bits is
1072 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1075 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1076 If defined, a C expression to compute the alignment for a variable in
1077 the static store. @var{type} is the data type, and @var{basic-align} is
1078 the alignment that the object would ordinarily have. The value of this
1079 macro is used instead of that alignment to align the object.
1081 If this macro is not defined, then @var{basic-align} is used.
1084 One use of this macro is to increase alignment of medium-size data to
1085 make it all fit in fewer cache lines. Another is to cause character
1086 arrays to be word-aligned so that @code{strcpy} calls that copy
1087 constants to character arrays can be done inline.
1090 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1091 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1092 some alignment increase, instead of optimization only purposes. E.g.@
1093 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1094 must be aligned to 16 byte boundaries.
1096 If this macro is not defined, then @var{basic-align} is used.
1099 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1100 If defined, a C expression to compute the alignment given to a constant
1101 that is being placed in memory. @var{constant} is the constant and
1102 @var{basic-align} is the alignment that the object would ordinarily
1103 have. The value of this macro is used instead of that alignment to
1106 The default definition just returns @var{basic-align}.
1108 The typical use of this macro is to increase alignment for string
1109 constants to be word aligned so that @code{strcpy} calls that copy
1110 constants can be done inline.
1113 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1114 If defined, a C expression to compute the alignment for a variable in
1115 the local store. @var{type} is the data type, and @var{basic-align} is
1116 the alignment that the object would ordinarily have. The value of this
1117 macro is used instead of that alignment to align the object.
1119 If this macro is not defined, then @var{basic-align} is used.
1121 One use of this macro is to increase alignment of medium-size data to
1122 make it all fit in fewer cache lines.
1124 If the value of this macro has a type, it should be an unsigned type.
1127 @deftypefn {Target Hook} HOST_WIDE_INT TARGET_VECTOR_ALIGNMENT (const_tree @var{type})
1128 This hook can be used to define the alignment for a vector of type
1129 @var{type}, in order to comply with a platform ABI. The default is to
1130 require natural alignment for vector types. The alignment returned by
1131 this hook must be a power-of-two multiple of the default alignment of
1132 the vector element type.
1135 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1136 If defined, a C expression to compute the alignment for stack slot.
1137 @var{type} is the data type, @var{mode} is the widest mode available,
1138 and @var{basic-align} is the alignment that the slot would ordinarily
1139 have. The value of this macro is used instead of that alignment to
1142 If this macro is not defined, then @var{basic-align} is used when
1143 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1146 This macro is to set alignment of stack slot to the maximum alignment
1147 of all possible modes which the slot may have.
1149 If the value of this macro has a type, it should be an unsigned type.
1152 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1153 If defined, a C expression to compute the alignment for a local
1154 variable @var{decl}.
1156 If this macro is not defined, then
1157 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1160 One use of this macro is to increase alignment of medium-size data to
1161 make it all fit in fewer cache lines.
1163 If the value of this macro has a type, it should be an unsigned type.
1166 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1167 If defined, a C expression to compute the minimum required alignment
1168 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1169 @var{mode}, assuming normal alignment @var{align}.
1171 If this macro is not defined, then @var{align} will be used.
1174 @defmac EMPTY_FIELD_BOUNDARY
1175 Alignment in bits to be given to a structure bit-field that follows an
1176 empty field such as @code{int : 0;}.
1178 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1181 @defmac STRUCTURE_SIZE_BOUNDARY
1182 Number of bits which any structure or union's size must be a multiple of.
1183 Each structure or union's size is rounded up to a multiple of this.
1185 If you do not define this macro, the default is the same as
1186 @code{BITS_PER_UNIT}.
1189 @defmac STRICT_ALIGNMENT
1190 Define this macro to be the value 1 if instructions will fail to work
1191 if given data not on the nominal alignment. If instructions will merely
1192 go slower in that case, define this macro as 0.
1195 @defmac PCC_BITFIELD_TYPE_MATTERS
1196 Define this if you wish to imitate the way many other C compilers handle
1197 alignment of bit-fields and the structures that contain them.
1199 The behavior is that the type written for a named bit-field (@code{int},
1200 @code{short}, or other integer type) imposes an alignment for the entire
1201 structure, as if the structure really did contain an ordinary field of
1202 that type. In addition, the bit-field is placed within the structure so
1203 that it would fit within such a field, not crossing a boundary for it.
1205 Thus, on most machines, a named bit-field whose type is written as
1206 @code{int} would not cross a four-byte boundary, and would force
1207 four-byte alignment for the whole structure. (The alignment used may
1208 not be four bytes; it is controlled by the other alignment parameters.)
1210 An unnamed bit-field will not affect the alignment of the containing
1213 If the macro is defined, its definition should be a C expression;
1214 a nonzero value for the expression enables this behavior.
1216 Note that if this macro is not defined, or its value is zero, some
1217 bit-fields may cross more than one alignment boundary. The compiler can
1218 support such references if there are @samp{insv}, @samp{extv}, and
1219 @samp{extzv} insns that can directly reference memory.
1221 The other known way of making bit-fields work is to define
1222 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1223 Then every structure can be accessed with fullwords.
1225 Unless the machine has bit-field instructions or you define
1226 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1227 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1229 If your aim is to make GCC use the same conventions for laying out
1230 bit-fields as are used by another compiler, here is how to investigate
1231 what the other compiler does. Compile and run this program:
1250 printf ("Size of foo1 is %d\n",
1251 sizeof (struct foo1));
1252 printf ("Size of foo2 is %d\n",
1253 sizeof (struct foo2));
1258 If this prints 2 and 5, then the compiler's behavior is what you would
1259 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1262 @defmac BITFIELD_NBYTES_LIMITED
1263 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1264 to aligning a bit-field within the structure.
1267 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1268 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1269 whether unnamed bitfields affect the alignment of the containing
1270 structure. The hook should return true if the structure should inherit
1271 the alignment requirements of an unnamed bitfield's type.
1274 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1275 This target hook should return @code{true} if accesses to volatile bitfields
1276 should use the narrowest mode possible. It should return @code{false} if
1277 these accesses should use the bitfield container type.
1279 The default is @code{false}.
1282 @deftypefn {Target Hook} bool TARGET_MEMBER_TYPE_FORCES_BLK (const_tree @var{field}, machine_mode @var{mode})
1283 Return true if a structure, union or array containing @var{field} should
1284 be accessed using @code{BLKMODE}.
1286 If @var{field} is the only field in the structure, @var{mode} is its
1287 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1288 case where structures of one field would require the structure's mode to
1289 retain the field's mode.
1291 Normally, this is not needed.
1294 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1295 Define this macro as an expression for the alignment of a type (given
1296 by @var{type} as a tree node) if the alignment computed in the usual
1297 way is @var{computed} and the alignment explicitly specified was
1300 The default is to use @var{specified} if it is larger; otherwise, use
1301 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1304 @defmac MAX_FIXED_MODE_SIZE
1305 An integer expression for the size in bits of the largest integer
1306 machine mode that should actually be used. All integer machine modes of
1307 this size or smaller can be used for structures and unions with the
1308 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1309 (DImode)} is assumed.
1312 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1313 If defined, an expression of type @code{machine_mode} that
1314 specifies the mode of the save area operand of a
1315 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1316 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1317 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1318 having its mode specified.
1320 You need not define this macro if it always returns @code{Pmode}. You
1321 would most commonly define this macro if the
1322 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1326 @defmac STACK_SIZE_MODE
1327 If defined, an expression of type @code{machine_mode} that
1328 specifies the mode of the size increment operand of an
1329 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1331 You need not define this macro if it always returns @code{word_mode}.
1332 You would most commonly define this macro if the @code{allocate_stack}
1333 pattern needs to support both a 32- and a 64-bit mode.
1336 @deftypefn {Target Hook} machine_mode TARGET_LIBGCC_CMP_RETURN_MODE (void)
1337 This target hook should return the mode to be used for the return value
1338 of compare instructions expanded to libgcc calls. If not defined
1339 @code{word_mode} is returned which is the right choice for a majority of
1343 @deftypefn {Target Hook} machine_mode TARGET_LIBGCC_SHIFT_COUNT_MODE (void)
1344 This target hook should return the mode to be used for the shift count operand
1345 of shift instructions expanded to libgcc calls. If not defined
1346 @code{word_mode} is returned which is the right choice for a majority of
1350 @deftypefn {Target Hook} machine_mode TARGET_UNWIND_WORD_MODE (void)
1351 Return machine mode to be used for @code{_Unwind_Word} type.
1352 The default is to use @code{word_mode}.
1355 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (const_tree @var{record_type})
1356 This target hook returns @code{true} if bit-fields in the given
1357 @var{record_type} are to be laid out following the rules of Microsoft
1358 Visual C/C++, namely: (i) a bit-field won't share the same storage
1359 unit with the previous bit-field if their underlying types have
1360 different sizes, and the bit-field will be aligned to the highest
1361 alignment of the underlying types of itself and of the previous
1362 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1363 the whole enclosing structure, even if it is unnamed; except that
1364 (iii) a zero-sized bit-field will be disregarded unless it follows
1365 another bit-field of nonzero size. If this hook returns @code{true},
1366 other macros that control bit-field layout are ignored.
1368 When a bit-field is inserted into a packed record, the whole size
1369 of the underlying type is used by one or more same-size adjacent
1370 bit-fields (that is, if its long:3, 32 bits is used in the record,
1371 and any additional adjacent long bit-fields are packed into the same
1372 chunk of 32 bits. However, if the size changes, a new field of that
1373 size is allocated). In an unpacked record, this is the same as using
1374 alignment, but not equivalent when packing.
1376 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1377 the latter will take precedence. If @samp{__attribute__((packed))} is
1378 used on a single field when MS bit-fields are in use, it will take
1379 precedence for that field, but the alignment of the rest of the structure
1380 may affect its placement.
1383 @deftypefn {Target Hook} bool TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1384 Returns true if the target supports decimal floating point.
1387 @deftypefn {Target Hook} bool TARGET_FIXED_POINT_SUPPORTED_P (void)
1388 Returns true if the target supports fixed-point arithmetic.
1391 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1392 This hook is called just before expansion into rtl, allowing the target
1393 to perform additional initializations or analysis before the expansion.
1394 For example, the rs6000 port uses it to allocate a scratch stack slot
1395 for use in copying SDmode values between memory and floating point
1396 registers whenever the function being expanded has any SDmode
1400 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1401 This hook allows the backend to perform additional instantiations on rtl
1402 that are not actually in any insns yet, but will be later.
1405 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (const_tree @var{type})
1406 If your target defines any fundamental types, or any types your target
1407 uses should be mangled differently from the default, define this hook
1408 to return the appropriate encoding for these types as part of a C++
1409 mangled name. The @var{type} argument is the tree structure representing
1410 the type to be mangled. The hook may be applied to trees which are
1411 not target-specific fundamental types; it should return @code{NULL}
1412 for all such types, as well as arguments it does not recognize. If the
1413 return value is not @code{NULL}, it must point to a statically-allocated
1416 Target-specific fundamental types might be new fundamental types or
1417 qualified versions of ordinary fundamental types. Encode new
1418 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1419 is the name used for the type in source code, and @var{n} is the
1420 length of @var{name} in decimal. Encode qualified versions of
1421 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1422 @var{name} is the name used for the type qualifier in source code,
1423 @var{n} is the length of @var{name} as above, and @var{code} is the
1424 code used to represent the unqualified version of this type. (See
1425 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1426 codes.) In both cases the spaces are for clarity; do not include any
1427 spaces in your string.
1429 This hook is applied to types prior to typedef resolution. If the mangled
1430 name for a particular type depends only on that type's main variant, you
1431 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1434 The default version of this hook always returns @code{NULL}, which is
1435 appropriate for a target that does not define any new fundamental
1440 @section Layout of Source Language Data Types
1442 These macros define the sizes and other characteristics of the standard
1443 basic data types used in programs being compiled. Unlike the macros in
1444 the previous section, these apply to specific features of C and related
1445 languages, rather than to fundamental aspects of storage layout.
1447 @defmac INT_TYPE_SIZE
1448 A C expression for the size in bits of the type @code{int} on the
1449 target machine. If you don't define this, the default is one word.
1452 @defmac SHORT_TYPE_SIZE
1453 A C expression for the size in bits of the type @code{short} on the
1454 target machine. If you don't define this, the default is half a word.
1455 (If this would be less than one storage unit, it is rounded up to one
1459 @defmac LONG_TYPE_SIZE
1460 A C expression for the size in bits of the type @code{long} on the
1461 target machine. If you don't define this, the default is one word.
1464 @defmac ADA_LONG_TYPE_SIZE
1465 On some machines, the size used for the Ada equivalent of the type
1466 @code{long} by a native Ada compiler differs from that used by C@. In
1467 that situation, define this macro to be a C expression to be used for
1468 the size of that type. If you don't define this, the default is the
1469 value of @code{LONG_TYPE_SIZE}.
1472 @defmac LONG_LONG_TYPE_SIZE
1473 A C expression for the size in bits of the type @code{long long} on the
1474 target machine. If you don't define this, the default is two
1475 words. If you want to support GNU Ada on your machine, the value of this
1476 macro must be at least 64.
1479 @defmac CHAR_TYPE_SIZE
1480 A C expression for the size in bits of the type @code{char} on the
1481 target machine. If you don't define this, the default is
1482 @code{BITS_PER_UNIT}.
1485 @defmac BOOL_TYPE_SIZE
1486 A C expression for the size in bits of the C++ type @code{bool} and
1487 C99 type @code{_Bool} on the target machine. If you don't define
1488 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1491 @defmac FLOAT_TYPE_SIZE
1492 A C expression for the size in bits of the type @code{float} on the
1493 target machine. If you don't define this, the default is one word.
1496 @defmac DOUBLE_TYPE_SIZE
1497 A C expression for the size in bits of the type @code{double} on the
1498 target machine. If you don't define this, the default is two
1502 @defmac LONG_DOUBLE_TYPE_SIZE
1503 A C expression for the size in bits of the type @code{long double} on
1504 the target machine. If you don't define this, the default is two
1508 @defmac SHORT_FRACT_TYPE_SIZE
1509 A C expression for the size in bits of the type @code{short _Fract} on
1510 the target machine. If you don't define this, the default is
1511 @code{BITS_PER_UNIT}.
1514 @defmac FRACT_TYPE_SIZE
1515 A C expression for the size in bits of the type @code{_Fract} on
1516 the target machine. If you don't define this, the default is
1517 @code{BITS_PER_UNIT * 2}.
1520 @defmac LONG_FRACT_TYPE_SIZE
1521 A C expression for the size in bits of the type @code{long _Fract} on
1522 the target machine. If you don't define this, the default is
1523 @code{BITS_PER_UNIT * 4}.
1526 @defmac LONG_LONG_FRACT_TYPE_SIZE
1527 A C expression for the size in bits of the type @code{long long _Fract} on
1528 the target machine. If you don't define this, the default is
1529 @code{BITS_PER_UNIT * 8}.
1532 @defmac SHORT_ACCUM_TYPE_SIZE
1533 A C expression for the size in bits of the type @code{short _Accum} on
1534 the target machine. If you don't define this, the default is
1535 @code{BITS_PER_UNIT * 2}.
1538 @defmac ACCUM_TYPE_SIZE
1539 A C expression for the size in bits of the type @code{_Accum} on
1540 the target machine. If you don't define this, the default is
1541 @code{BITS_PER_UNIT * 4}.
1544 @defmac LONG_ACCUM_TYPE_SIZE
1545 A C expression for the size in bits of the type @code{long _Accum} on
1546 the target machine. If you don't define this, the default is
1547 @code{BITS_PER_UNIT * 8}.
1550 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1551 A C expression for the size in bits of the type @code{long long _Accum} on
1552 the target machine. If you don't define this, the default is
1553 @code{BITS_PER_UNIT * 16}.
1556 @defmac LIBGCC2_GNU_PREFIX
1557 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1558 hook and should be defined if that hook is overriden to be true. It
1559 causes function names in libgcc to be changed to use a @code{__gnu_}
1560 prefix for their name rather than the default @code{__}. A port which
1561 uses this macro should also arrange to use @file{t-gnu-prefix} in
1562 the libgcc @file{config.host}.
1565 @defmac TARGET_FLT_EVAL_METHOD
1566 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1567 assuming, if applicable, that the floating-point control word is in its
1568 default state. If you do not define this macro the value of
1569 @code{FLT_EVAL_METHOD} will be zero.
1572 @defmac WIDEST_HARDWARE_FP_SIZE
1573 A C expression for the size in bits of the widest floating-point format
1574 supported by the hardware. If you define this macro, you must specify a
1575 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1576 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1580 @defmac DEFAULT_SIGNED_CHAR
1581 An expression whose value is 1 or 0, according to whether the type
1582 @code{char} should be signed or unsigned by default. The user can
1583 always override this default with the options @option{-fsigned-char}
1584 and @option{-funsigned-char}.
1587 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1588 This target hook should return true if the compiler should give an
1589 @code{enum} type only as many bytes as it takes to represent the range
1590 of possible values of that type. It should return false if all
1591 @code{enum} types should be allocated like @code{int}.
1593 The default is to return false.
1597 A C expression for a string describing the name of the data type to use
1598 for size values. The typedef name @code{size_t} is defined using the
1599 contents of the string.
1601 The string can contain more than one keyword. If so, separate them with
1602 spaces, and write first any length keyword, then @code{unsigned} if
1603 appropriate, and finally @code{int}. The string must exactly match one
1604 of the data type names defined in the function
1605 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1606 You may not omit @code{int} or change the order---that would cause the
1607 compiler to crash on startup.
1609 If you don't define this macro, the default is @code{"long unsigned
1614 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1615 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1616 dealing with size. This macro is a C expression for a string describing
1617 the name of the data type from which the precision of @code{sizetype}
1620 The string has the same restrictions as @code{SIZE_TYPE} string.
1622 If you don't define this macro, the default is @code{SIZE_TYPE}.
1625 @defmac PTRDIFF_TYPE
1626 A C expression for a string describing the name of the data type to use
1627 for the result of subtracting two pointers. The typedef name
1628 @code{ptrdiff_t} is defined using the contents of the string. See
1629 @code{SIZE_TYPE} above for more information.
1631 If you don't define this macro, the default is @code{"long int"}.
1635 A C expression for a string describing the name of the data type to use
1636 for wide characters. The typedef name @code{wchar_t} is defined using
1637 the contents of the string. See @code{SIZE_TYPE} above for more
1640 If you don't define this macro, the default is @code{"int"}.
1643 @defmac WCHAR_TYPE_SIZE
1644 A C expression for the size in bits of the data type for wide
1645 characters. This is used in @code{cpp}, which cannot make use of
1650 A C expression for a string describing the name of the data type to
1651 use for wide characters passed to @code{printf} and returned from
1652 @code{getwc}. The typedef name @code{wint_t} is defined using the
1653 contents of the string. See @code{SIZE_TYPE} above for more
1656 If you don't define this macro, the default is @code{"unsigned int"}.
1660 A C expression for a string describing the name of the data type that
1661 can represent any value of any standard or extended signed integer type.
1662 The typedef name @code{intmax_t} is defined using the contents of the
1663 string. See @code{SIZE_TYPE} above for more information.
1665 If you don't define this macro, the default is the first of
1666 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1667 much precision as @code{long long int}.
1670 @defmac UINTMAX_TYPE
1671 A C expression for a string describing the name of the data type that
1672 can represent any value of any standard or extended unsigned integer
1673 type. The typedef name @code{uintmax_t} is defined using the contents
1674 of the string. See @code{SIZE_TYPE} above for more information.
1676 If you don't define this macro, the default is the first of
1677 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1678 unsigned int"} that has as much precision as @code{long long unsigned
1682 @defmac SIG_ATOMIC_TYPE
1688 @defmacx UINT16_TYPE
1689 @defmacx UINT32_TYPE
1690 @defmacx UINT64_TYPE
1691 @defmacx INT_LEAST8_TYPE
1692 @defmacx INT_LEAST16_TYPE
1693 @defmacx INT_LEAST32_TYPE
1694 @defmacx INT_LEAST64_TYPE
1695 @defmacx UINT_LEAST8_TYPE
1696 @defmacx UINT_LEAST16_TYPE
1697 @defmacx UINT_LEAST32_TYPE
1698 @defmacx UINT_LEAST64_TYPE
1699 @defmacx INT_FAST8_TYPE
1700 @defmacx INT_FAST16_TYPE
1701 @defmacx INT_FAST32_TYPE
1702 @defmacx INT_FAST64_TYPE
1703 @defmacx UINT_FAST8_TYPE
1704 @defmacx UINT_FAST16_TYPE
1705 @defmacx UINT_FAST32_TYPE
1706 @defmacx UINT_FAST64_TYPE
1707 @defmacx INTPTR_TYPE
1708 @defmacx UINTPTR_TYPE
1709 C expressions for the standard types @code{sig_atomic_t},
1710 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1711 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1712 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1713 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1714 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1715 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1716 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1717 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1718 @code{SIZE_TYPE} above for more information.
1720 If any of these macros evaluates to a null pointer, the corresponding
1721 type is not supported; if GCC is configured to provide
1722 @code{<stdint.h>} in such a case, the header provided may not conform
1723 to C99, depending on the type in question. The defaults for all of
1724 these macros are null pointers.
1727 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1728 The C++ compiler represents a pointer-to-member-function with a struct
1735 ptrdiff_t vtable_index;
1742 The C++ compiler must use one bit to indicate whether the function that
1743 will be called through a pointer-to-member-function is virtual.
1744 Normally, we assume that the low-order bit of a function pointer must
1745 always be zero. Then, by ensuring that the vtable_index is odd, we can
1746 distinguish which variant of the union is in use. But, on some
1747 platforms function pointers can be odd, and so this doesn't work. In
1748 that case, we use the low-order bit of the @code{delta} field, and shift
1749 the remainder of the @code{delta} field to the left.
1751 GCC will automatically make the right selection about where to store
1752 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1753 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1754 set such that functions always start at even addresses, but the lowest
1755 bit of pointers to functions indicate whether the function at that
1756 address is in ARM or Thumb mode. If this is the case of your
1757 architecture, you should define this macro to
1758 @code{ptrmemfunc_vbit_in_delta}.
1760 In general, you should not have to define this macro. On architectures
1761 in which function addresses are always even, according to
1762 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1763 @code{ptrmemfunc_vbit_in_pfn}.
1766 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1767 Normally, the C++ compiler uses function pointers in vtables. This
1768 macro allows the target to change to use ``function descriptors''
1769 instead. Function descriptors are found on targets for whom a
1770 function pointer is actually a small data structure. Normally the
1771 data structure consists of the actual code address plus a data
1772 pointer to which the function's data is relative.
1774 If vtables are used, the value of this macro should be the number
1775 of words that the function descriptor occupies.
1778 @defmac TARGET_VTABLE_ENTRY_ALIGN
1779 By default, the vtable entries are void pointers, the so the alignment
1780 is the same as pointer alignment. The value of this macro specifies
1781 the alignment of the vtable entry in bits. It should be defined only
1782 when special alignment is necessary. */
1785 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1786 There are a few non-descriptor entries in the vtable at offsets below
1787 zero. If these entries must be padded (say, to preserve the alignment
1788 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1789 of words in each data entry.
1793 @section Register Usage
1794 @cindex register usage
1796 This section explains how to describe what registers the target machine
1797 has, and how (in general) they can be used.
1799 The description of which registers a specific instruction can use is
1800 done with register classes; see @ref{Register Classes}. For information
1801 on using registers to access a stack frame, see @ref{Frame Registers}.
1802 For passing values in registers, see @ref{Register Arguments}.
1803 For returning values in registers, see @ref{Scalar Return}.
1806 * Register Basics:: Number and kinds of registers.
1807 * Allocation Order:: Order in which registers are allocated.
1808 * Values in Registers:: What kinds of values each reg can hold.
1809 * Leaf Functions:: Renumbering registers for leaf functions.
1810 * Stack Registers:: Handling a register stack such as 80387.
1813 @node Register Basics
1814 @subsection Basic Characteristics of Registers
1816 @c prevent bad page break with this line
1817 Registers have various characteristics.
1819 @defmac FIRST_PSEUDO_REGISTER
1820 Number of hardware registers known to the compiler. They receive
1821 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1822 pseudo register's number really is assigned the number
1823 @code{FIRST_PSEUDO_REGISTER}.
1826 @defmac FIXED_REGISTERS
1827 @cindex fixed register
1828 An initializer that says which registers are used for fixed purposes
1829 all throughout the compiled code and are therefore not available for
1830 general allocation. These would include the stack pointer, the frame
1831 pointer (except on machines where that can be used as a general
1832 register when no frame pointer is needed), the program counter on
1833 machines where that is considered one of the addressable registers,
1834 and any other numbered register with a standard use.
1836 This information is expressed as a sequence of numbers, separated by
1837 commas and surrounded by braces. The @var{n}th number is 1 if
1838 register @var{n} is fixed, 0 otherwise.
1840 The table initialized from this macro, and the table initialized by
1841 the following one, may be overridden at run time either automatically,
1842 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1843 the user with the command options @option{-ffixed-@var{reg}},
1844 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1847 @defmac CALL_USED_REGISTERS
1848 @cindex call-used register
1849 @cindex call-clobbered register
1850 @cindex call-saved register
1851 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1852 clobbered (in general) by function calls as well as for fixed
1853 registers. This macro therefore identifies the registers that are not
1854 available for general allocation of values that must live across
1857 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1858 automatically saves it on function entry and restores it on function
1859 exit, if the register is used within the function.
1862 @defmac CALL_REALLY_USED_REGISTERS
1863 @cindex call-used register
1864 @cindex call-clobbered register
1865 @cindex call-saved register
1866 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1867 that the entire set of @code{FIXED_REGISTERS} be included.
1868 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1869 This macro is optional. If not specified, it defaults to the value
1870 of @code{CALL_USED_REGISTERS}.
1873 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1874 @cindex call-used register
1875 @cindex call-clobbered register
1876 @cindex call-saved register
1877 A C expression that is nonzero if it is not permissible to store a
1878 value of mode @var{mode} in hard register number @var{regno} across a
1879 call without some part of it being clobbered. For most machines this
1880 macro need not be defined. It is only required for machines that do not
1881 preserve the entire contents of a register across a call.
1885 @findex call_used_regs
1888 @findex reg_class_contents
1889 @deftypefn {Target Hook} void TARGET_CONDITIONAL_REGISTER_USAGE (void)
1890 This hook may conditionally modify five variables
1891 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1892 @code{reg_names}, and @code{reg_class_contents}, to take into account
1893 any dependence of these register sets on target flags. The first three
1894 of these are of type @code{char []} (interpreted as Boolean vectors).
1895 @code{global_regs} is a @code{const char *[]}, and
1896 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1897 called, @code{fixed_regs}, @code{call_used_regs},
1898 @code{reg_class_contents}, and @code{reg_names} have been initialized
1899 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
1900 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
1901 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
1902 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
1903 command options have been applied.
1905 @cindex disabling certain registers
1906 @cindex controlling register usage
1907 If the usage of an entire class of registers depends on the target
1908 flags, you may indicate this to GCC by using this macro to modify
1909 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
1910 registers in the classes which should not be used by GCC@. Also make
1911 @code{define_register_constraint}s return @code{NO_REGS} for constraints
1912 that shouldn't be used.
1914 (However, if this class is not included in @code{GENERAL_REGS} and all
1915 of the insn patterns whose constraints permit this class are
1916 controlled by target switches, then GCC will automatically avoid using
1917 these registers when the target switches are opposed to them.)
1920 @defmac INCOMING_REGNO (@var{out})
1921 Define this macro if the target machine has register windows. This C
1922 expression returns the register number as seen by the called function
1923 corresponding to the register number @var{out} as seen by the calling
1924 function. Return @var{out} if register number @var{out} is not an
1928 @defmac OUTGOING_REGNO (@var{in})
1929 Define this macro if the target machine has register windows. This C
1930 expression returns the register number as seen by the calling function
1931 corresponding to the register number @var{in} as seen by the called
1932 function. Return @var{in} if register number @var{in} is not an inbound
1936 @defmac LOCAL_REGNO (@var{regno})
1937 Define this macro if the target machine has register windows. This C
1938 expression returns true if the register is call-saved but is in the
1939 register window. Unlike most call-saved registers, such registers
1940 need not be explicitly restored on function exit or during non-local
1945 If the program counter has a register number, define this as that
1946 register number. Otherwise, do not define it.
1949 @node Allocation Order
1950 @subsection Order of Allocation of Registers
1951 @cindex order of register allocation
1952 @cindex register allocation order
1954 @c prevent bad page break with this line
1955 Registers are allocated in order.
1957 @defmac REG_ALLOC_ORDER
1958 If defined, an initializer for a vector of integers, containing the
1959 numbers of hard registers in the order in which GCC should prefer
1960 to use them (from most preferred to least).
1962 If this macro is not defined, registers are used lowest numbered first
1963 (all else being equal).
1965 One use of this macro is on machines where the highest numbered
1966 registers must always be saved and the save-multiple-registers
1967 instruction supports only sequences of consecutive registers. On such
1968 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1969 the highest numbered allocable register first.
1972 @defmac ADJUST_REG_ALLOC_ORDER
1973 A C statement (sans semicolon) to choose the order in which to allocate
1974 hard registers for pseudo-registers local to a basic block.
1976 Store the desired register order in the array @code{reg_alloc_order}.
1977 Element 0 should be the register to allocate first; element 1, the next
1978 register; and so on.
1980 The macro body should not assume anything about the contents of
1981 @code{reg_alloc_order} before execution of the macro.
1983 On most machines, it is not necessary to define this macro.
1986 @defmac HONOR_REG_ALLOC_ORDER
1987 Normally, IRA tries to estimate the costs for saving a register in the
1988 prologue and restoring it in the epilogue. This discourages it from
1989 using call-saved registers. If a machine wants to ensure that IRA
1990 allocates registers in the order given by REG_ALLOC_ORDER even if some
1991 call-saved registers appear earlier than call-used ones, then define this
1992 macro as a C expression to nonzero. Default is 0.
1995 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
1996 In some case register allocation order is not enough for the
1997 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
1998 If this macro is defined, it should return a floating point value
1999 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2000 be increased by approximately the pseudo's usage frequency times the
2001 value returned by this macro. Not defining this macro is equivalent
2002 to having it always return @code{0.0}.
2004 On most machines, it is not necessary to define this macro.
2007 @node Values in Registers
2008 @subsection How Values Fit in Registers
2010 This section discusses the macros that describe which kinds of values
2011 (specifically, which machine modes) each register can hold, and how many
2012 consecutive registers are needed for a given mode.
2014 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2015 A C expression for the number of consecutive hard registers, starting
2016 at register number @var{regno}, required to hold a value of mode
2017 @var{mode}. This macro must never return zero, even if a register
2018 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2019 and/or CANNOT_CHANGE_MODE_CLASS instead.
2021 On a machine where all registers are exactly one word, a suitable
2022 definition of this macro is
2025 #define HARD_REGNO_NREGS(REGNO, MODE) \
2026 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2031 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2032 A C expression that is nonzero if a value of mode @var{mode}, stored
2033 in memory, ends with padding that causes it to take up more space than
2034 in registers starting at register number @var{regno} (as determined by
2035 multiplying GCC's notion of the size of the register when containing
2036 this mode by the number of registers returned by
2037 @code{HARD_REGNO_NREGS}). By default this is zero.
2039 For example, if a floating-point value is stored in three 32-bit
2040 registers but takes up 128 bits in memory, then this would be
2043 This macros only needs to be defined if there are cases where
2044 @code{subreg_get_info}
2045 would otherwise wrongly determine that a @code{subreg} can be
2046 represented by an offset to the register number, when in fact such a
2047 @code{subreg} would contain some of the padding not stored in
2048 registers and so not be representable.
2051 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2052 For values of @var{regno} and @var{mode} for which
2053 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2054 returning the greater number of registers required to hold the value
2055 including any padding. In the example above, the value would be four.
2058 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2059 Define this macro if the natural size of registers that hold values
2060 of mode @var{mode} is not the word size. It is a C expression that
2061 should give the natural size in bytes for the specified mode. It is
2062 used by the register allocator to try to optimize its results. This
2063 happens for example on SPARC 64-bit where the natural size of
2064 floating-point registers is still 32-bit.
2067 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2068 A C expression that is nonzero if it is permissible to store a value
2069 of mode @var{mode} in hard register number @var{regno} (or in several
2070 registers starting with that one). For a machine where all registers
2071 are equivalent, a suitable definition is
2074 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2077 You need not include code to check for the numbers of fixed registers,
2078 because the allocation mechanism considers them to be always occupied.
2080 @cindex register pairs
2081 On some machines, double-precision values must be kept in even/odd
2082 register pairs. You can implement that by defining this macro to reject
2083 odd register numbers for such modes.
2085 The minimum requirement for a mode to be OK in a register is that the
2086 @samp{mov@var{mode}} instruction pattern support moves between the
2087 register and other hard register in the same class and that moving a
2088 value into the register and back out not alter it.
2090 Since the same instruction used to move @code{word_mode} will work for
2091 all narrower integer modes, it is not necessary on any machine for
2092 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2093 you define patterns @samp{movhi}, etc., to take advantage of this. This
2094 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2095 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2098 Many machines have special registers for floating point arithmetic.
2099 Often people assume that floating point machine modes are allowed only
2100 in floating point registers. This is not true. Any registers that
2101 can hold integers can safely @emph{hold} a floating point machine
2102 mode, whether or not floating arithmetic can be done on it in those
2103 registers. Integer move instructions can be used to move the values.
2105 On some machines, though, the converse is true: fixed-point machine
2106 modes may not go in floating registers. This is true if the floating
2107 registers normalize any value stored in them, because storing a
2108 non-floating value there would garble it. In this case,
2109 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2110 floating registers. But if the floating registers do not automatically
2111 normalize, if you can store any bit pattern in one and retrieve it
2112 unchanged without a trap, then any machine mode may go in a floating
2113 register, so you can define this macro to say so.
2115 The primary significance of special floating registers is rather that
2116 they are the registers acceptable in floating point arithmetic
2117 instructions. However, this is of no concern to
2118 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2119 constraints for those instructions.
2121 On some machines, the floating registers are especially slow to access,
2122 so that it is better to store a value in a stack frame than in such a
2123 register if floating point arithmetic is not being done. As long as the
2124 floating registers are not in class @code{GENERAL_REGS}, they will not
2125 be used unless some pattern's constraint asks for one.
2128 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2129 A C expression that is nonzero if it is OK to rename a hard register
2130 @var{from} to another hard register @var{to}.
2132 One common use of this macro is to prevent renaming of a register to
2133 another register that is not saved by a prologue in an interrupt
2136 The default is always nonzero.
2139 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2140 A C expression that is nonzero if a value of mode
2141 @var{mode1} is accessible in mode @var{mode2} without copying.
2143 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2144 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2145 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2146 should be nonzero. If they differ for any @var{r}, you should define
2147 this macro to return zero unless some other mechanism ensures the
2148 accessibility of the value in a narrower mode.
2150 You should define this macro to return nonzero in as many cases as
2151 possible since doing so will allow GCC to perform better register
2155 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2156 This target hook should return @code{true} if it is OK to use a hard register
2157 @var{regno} as scratch reg in peephole2.
2159 One common use of this macro is to prevent using of a register that
2160 is not saved by a prologue in an interrupt handler.
2162 The default version of this hook always returns @code{true}.
2165 @defmac AVOID_CCMODE_COPIES
2166 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2167 registers. You should only define this macro if support for copying to/from
2168 @code{CCmode} is incomplete.
2171 @node Leaf Functions
2172 @subsection Handling Leaf Functions
2174 @cindex leaf functions
2175 @cindex functions, leaf
2176 On some machines, a leaf function (i.e., one which makes no calls) can run
2177 more efficiently if it does not make its own register window. Often this
2178 means it is required to receive its arguments in the registers where they
2179 are passed by the caller, instead of the registers where they would
2182 The special treatment for leaf functions generally applies only when
2183 other conditions are met; for example, often they may use only those
2184 registers for its own variables and temporaries. We use the term ``leaf
2185 function'' to mean a function that is suitable for this special
2186 handling, so that functions with no calls are not necessarily ``leaf
2189 GCC assigns register numbers before it knows whether the function is
2190 suitable for leaf function treatment. So it needs to renumber the
2191 registers in order to output a leaf function. The following macros
2194 @defmac LEAF_REGISTERS
2195 Name of a char vector, indexed by hard register number, which
2196 contains 1 for a register that is allowable in a candidate for leaf
2199 If leaf function treatment involves renumbering the registers, then the
2200 registers marked here should be the ones before renumbering---those that
2201 GCC would ordinarily allocate. The registers which will actually be
2202 used in the assembler code, after renumbering, should not be marked with 1
2205 Define this macro only if the target machine offers a way to optimize
2206 the treatment of leaf functions.
2209 @defmac LEAF_REG_REMAP (@var{regno})
2210 A C expression whose value is the register number to which @var{regno}
2211 should be renumbered, when a function is treated as a leaf function.
2213 If @var{regno} is a register number which should not appear in a leaf
2214 function before renumbering, then the expression should yield @minus{}1, which
2215 will cause the compiler to abort.
2217 Define this macro only if the target machine offers a way to optimize the
2218 treatment of leaf functions, and registers need to be renumbered to do
2222 @findex current_function_is_leaf
2223 @findex current_function_uses_only_leaf_regs
2224 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2225 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2226 specially. They can test the C variable @code{current_function_is_leaf}
2227 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2228 set prior to local register allocation and is valid for the remaining
2229 compiler passes. They can also test the C variable
2230 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2231 functions which only use leaf registers.
2232 @code{current_function_uses_only_leaf_regs} is valid after all passes
2233 that modify the instructions have been run and is only useful if
2234 @code{LEAF_REGISTERS} is defined.
2235 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2236 @c of the next paragraph?! --mew 2feb93
2238 @node Stack Registers
2239 @subsection Registers That Form a Stack
2241 There are special features to handle computers where some of the
2242 ``registers'' form a stack. Stack registers are normally written by
2243 pushing onto the stack, and are numbered relative to the top of the
2246 Currently, GCC can only handle one group of stack-like registers, and
2247 they must be consecutively numbered. Furthermore, the existing
2248 support for stack-like registers is specific to the 80387 floating
2249 point coprocessor. If you have a new architecture that uses
2250 stack-like registers, you will need to do substantial work on
2251 @file{reg-stack.c} and write your machine description to cooperate
2252 with it, as well as defining these macros.
2255 Define this if the machine has any stack-like registers.
2258 @defmac STACK_REG_COVER_CLASS
2259 This is a cover class containing the stack registers. Define this if
2260 the machine has any stack-like registers.
2263 @defmac FIRST_STACK_REG
2264 The number of the first stack-like register. This one is the top
2268 @defmac LAST_STACK_REG
2269 The number of the last stack-like register. This one is the bottom of
2273 @node Register Classes
2274 @section Register Classes
2275 @cindex register class definitions
2276 @cindex class definitions, register
2278 On many machines, the numbered registers are not all equivalent.
2279 For example, certain registers may not be allowed for indexed addressing;
2280 certain registers may not be allowed in some instructions. These machine
2281 restrictions are described to the compiler using @dfn{register classes}.
2283 You define a number of register classes, giving each one a name and saying
2284 which of the registers belong to it. Then you can specify register classes
2285 that are allowed as operands to particular instruction patterns.
2289 In general, each register will belong to several classes. In fact, one
2290 class must be named @code{ALL_REGS} and contain all the registers. Another
2291 class must be named @code{NO_REGS} and contain no registers. Often the
2292 union of two classes will be another class; however, this is not required.
2294 @findex GENERAL_REGS
2295 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2296 terribly special about the name, but the operand constraint letters
2297 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2298 the same as @code{ALL_REGS}, just define it as a macro which expands
2301 Order the classes so that if class @var{x} is contained in class @var{y}
2302 then @var{x} has a lower class number than @var{y}.
2304 The way classes other than @code{GENERAL_REGS} are specified in operand
2305 constraints is through machine-dependent operand constraint letters.
2306 You can define such letters to correspond to various classes, then use
2307 them in operand constraints.
2309 You must define the narrowest register classes for allocatable
2310 registers, so that each class either has no subclasses, or that for
2311 some mode, the move cost between registers within the class is
2312 cheaper than moving a register in the class to or from memory
2315 You should define a class for the union of two classes whenever some
2316 instruction allows both classes. For example, if an instruction allows
2317 either a floating point (coprocessor) register or a general register for a
2318 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2319 which includes both of them. Otherwise you will get suboptimal code,
2320 or even internal compiler errors when reload cannot find a register in the
2321 class computed via @code{reg_class_subunion}.
2323 You must also specify certain redundant information about the register
2324 classes: for each class, which classes contain it and which ones are
2325 contained in it; for each pair of classes, the largest class contained
2328 When a value occupying several consecutive registers is expected in a
2329 certain class, all the registers used must belong to that class.
2330 Therefore, register classes cannot be used to enforce a requirement for
2331 a register pair to start with an even-numbered register. The way to
2332 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2334 Register classes used for input-operands of bitwise-and or shift
2335 instructions have a special requirement: each such class must have, for
2336 each fixed-point machine mode, a subclass whose registers can transfer that
2337 mode to or from memory. For example, on some machines, the operations for
2338 single-byte values (@code{QImode}) are limited to certain registers. When
2339 this is so, each register class that is used in a bitwise-and or shift
2340 instruction must have a subclass consisting of registers from which
2341 single-byte values can be loaded or stored. This is so that
2342 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2344 @deftp {Data type} {enum reg_class}
2345 An enumerated type that must be defined with all the register class names
2346 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2347 must be the last register class, followed by one more enumerated value,
2348 @code{LIM_REG_CLASSES}, which is not a register class but rather
2349 tells how many classes there are.
2351 Each register class has a number, which is the value of casting
2352 the class name to type @code{int}. The number serves as an index
2353 in many of the tables described below.
2356 @defmac N_REG_CLASSES
2357 The number of distinct register classes, defined as follows:
2360 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2364 @defmac REG_CLASS_NAMES
2365 An initializer containing the names of the register classes as C string
2366 constants. These names are used in writing some of the debugging dumps.
2369 @defmac REG_CLASS_CONTENTS
2370 An initializer containing the contents of the register classes, as integers
2371 which are bit masks. The @var{n}th integer specifies the contents of class
2372 @var{n}. The way the integer @var{mask} is interpreted is that
2373 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2375 When the machine has more than 32 registers, an integer does not suffice.
2376 Then the integers are replaced by sub-initializers, braced groupings containing
2377 several integers. Each sub-initializer must be suitable as an initializer
2378 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2379 In this situation, the first integer in each sub-initializer corresponds to
2380 registers 0 through 31, the second integer to registers 32 through 63, and
2384 @defmac REGNO_REG_CLASS (@var{regno})
2385 A C expression whose value is a register class containing hard register
2386 @var{regno}. In general there is more than one such class; choose a class
2387 which is @dfn{minimal}, meaning that no smaller class also contains the
2391 @defmac BASE_REG_CLASS
2392 A macro whose definition is the name of the class to which a valid
2393 base register must belong. A base register is one used in an address
2394 which is the register value plus a displacement.
2397 @defmac MODE_BASE_REG_CLASS (@var{mode})
2398 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2399 the selection of a base register in a mode dependent manner. If
2400 @var{mode} is VOIDmode then it should return the same value as
2401 @code{BASE_REG_CLASS}.
2404 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2405 A C expression whose value is the register class to which a valid
2406 base register must belong in order to be used in a base plus index
2407 register address. You should define this macro if base plus index
2408 addresses have different requirements than other base register uses.
2411 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2412 A C expression whose value is the register class to which a valid
2413 base register for a memory reference in mode @var{mode} to address
2414 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2415 define the context in which the base register occurs. @var{outer_code} is
2416 the code of the immediately enclosing expression (@code{MEM} for the top level
2417 of an address, @code{ADDRESS} for something that occurs in an
2418 @code{address_operand}). @var{index_code} is the code of the corresponding
2419 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2422 @defmac INDEX_REG_CLASS
2423 A macro whose definition is the name of the class to which a valid
2424 index register must belong. An index register is one used in an
2425 address where its value is either multiplied by a scale factor or
2426 added to another register (as well as added to a displacement).
2429 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2430 A C expression which is nonzero if register number @var{num} is
2431 suitable for use as a base register in operand addresses.
2434 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2435 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2436 that expression may examine the mode of the memory reference in
2437 @var{mode}. You should define this macro if the mode of the memory
2438 reference affects whether a register may be used as a base register. If
2439 you define this macro, the compiler will use it instead of
2440 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2441 addresses that appear outside a @code{MEM}, i.e., as an
2442 @code{address_operand}.
2445 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2446 A C expression which is nonzero if register number @var{num} is suitable for
2447 use as a base register in base plus index operand addresses, accessing
2448 memory in mode @var{mode}. It may be either a suitable hard register or a
2449 pseudo register that has been allocated such a hard register. You should
2450 define this macro if base plus index addresses have different requirements
2451 than other base register uses.
2453 Use of this macro is deprecated; please use the more general
2454 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2457 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2458 A C expression which is nonzero if register number @var{num} is
2459 suitable for use as a base register in operand addresses, accessing
2460 memory in mode @var{mode} in address space @var{address_space}.
2461 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2462 that that expression may examine the context in which the register
2463 appears in the memory reference. @var{outer_code} is the code of the
2464 immediately enclosing expression (@code{MEM} if at the top level of the
2465 address, @code{ADDRESS} for something that occurs in an
2466 @code{address_operand}). @var{index_code} is the code of the
2467 corresponding index expression if @var{outer_code} is @code{PLUS};
2468 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2469 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2472 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2473 A C expression which is nonzero if register number @var{num} is
2474 suitable for use as an index register in operand addresses. It may be
2475 either a suitable hard register or a pseudo register that has been
2476 allocated such a hard register.
2478 The difference between an index register and a base register is that
2479 the index register may be scaled. If an address involves the sum of
2480 two registers, neither one of them scaled, then either one may be
2481 labeled the ``base'' and the other the ``index''; but whichever
2482 labeling is used must fit the machine's constraints of which registers
2483 may serve in each capacity. The compiler will try both labelings,
2484 looking for one that is valid, and will reload one or both registers
2485 only if neither labeling works.
2488 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RENAME_CLASS (reg_class_t @var{rclass})
2489 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.
2492 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2493 A target hook that places additional restrictions on the register class
2494 to use when it is necessary to copy value @var{x} into a register in class
2495 @var{rclass}. The value is a register class; perhaps @var{rclass}, or perhaps
2496 another, smaller class.
2498 The default version of this hook always returns value of @code{rclass} argument.
2500 Sometimes returning a more restrictive class makes better code. For
2501 example, on the 68000, when @var{x} is an integer constant that is in range
2502 for a @samp{moveq} instruction, the value of this macro is always
2503 @code{DATA_REGS} as long as @var{rclass} includes the data registers.
2504 Requiring a data register guarantees that a @samp{moveq} will be used.
2506 One case where @code{TARGET_PREFERRED_RELOAD_CLASS} must not return
2507 @var{rclass} is if @var{x} is a legitimate constant which cannot be
2508 loaded into some register class. By returning @code{NO_REGS} you can
2509 force @var{x} into a memory location. For example, rs6000 can load
2510 immediate values into general-purpose registers, but does not have an
2511 instruction for loading an immediate value into a floating-point
2512 register, so @code{TARGET_PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2513 @var{x} is a floating-point constant. If the constant can't be loaded
2514 into any kind of register, code generation will be better if
2515 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2516 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2518 If an insn has pseudos in it after register allocation, reload will go
2519 through the alternatives and call repeatedly @code{TARGET_PREFERRED_RELOAD_CLASS}
2520 to find the best one. Returning @code{NO_REGS}, in this case, makes
2521 reload add a @code{!} in front of the constraint: the x86 back-end uses
2522 this feature to discourage usage of 387 registers when math is done in
2523 the SSE registers (and vice versa).
2526 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2527 A C expression that places additional restrictions on the register class
2528 to use when it is necessary to copy value @var{x} into a register in class
2529 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2530 another, smaller class. On many machines, the following definition is
2534 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2537 Sometimes returning a more restrictive class makes better code. For
2538 example, on the 68000, when @var{x} is an integer constant that is in range
2539 for a @samp{moveq} instruction, the value of this macro is always
2540 @code{DATA_REGS} as long as @var{class} includes the data registers.
2541 Requiring a data register guarantees that a @samp{moveq} will be used.
2543 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2544 @var{class} is if @var{x} is a legitimate constant which cannot be
2545 loaded into some register class. By returning @code{NO_REGS} you can
2546 force @var{x} into a memory location. For example, rs6000 can load
2547 immediate values into general-purpose registers, but does not have an
2548 instruction for loading an immediate value into a floating-point
2549 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2550 @var{x} is a floating-point constant. If the constant can't be loaded
2551 into any kind of register, code generation will be better if
2552 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2553 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2555 If an insn has pseudos in it after register allocation, reload will go
2556 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2557 to find the best one. Returning @code{NO_REGS}, in this case, makes
2558 reload add a @code{!} in front of the constraint: the x86 back-end uses
2559 this feature to discourage usage of 387 registers when math is done in
2560 the SSE registers (and vice versa).
2563 @deftypefn {Target Hook} reg_class_t TARGET_PREFERRED_OUTPUT_RELOAD_CLASS (rtx @var{x}, reg_class_t @var{rclass})
2564 Like @code{TARGET_PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2567 The default version of this hook always returns value of @code{rclass}
2570 You can also use @code{TARGET_PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2571 reload from using some alternatives, like @code{TARGET_PREFERRED_RELOAD_CLASS}.
2574 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2575 A C expression that places additional restrictions on the register class
2576 to use when it is necessary to be able to hold a value of mode
2577 @var{mode} in a reload register for which class @var{class} would
2580 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2581 there are certain modes that simply can't go in certain reload classes.
2583 The value is a register class; perhaps @var{class}, or perhaps another,
2586 Don't define this macro unless the target machine has limitations which
2587 require the macro to do something nontrivial.
2590 @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})
2591 Many machines have some registers that cannot be copied directly to or
2592 from memory or even from other types of registers. An example is the
2593 @samp{MQ} register, which on most machines, can only be copied to or
2594 from general registers, but not memory. Below, we shall be using the
2595 term 'intermediate register' when a move operation cannot be performed
2596 directly, but has to be done by copying the source into the intermediate
2597 register first, and then copying the intermediate register to the
2598 destination. An intermediate register always has the same mode as
2599 source and destination. Since it holds the actual value being copied,
2600 reload might apply optimizations to re-use an intermediate register
2601 and eliding the copy from the source when it can determine that the
2602 intermediate register still holds the required value.
2604 Another kind of secondary reload is required on some machines which
2605 allow copying all registers to and from memory, but require a scratch
2606 register for stores to some memory locations (e.g., those with symbolic
2607 address on the RT, and those with certain symbolic address on the SPARC
2608 when compiling PIC)@. Scratch registers need not have the same mode
2609 as the value being copied, and usually hold a different value than
2610 that being copied. Special patterns in the md file are needed to
2611 describe how the copy is performed with the help of the scratch register;
2612 these patterns also describe the number, register class(es) and mode(s)
2613 of the scratch register(s).
2615 In some cases, both an intermediate and a scratch register are required.
2617 For input reloads, this target hook is called with nonzero @var{in_p},
2618 and @var{x} is an rtx that needs to be copied to a register of class
2619 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2620 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2621 needs to be copied to rtx @var{x} in @var{reload_mode}.
2623 If copying a register of @var{reload_class} from/to @var{x} requires
2624 an intermediate register, the hook @code{secondary_reload} should
2625 return the register class required for this intermediate register.
2626 If no intermediate register is required, it should return NO_REGS.
2627 If more than one intermediate register is required, describe the one
2628 that is closest in the copy chain to the reload register.
2630 If scratch registers are needed, you also have to describe how to
2631 perform the copy from/to the reload register to/from this
2632 closest intermediate register. Or if no intermediate register is
2633 required, but still a scratch register is needed, describe the
2634 copy from/to the reload register to/from the reload operand @var{x}.
2636 You do this by setting @code{sri->icode} to the instruction code of a pattern
2637 in the md file which performs the move. Operands 0 and 1 are the output
2638 and input of this copy, respectively. Operands from operand 2 onward are
2639 for scratch operands. These scratch operands must have a mode, and a
2640 single-register-class
2641 @c [later: or memory]
2644 When an intermediate register is used, the @code{secondary_reload}
2645 hook will be called again to determine how to copy the intermediate
2646 register to/from the reload operand @var{x}, so your hook must also
2647 have code to handle the register class of the intermediate operand.
2649 @c [For later: maybe we'll allow multi-alternative reload patterns -
2650 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2651 @c and match the constraints of input and output to determine the required
2652 @c alternative. A restriction would be that constraints used to match
2653 @c against reloads registers would have to be written as register class
2654 @c constraints, or we need a new target macro / hook that tells us if an
2655 @c arbitrary constraint can match an unknown register of a given class.
2656 @c Such a macro / hook would also be useful in other places.]
2659 @var{x} might be a pseudo-register or a @code{subreg} of a
2660 pseudo-register, which could either be in a hard register or in memory.
2661 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2662 in memory and the hard register number if it is in a register.
2664 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2665 currently not supported. For the time being, you will have to continue
2666 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2668 @code{copy_cost} also uses this target hook to find out how values are
2669 copied. If you want it to include some extra cost for the need to allocate
2670 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2671 Or if two dependent moves are supposed to have a lower cost than the sum
2672 of the individual moves due to expected fortuitous scheduling and/or special
2673 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2676 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2677 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2678 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2679 These macros are obsolete, new ports should use the target hook
2680 @code{TARGET_SECONDARY_RELOAD} instead.
2682 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2683 target hook. Older ports still define these macros to indicate to the
2684 reload phase that it may
2685 need to allocate at least one register for a reload in addition to the
2686 register to contain the data. Specifically, if copying @var{x} to a
2687 register @var{class} in @var{mode} requires an intermediate register,
2688 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2689 largest register class all of whose registers can be used as
2690 intermediate registers or scratch registers.
2692 If copying a register @var{class} in @var{mode} to @var{x} requires an
2693 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2694 was supposed to be defined be defined to return the largest register
2695 class required. If the
2696 requirements for input and output reloads were the same, the macro
2697 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2700 The values returned by these macros are often @code{GENERAL_REGS}.
2701 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2702 can be directly copied to or from a register of @var{class} in
2703 @var{mode} without requiring a scratch register. Do not define this
2704 macro if it would always return @code{NO_REGS}.
2706 If a scratch register is required (either with or without an
2707 intermediate register), you were supposed to define patterns for
2708 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2709 (@pxref{Standard Names}. These patterns, which were normally
2710 implemented with a @code{define_expand}, should be similar to the
2711 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2714 These patterns need constraints for the reload register and scratch
2716 contain a single register class. If the original reload register (whose
2717 class is @var{class}) can meet the constraint given in the pattern, the
2718 value returned by these macros is used for the class of the scratch
2719 register. Otherwise, two additional reload registers are required.
2720 Their classes are obtained from the constraints in the insn pattern.
2722 @var{x} might be a pseudo-register or a @code{subreg} of a
2723 pseudo-register, which could either be in a hard register or in memory.
2724 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2725 in memory and the hard register number if it is in a register.
2727 These macros should not be used in the case where a particular class of
2728 registers can only be copied to memory and not to another class of
2729 registers. In that case, secondary reload registers are not needed and
2730 would not be helpful. Instead, a stack location must be used to perform
2731 the copy and the @code{mov@var{m}} pattern should use memory as an
2732 intermediate storage. This case often occurs between floating-point and
2736 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2737 Certain machines have the property that some registers cannot be copied
2738 to some other registers without using memory. Define this macro on
2739 those machines to be a C expression that is nonzero if objects of mode
2740 @var{m} in registers of @var{class1} can only be copied to registers of
2741 class @var{class2} by storing a register of @var{class1} into memory
2742 and loading that memory location into a register of @var{class2}.
2744 Do not define this macro if its value would always be zero.
2747 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2748 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2749 allocates a stack slot for a memory location needed for register copies.
2750 If this macro is defined, the compiler instead uses the memory location
2751 defined by this macro.
2753 Do not define this macro if you do not define
2754 @code{SECONDARY_MEMORY_NEEDED}.
2757 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2758 When the compiler needs a secondary memory location to copy between two
2759 registers of mode @var{mode}, it normally allocates sufficient memory to
2760 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2761 load operations in a mode that many bits wide and whose class is the
2762 same as that of @var{mode}.
2764 This is right thing to do on most machines because it ensures that all
2765 bits of the register are copied and prevents accesses to the registers
2766 in a narrower mode, which some machines prohibit for floating-point
2769 However, this default behavior is not correct on some machines, such as
2770 the DEC Alpha, that store short integers in floating-point registers
2771 differently than in integer registers. On those machines, the default
2772 widening will not work correctly and you must define this macro to
2773 suppress that widening in some cases. See the file @file{alpha.h} for
2776 Do not define this macro if you do not define
2777 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2778 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2781 @deftypefn {Target Hook} bool TARGET_CLASS_LIKELY_SPILLED_P (reg_class_t @var{rclass})
2782 A target hook which returns @code{true} if pseudos that have been assigned
2783 to registers of class @var{rclass} would likely be spilled because
2784 registers of @var{rclass} are needed for spill registers.
2786 The default version of this target hook returns @code{true} if @var{rclass}
2787 has exactly one register and @code{false} otherwise. On most machines, this
2788 default should be used. For generally register-starved machines, such as
2789 i386, or machines with right register constraints, such as SH, this hook
2790 can be used to avoid excessive spilling.
2792 This hook is also used by some of the global intra-procedural code
2793 transformations to throtle code motion, to avoid increasing register
2797 @deftypefn {Target Hook} {unsigned char} TARGET_CLASS_MAX_NREGS (reg_class_t @var{rclass}, machine_mode @var{mode})
2798 A target hook returns the maximum number of consecutive registers
2799 of class @var{rclass} needed to hold a value of mode @var{mode}.
2801 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2802 the value returned by @code{TARGET_CLASS_MAX_NREGS (@var{rclass},
2803 @var{mode})} target hook should be the maximum value of
2804 @code{HARD_REGNO_NREGS (@var{regno}, @var{mode})} for all @var{regno}
2805 values in the class @var{rclass}.
2807 This target hook helps control the handling of multiple-word values
2810 The default version of this target hook returns the size of @var{mode}
2814 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2815 A C expression for the maximum number of consecutive registers
2816 of class @var{class} needed to hold a value of mode @var{mode}.
2818 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2819 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2820 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2821 @var{mode})} for all @var{regno} values in the class @var{class}.
2823 This macro helps control the handling of multiple-word values
2827 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2828 If defined, a C expression that returns nonzero for a @var{class} for which
2829 a change from mode @var{from} to mode @var{to} is invalid.
2831 For example, loading 32-bit integer or floating-point objects into
2832 floating-point registers on Alpha extends them to 64 bits.
2833 Therefore loading a 64-bit object and then storing it as a 32-bit object
2834 does not store the low-order 32 bits, as would be the case for a normal
2835 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2839 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2840 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2841 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2844 Even if storing from a register in mode @var{to} would be valid,
2845 if both @var{from} and @code{raw_reg_mode} for @var{class} are wider
2846 than @code{word_mode}, then we must prevent @var{to} narrowing the
2847 mode. This happens when the middle-end assumes that it can load
2848 or store pieces of an @var{N}-word pseudo, and that the pseudo will
2849 eventually be allocated to @var{N} @code{word_mode} hard registers.
2850 Failure to prevent this kind of mode change will result in the
2851 entire @code{raw_reg_mode} being modified instead of the partial
2852 value that the middle-end intended.
2856 @deftypefn {Target Hook} reg_class_t TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS (int, @var{reg_class_t}, @var{reg_class_t})
2857 A target hook which can change allocno class for given pseudo from
2858 allocno and best class calculated by IRA.
2860 The default version of this target hook always returns given class.
2863 @deftypefn {Target Hook} bool TARGET_LRA_P (void)
2864 A target hook which returns true if we use LRA instead of reload pass. The default version of this target hook returns always false, but new ports should use LRA.
2867 @deftypefn {Target Hook} int TARGET_REGISTER_PRIORITY (int)
2868 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.
2871 @deftypefn {Target Hook} bool TARGET_REGISTER_USAGE_LEVELING_P (void)
2872 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.
2875 @deftypefn {Target Hook} bool TARGET_DIFFERENT_ADDR_DISPLACEMENT_P (void)
2876 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.
2879 @deftypefn {Target Hook} bool TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P (rtx @var{subst})
2880 A target hook which returns @code{true} if @var{subst} can't
2881 substitute safely pseudos with equivalent memory values during
2882 register allocation.
2883 The default version of this target hook returns @code{false}.
2884 On most machines, this default should be used. For generally
2885 machines with non orthogonal register usage for addressing, such
2886 as SH, this hook can be used to avoid excessive spilling.
2889 @deftypefn {Target Hook} bool TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT (rtx *@var{disp}, rtx *@var{offset}, machine_mode @var{mode})
2890 A target hook which returns @code{true} if *@var{disp} is
2891 legitimezed to valid address displacement with subtracting *@var{offset}
2892 at memory mode @var{mode}.
2893 The default version of this target hook returns @code{false}.
2894 This hook will benefit machines with limited base plus displacement
2898 @deftypefn {Target Hook} reg_class_t TARGET_SPILL_CLASS (reg_class_t, @var{machine_mode})
2899 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.
2902 @deftypefn {Target Hook} machine_mode TARGET_CSTORE_MODE (enum insn_code @var{icode})
2903 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.
2906 @node Stack and Calling
2907 @section Stack Layout and Calling Conventions
2908 @cindex calling conventions
2910 @c prevent bad page break with this line
2911 This describes the stack layout and calling conventions.
2915 * Exception Handling::
2920 * Register Arguments::
2922 * Aggregate Return::
2927 * Stack Smashing Protection::
2928 * Miscellaneous Register Hooks::
2932 @subsection Basic Stack Layout
2933 @cindex stack frame layout
2934 @cindex frame layout
2936 @c prevent bad page break with this line
2937 Here is the basic stack layout.
2939 @defmac STACK_GROWS_DOWNWARD
2940 Define this macro to be true if pushing a word onto the stack moves the stack
2941 pointer to a smaller address, and false otherwise.
2944 @defmac STACK_PUSH_CODE
2945 This macro defines the operation used when something is pushed
2946 on the stack. In RTL, a push operation will be
2947 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2949 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2950 and @code{POST_INC}. Which of these is correct depends on
2951 the stack direction and on whether the stack pointer points
2952 to the last item on the stack or whether it points to the
2953 space for the next item on the stack.
2955 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2956 true, which is almost always right, and @code{PRE_INC} otherwise,
2957 which is often wrong.
2960 @defmac FRAME_GROWS_DOWNWARD
2961 Define this macro to nonzero value if the addresses of local variable slots
2962 are at negative offsets from the frame pointer.
2965 @defmac ARGS_GROW_DOWNWARD
2966 Define this macro if successive arguments to a function occupy decreasing
2967 addresses on the stack.
2970 @defmac STARTING_FRAME_OFFSET
2971 Offset from the frame pointer to the first local variable slot to be allocated.
2973 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2974 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2975 Otherwise, it is found by adding the length of the first slot to the
2976 value @code{STARTING_FRAME_OFFSET}.
2977 @c i'm not sure if the above is still correct.. had to change it to get
2978 @c rid of an overfull. --mew 2feb93
2981 @defmac STACK_ALIGNMENT_NEEDED
2982 Define to zero to disable final alignment of the stack during reload.
2983 The nonzero default for this macro is suitable for most ports.
2985 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2986 is a register save block following the local block that doesn't require
2987 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2988 stack alignment and do it in the backend.
2991 @defmac STACK_POINTER_OFFSET
2992 Offset from the stack pointer register to the first location at which
2993 outgoing arguments are placed. If not specified, the default value of
2994 zero is used. This is the proper value for most machines.
2996 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2997 the first location at which outgoing arguments are placed.
3000 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3001 Offset from the argument pointer register to the first argument's
3002 address. On some machines it may depend on the data type of the
3005 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3006 the first argument's address.
3009 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3010 Offset from the stack pointer register to an item dynamically allocated
3011 on the stack, e.g., by @code{alloca}.
3013 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3014 length of the outgoing arguments. The default is correct for most
3015 machines. See @file{function.c} for details.
3018 @defmac INITIAL_FRAME_ADDRESS_RTX
3019 A C expression whose value is RTL representing the address of the initial
3020 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3021 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3022 default value will be used. Define this macro in order to make frame pointer
3023 elimination work in the presence of @code{__builtin_frame_address (count)} and
3024 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3027 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3028 A C expression whose value is RTL representing the address in a stack
3029 frame where the pointer to the caller's frame is stored. Assume that
3030 @var{frameaddr} is an RTL expression for the address of the stack frame
3033 If you don't define this macro, the default is to return the value
3034 of @var{frameaddr}---that is, the stack frame address is also the
3035 address of the stack word that points to the previous frame.
3038 @defmac SETUP_FRAME_ADDRESSES
3039 A C expression that produces the machine-specific code to
3040 setup the stack so that arbitrary frames can be accessed. For example,
3041 on the SPARC, we must flush all of the register windows to the stack
3042 before we can access arbitrary stack frames. You will seldom need to
3043 define this macro. The default is to do nothing.
3046 @deftypefn {Target Hook} rtx TARGET_BUILTIN_SETJMP_FRAME_VALUE (void)
3047 This target hook should return an rtx that is used to store
3048 the address of the current frame into the built in @code{setjmp} buffer.
3049 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3050 machines. One reason you may need to define this target hook is if
3051 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3054 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3055 A C expression whose value is RTL representing the value of the frame
3056 address for the current frame. @var{frameaddr} is the frame pointer
3057 of the current frame. This is used for __builtin_frame_address.
3058 You need only define this macro if the frame address is not the same
3059 as the frame pointer. Most machines do not need to define it.
3062 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3063 A C expression whose value is RTL representing the value of the return
3064 address for the frame @var{count} steps up from the current frame, after
3065 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3066 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3067 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero.
3069 The value of the expression must always be the correct address when
3070 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3071 determine the return address of other frames.
3074 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3075 Define this macro to nonzero value if the return address of a particular
3076 stack frame is accessed from the frame pointer of the previous stack
3077 frame. The zero default for this macro is suitable for most ports.
3080 @defmac INCOMING_RETURN_ADDR_RTX
3081 A C expression whose value is RTL representing the location of the
3082 incoming return address at the beginning of any function, before the
3083 prologue. This RTL is either a @code{REG}, indicating that the return
3084 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3087 You only need to define this macro if you want to support call frame
3088 debugging information like that provided by DWARF 2.
3090 If this RTL is a @code{REG}, you should also define
3091 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3094 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3095 A C expression whose value is an integer giving a DWARF 2 column
3096 number that may be used as an alternative return column. The column
3097 must not correspond to any gcc hard register (that is, it must not
3098 be in the range of @code{DWARF_FRAME_REGNUM}).
3100 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3101 general register, but an alternative column needs to be used for signal
3102 frames. Some targets have also used different frame return columns
3106 @defmac DWARF_ZERO_REG
3107 A C expression whose value is an integer giving a DWARF 2 register
3108 number that is considered to always have the value zero. This should
3109 only be defined if the target has an architected zero register, and
3110 someone decided it was a good idea to use that register number to
3111 terminate the stack backtrace. New ports should avoid this.
3114 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3115 This target hook allows the backend to emit frame-related insns that
3116 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3117 info engine will invoke it on insns of the form
3119 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3123 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3125 to let the backend emit the call frame instructions. @var{label} is
3126 the CFI label attached to the insn, @var{pattern} is the pattern of
3127 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3130 @defmac INCOMING_FRAME_SP_OFFSET
3131 A C expression whose value is an integer giving the offset, in bytes,
3132 from the value of the stack pointer register to the top of the stack
3133 frame at the beginning of any function, before the prologue. The top of
3134 the frame is defined to be the value of the stack pointer in the
3135 previous frame, just before the call instruction.
3137 You only need to define this macro if you want to support call frame
3138 debugging information like that provided by DWARF 2.
3141 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3142 A C expression whose value is an integer giving the offset, in bytes,
3143 from the argument pointer to the canonical frame address (cfa). The
3144 final value should coincide with that calculated by
3145 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3146 during virtual register instantiation.
3148 The default value for this macro is
3149 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
3150 which is correct for most machines; in general, the arguments are found
3151 immediately before the stack frame. Note that this is not the case on
3152 some targets that save registers into the caller's frame, such as SPARC
3153 and rs6000, and so such targets need to define this macro.
3155 You only need to define this macro if the default is incorrect, and you
3156 want to support call frame debugging information like that provided by
3160 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3161 If defined, a C expression whose value is an integer giving the offset
3162 in bytes from the frame pointer to the canonical frame address (cfa).
3163 The final value should coincide with that calculated by
3164 @code{INCOMING_FRAME_SP_OFFSET}.
3166 Normally the CFA is calculated as an offset from the argument pointer,
3167 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3168 variable due to the ABI, this may not be possible. If this macro is
3169 defined, it implies that the virtual register instantiation should be
3170 based on the frame pointer instead of the argument pointer. Only one
3171 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3175 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3176 If defined, a C expression whose value is an integer giving the offset
3177 in bytes from the canonical frame address (cfa) to the frame base used
3178 in DWARF 2 debug information. The default is zero. A different value
3179 may reduce the size of debug information on some ports.
3182 @node Exception Handling
3183 @subsection Exception Handling Support
3184 @cindex exception handling
3186 @defmac EH_RETURN_DATA_REGNO (@var{N})
3187 A C expression whose value is the @var{N}th register number used for
3188 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3189 @var{N} registers are usable.
3191 The exception handling library routines communicate with the exception
3192 handlers via a set of agreed upon registers. Ideally these registers
3193 should be call-clobbered; it is possible to use call-saved registers,
3194 but may negatively impact code size. The target must support at least
3195 2 data registers, but should define 4 if there are enough free registers.
3197 You must define this macro if you want to support call frame exception
3198 handling like that provided by DWARF 2.
3201 @defmac EH_RETURN_STACKADJ_RTX
3202 A C expression whose value is RTL representing a location in which
3203 to store a stack adjustment to be applied before function return.
3204 This is used to unwind the stack to an exception handler's call frame.
3205 It will be assigned zero on code paths that return normally.
3207 Typically this is a call-clobbered hard register that is otherwise
3208 untouched by the epilogue, but could also be a stack slot.
3210 Do not define this macro if the stack pointer is saved and restored
3211 by the regular prolog and epilog code in the call frame itself; in
3212 this case, the exception handling library routines will update the
3213 stack location to be restored in place. Otherwise, you must define
3214 this macro if you want to support call frame exception handling like
3215 that provided by DWARF 2.
3218 @defmac EH_RETURN_HANDLER_RTX
3219 A C expression whose value is RTL representing a location in which
3220 to store the address of an exception handler to which we should
3221 return. It will not be assigned on code paths that return normally.
3223 Typically this is the location in the call frame at which the normal
3224 return address is stored. For targets that return by popping an
3225 address off the stack, this might be a memory address just below
3226 the @emph{target} call frame rather than inside the current call
3227 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3228 been assigned, so it may be used to calculate the location of the
3231 Some targets have more complex requirements than storing to an
3232 address calculable during initial code generation. In that case
3233 the @code{eh_return} instruction pattern should be used instead.
3235 If you want to support call frame exception handling, you must
3236 define either this macro or the @code{eh_return} instruction pattern.
3239 @defmac RETURN_ADDR_OFFSET
3240 If defined, an integer-valued C expression for which rtl will be generated
3241 to add it to the exception handler address before it is searched in the
3242 exception handling tables, and to subtract it again from the address before
3243 using it to return to the exception handler.
3246 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3247 This macro chooses the encoding of pointers embedded in the exception
3248 handling sections. If at all possible, this should be defined such
3249 that the exception handling section will not require dynamic relocations,
3250 and so may be read-only.
3252 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3253 @var{global} is true if the symbol may be affected by dynamic relocations.
3254 The macro should return a combination of the @code{DW_EH_PE_*} defines
3255 as found in @file{dwarf2.h}.
3257 If this macro is not defined, pointers will not be encoded but
3258 represented directly.
3261 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3262 This macro allows the target to emit whatever special magic is required
3263 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3264 Generic code takes care of pc-relative and indirect encodings; this must
3265 be defined if the target uses text-relative or data-relative encodings.
3267 This is a C statement that branches to @var{done} if the format was
3268 handled. @var{encoding} is the format chosen, @var{size} is the number
3269 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3273 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3274 This macro allows the target to add CPU and operating system specific
3275 code to the call-frame unwinder for use when there is no unwind data
3276 available. The most common reason to implement this macro is to unwind
3277 through signal frames.
3279 This macro is called from @code{uw_frame_state_for} in
3280 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3281 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3282 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3283 for the address of the code being executed and @code{context->cfa} for
3284 the stack pointer value. If the frame can be decoded, the register
3285 save addresses should be updated in @var{fs} and the macro should
3286 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3287 the macro should evaluate to @code{_URC_END_OF_STACK}.
3289 For proper signal handling in Java this macro is accompanied by
3290 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3293 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3294 This macro allows the target to add operating system specific code to the
3295 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3296 usually used for signal or interrupt frames.
3298 This macro is called from @code{uw_update_context} in libgcc's
3299 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3300 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3301 for the abi and context in the @code{.unwabi} directive. If the
3302 @code{.unwabi} directive can be handled, the register save addresses should
3303 be updated in @var{fs}.
3306 @defmac TARGET_USES_WEAK_UNWIND_INFO
3307 A C expression that evaluates to true if the target requires unwind
3308 info to be given comdat linkage. Define it to be @code{1} if comdat
3309 linkage is necessary. The default is @code{0}.
3312 @node Stack Checking
3313 @subsection Specifying How Stack Checking is Done
3315 GCC will check that stack references are within the boundaries of the
3316 stack, if the option @option{-fstack-check} is specified, in one of
3321 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3322 will assume that you have arranged for full stack checking to be done
3323 at appropriate places in the configuration files. GCC will not do
3324 other special processing.
3327 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3328 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3329 that you have arranged for static stack checking (checking of the
3330 static stack frame of functions) to be done at appropriate places
3331 in the configuration files. GCC will only emit code to do dynamic
3332 stack checking (checking on dynamic stack allocations) using the third
3336 If neither of the above are true, GCC will generate code to periodically
3337 ``probe'' the stack pointer using the values of the macros defined below.
3340 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3341 GCC will change its allocation strategy for large objects if the option
3342 @option{-fstack-check} is specified: they will always be allocated
3343 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3345 @defmac STACK_CHECK_BUILTIN
3346 A nonzero value if stack checking is done by the configuration files in a
3347 machine-dependent manner. You should define this macro if stack checking
3348 is required by the ABI of your machine or if you would like to do stack
3349 checking in some more efficient way than the generic approach. The default
3350 value of this macro is zero.
3353 @defmac STACK_CHECK_STATIC_BUILTIN
3354 A nonzero value if static stack checking is done by the configuration files
3355 in a machine-dependent manner. You should define this macro if you would
3356 like to do static stack checking in some more efficient way than the generic
3357 approach. The default value of this macro is zero.
3360 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
3361 An integer specifying the interval at which GCC must generate stack probe
3362 instructions, defined as 2 raised to this integer. You will normally
3363 define this macro so that the interval be no larger than the size of
3364 the ``guard pages'' at the end of a stack area. The default value
3365 of 12 (4096-byte interval) is suitable for most systems.
3368 @defmac STACK_CHECK_MOVING_SP
3369 An integer which is nonzero if GCC should move the stack pointer page by page
3370 when doing probes. This can be necessary on systems where the stack pointer
3371 contains the bottom address of the memory area accessible to the executing
3372 thread at any point in time. In this situation an alternate signal stack
3373 is required in order to be able to recover from a stack overflow. The
3374 default value of this macro is zero.
3377 @defmac STACK_CHECK_PROTECT
3378 The number of bytes of stack needed to recover from a stack overflow, for
3379 languages where such a recovery is supported. The default value of 4KB/8KB
3380 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
3381 8KB/12KB with other exception handling mechanisms should be adequate for most
3382 architectures and operating systems.
3385 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3386 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3387 in the opposite case.
3389 @defmac STACK_CHECK_MAX_FRAME_SIZE
3390 The maximum size of a stack frame, in bytes. GCC will generate probe
3391 instructions in non-leaf functions to ensure at least this many bytes of
3392 stack are available. If a stack frame is larger than this size, stack
3393 checking will not be reliable and GCC will issue a warning. The
3394 default is chosen so that GCC only generates one instruction on most
3395 systems. You should normally not change the default value of this macro.
3398 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3399 GCC uses this value to generate the above warning message. It
3400 represents the amount of fixed frame used by a function, not including
3401 space for any callee-saved registers, temporaries and user variables.
3402 You need only specify an upper bound for this amount and will normally
3403 use the default of four words.
3406 @defmac STACK_CHECK_MAX_VAR_SIZE
3407 The maximum size, in bytes, of an object that GCC will place in the
3408 fixed area of the stack frame when the user specifies
3409 @option{-fstack-check}.
3410 GCC computed the default from the values of the above macros and you will
3411 normally not need to override that default.
3415 @node Frame Registers
3416 @subsection Registers That Address the Stack Frame
3418 @c prevent bad page break with this line
3419 This discusses registers that address the stack frame.
3421 @defmac STACK_POINTER_REGNUM
3422 The register number of the stack pointer register, which must also be a
3423 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3424 the hardware determines which register this is.
3427 @defmac FRAME_POINTER_REGNUM
3428 The register number of the frame pointer register, which is used to
3429 access automatic variables in the stack frame. On some machines, the
3430 hardware determines which register this is. On other machines, you can
3431 choose any register you wish for this purpose.
3434 @defmac HARD_FRAME_POINTER_REGNUM
3435 On some machines the offset between the frame pointer and starting
3436 offset of the automatic variables is not known until after register
3437 allocation has been done (for example, because the saved registers are
3438 between these two locations). On those machines, define
3439 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3440 be used internally until the offset is known, and define
3441 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3442 used for the frame pointer.
3444 You should define this macro only in the very rare circumstances when it
3445 is not possible to calculate the offset between the frame pointer and
3446 the automatic variables until after register allocation has been
3447 completed. When this macro is defined, you must also indicate in your
3448 definition of @code{ELIMINABLE_REGS} how to eliminate
3449 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3450 or @code{STACK_POINTER_REGNUM}.
3452 Do not define this macro if it would be the same as
3453 @code{FRAME_POINTER_REGNUM}.
3456 @defmac ARG_POINTER_REGNUM
3457 The register number of the arg pointer register, which is used to access
3458 the function's argument list. On some machines, this is the same as the
3459 frame pointer register. On some machines, the hardware determines which
3460 register this is. On other machines, you can choose any register you
3461 wish for this purpose. If this is not the same register as the frame
3462 pointer register, then you must mark it as a fixed register according to
3463 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3464 (@pxref{Elimination}).
3467 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3468 Define this to a preprocessor constant that is nonzero if
3469 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3470 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3471 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3472 definition is not suitable for use in preprocessor conditionals.
3475 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3476 Define this to a preprocessor constant that is nonzero if
3477 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3478 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3479 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3480 definition is not suitable for use in preprocessor conditionals.
3483 @defmac RETURN_ADDRESS_POINTER_REGNUM
3484 The register number of the return address pointer register, which is used to
3485 access the current function's return address from the stack. On some
3486 machines, the return address is not at a fixed offset from the frame
3487 pointer or stack pointer or argument pointer. This register can be defined
3488 to point to the return address on the stack, and then be converted by
3489 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3491 Do not define this macro unless there is no other way to get the return
3492 address from the stack.
3495 @defmac STATIC_CHAIN_REGNUM
3496 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3497 Register numbers used for passing a function's static chain pointer. If
3498 register windows are used, the register number as seen by the called
3499 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3500 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3501 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3504 The static chain register need not be a fixed register.
3506 If the static chain is passed in memory, these macros should not be
3507 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3510 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl_or_type}, bool @var{incoming_p})
3511 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3512 targets that may use different static chain locations for different
3513 nested functions. This may be required if the target has function
3514 attributes that affect the calling conventions of the function and
3515 those calling conventions use different static chain locations.
3517 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3519 If the static chain is passed in memory, this hook should be used to
3520 provide rtx giving @code{mem} expressions that denote where they are stored.
3521 Often the @code{mem} expression as seen by the caller will be at an offset
3522 from the stack pointer and the @code{mem} expression as seen by the callee
3523 will be at an offset from the frame pointer.
3524 @findex stack_pointer_rtx
3525 @findex frame_pointer_rtx
3526 @findex arg_pointer_rtx
3527 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3528 @code{arg_pointer_rtx} will have been initialized and should be used
3529 to refer to those items.
3532 @defmac DWARF_FRAME_REGISTERS
3533 This macro specifies the maximum number of hard registers that can be
3534 saved in a call frame. This is used to size data structures used in
3535 DWARF2 exception handling.
3537 Prior to GCC 3.0, this macro was needed in order to establish a stable
3538 exception handling ABI in the face of adding new hard registers for ISA
3539 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3540 in the number of hard registers. Nevertheless, this macro can still be
3541 used to reduce the runtime memory requirements of the exception handling
3542 routines, which can be substantial if the ISA contains a lot of
3543 registers that are not call-saved.
3545 If this macro is not defined, it defaults to
3546 @code{FIRST_PSEUDO_REGISTER}.
3549 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3551 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3552 for backward compatibility in pre GCC 3.0 compiled code.
3554 If this macro is not defined, it defaults to
3555 @code{DWARF_FRAME_REGISTERS}.
3558 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3560 Define this macro if the target's representation for dwarf registers
3561 is different than the internal representation for unwind column.
3562 Given a dwarf register, this macro should return the internal unwind
3563 column number to use instead.
3565 See the PowerPC's SPE target for an example.
3568 @defmac DWARF_FRAME_REGNUM (@var{regno})
3570 Define this macro if the target's representation for dwarf registers
3571 used in .eh_frame or .debug_frame is different from that used in other
3572 debug info sections. Given a GCC hard register number, this macro
3573 should return the .eh_frame register number. The default is
3574 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3578 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3580 Define this macro to map register numbers held in the call frame info
3581 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3582 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3583 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3584 return @code{@var{regno}}.
3588 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3590 Define this macro if the target stores register values as
3591 @code{_Unwind_Word} type in unwind context. It should be defined if
3592 target register size is larger than the size of @code{void *}. The
3593 default is to store register values as @code{void *} type.
3597 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3599 Define this macro to be 1 if the target always uses extended unwind
3600 context with version, args_size and by_value fields. If it is undefined,
3601 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3602 defined and 0 otherwise.
3607 @subsection Eliminating Frame Pointer and Arg Pointer
3609 @c prevent bad page break with this line
3610 This is about eliminating the frame pointer and arg pointer.
3612 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3613 This target hook should return @code{true} if a function must have and use
3614 a frame pointer. This target hook is called in the reload pass. If its return
3615 value is @code{true} the function will have a frame pointer.
3617 This target hook can in principle examine the current function and decide
3618 according to the facts, but on most machines the constant @code{false} or the
3619 constant @code{true} suffices. Use @code{false} when the machine allows code
3620 to be generated with no frame pointer, and doing so saves some time or space.
3621 Use @code{true} when there is no possible advantage to avoiding a frame
3624 In certain cases, the compiler does not know how to produce valid code
3625 without a frame pointer. The compiler recognizes those cases and
3626 automatically gives the function a frame pointer regardless of what
3627 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3630 In a function that does not require a frame pointer, the frame pointer
3631 register can be allocated for ordinary usage, unless you mark it as a
3632 fixed register. See @code{FIXED_REGISTERS} for more information.
3634 Default return value is @code{false}.
3637 @findex get_frame_size
3638 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3639 A C statement to store in the variable @var{depth-var} the difference
3640 between the frame pointer and the stack pointer values immediately after
3641 the function prologue. The value would be computed from information
3642 such as the result of @code{get_frame_size ()} and the tables of
3643 registers @code{regs_ever_live} and @code{call_used_regs}.
3645 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3646 need not be defined. Otherwise, it must be defined even if
3647 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3648 case, you may set @var{depth-var} to anything.
3651 @defmac ELIMINABLE_REGS
3652 If defined, this macro specifies a table of register pairs used to
3653 eliminate unneeded registers that point into the stack frame. If it is not
3654 defined, the only elimination attempted by the compiler is to replace
3655 references to the frame pointer with references to the stack pointer.
3657 The definition of this macro is a list of structure initializations, each
3658 of which specifies an original and replacement register.
3660 On some machines, the position of the argument pointer is not known until
3661 the compilation is completed. In such a case, a separate hard register
3662 must be used for the argument pointer. This register can be eliminated by
3663 replacing it with either the frame pointer or the argument pointer,
3664 depending on whether or not the frame pointer has been eliminated.
3666 In this case, you might specify:
3668 #define ELIMINABLE_REGS \
3669 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3670 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3671 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3674 Note that the elimination of the argument pointer with the stack pointer is
3675 specified first since that is the preferred elimination.
3678 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from_reg}, const int @var{to_reg})
3679 This target hook should returns @code{true} if the compiler is allowed to
3680 try to replace register number @var{from_reg} with register number
3681 @var{to_reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3682 is defined, and will usually be @code{true}, since most of the cases
3683 preventing register elimination are things that the compiler already
3686 Default return value is @code{true}.
3689 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3690 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3691 specifies the initial difference between the specified pair of
3692 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3696 @node Stack Arguments
3697 @subsection Passing Function Arguments on the Stack
3698 @cindex arguments on stack
3699 @cindex stack arguments
3701 The macros in this section control how arguments are passed
3702 on the stack. See the following section for other macros that
3703 control passing certain arguments in registers.
3705 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (const_tree @var{fntype})
3706 This target hook returns @code{true} if an argument declared in a
3707 prototype as an integral type smaller than @code{int} should actually be
3708 passed as an @code{int}. In addition to avoiding errors in certain
3709 cases of mismatch, it also makes for better code on certain machines.
3710 The default is to not promote prototypes.
3714 A C expression. If nonzero, push insns will be used to pass
3716 If the target machine does not have a push instruction, set it to zero.
3717 That directs GCC to use an alternate strategy: to
3718 allocate the entire argument block and then store the arguments into
3719 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3722 @defmac PUSH_ARGS_REVERSED
3723 A C expression. If nonzero, function arguments will be evaluated from
3724 last to first, rather than from first to last. If this macro is not
3725 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3726 and args grow in opposite directions, and 0 otherwise.
3729 @defmac PUSH_ROUNDING (@var{npushed})
3730 A C expression that is the number of bytes actually pushed onto the
3731 stack when an instruction attempts to push @var{npushed} bytes.
3733 On some machines, the definition
3736 #define PUSH_ROUNDING(BYTES) (BYTES)
3740 will suffice. But on other machines, instructions that appear
3741 to push one byte actually push two bytes in an attempt to maintain
3742 alignment. Then the definition should be
3745 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3748 If the value of this macro has a type, it should be an unsigned type.
3751 @findex outgoing_args_size
3752 @findex crtl->outgoing_args_size
3753 @defmac ACCUMULATE_OUTGOING_ARGS
3754 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3755 will be computed and placed into
3756 @code{crtl->outgoing_args_size}. No space will be pushed
3757 onto the stack for each call; instead, the function prologue should
3758 increase the stack frame size by this amount.
3760 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3764 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3765 Define this macro if functions should assume that stack space has been
3766 allocated for arguments even when their values are passed in
3769 The value of this macro is the size, in bytes, of the area reserved for
3770 arguments passed in registers for the function represented by @var{fndecl},
3771 which can be zero if GCC is calling a library function.
3772 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3775 This space can be allocated by the caller, or be a part of the
3776 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3779 @c above is overfull. not sure what to do. --mew 5feb93 did
3780 @c something, not sure if it looks good. --mew 10feb93
3782 @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3783 Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3784 Define this macro if space guaranteed when compiling a function body
3785 is different to space required when making a call, a situation that
3786 can arise with K&R style function definitions.
3789 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3790 Define this to a nonzero value if it is the responsibility of the
3791 caller to allocate the area reserved for arguments passed in registers
3792 when calling a function of @var{fntype}. @var{fntype} may be NULL
3793 if the function called is a library function.
3795 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3796 whether the space for these arguments counts in the value of
3797 @code{crtl->outgoing_args_size}.
3800 @defmac STACK_PARMS_IN_REG_PARM_AREA
3801 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3802 stack parameters don't skip the area specified by it.
3803 @c i changed this, makes more sens and it should have taken care of the
3804 @c overfull.. not as specific, tho. --mew 5feb93
3806 Normally, when a parameter is not passed in registers, it is placed on the
3807 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3808 suppresses this behavior and causes the parameter to be passed on the
3809 stack in its natural location.
3812 @deftypefn {Target Hook} int TARGET_RETURN_POPS_ARGS (tree @var{fundecl}, tree @var{funtype}, int @var{size})
3813 This target hook returns the number of bytes of its own arguments that
3814 a function pops on returning, or 0 if the function pops no arguments
3815 and the caller must therefore pop them all after the function returns.
3817 @var{fundecl} is a C variable whose value is a tree node that describes
3818 the function in question. Normally it is a node of type
3819 @code{FUNCTION_DECL} that describes the declaration of the function.
3820 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3822 @var{funtype} is a C variable whose value is a tree node that
3823 describes the function in question. Normally it is a node of type
3824 @code{FUNCTION_TYPE} that describes the data type of the function.
3825 From this it is possible to obtain the data types of the value and
3826 arguments (if known).
3828 When a call to a library function is being considered, @var{fundecl}
3829 will contain an identifier node for the library function. Thus, if
3830 you need to distinguish among various library functions, you can do so
3831 by their names. Note that ``library function'' in this context means
3832 a function used to perform arithmetic, whose name is known specially
3833 in the compiler and was not mentioned in the C code being compiled.
3835 @var{size} is the number of bytes of arguments passed on the
3836 stack. If a variable number of bytes is passed, it is zero, and
3837 argument popping will always be the responsibility of the calling function.
3839 On the VAX, all functions always pop their arguments, so the definition
3840 of this macro is @var{size}. On the 68000, using the standard
3841 calling convention, no functions pop their arguments, so the value of
3842 the macro is always 0 in this case. But an alternative calling
3843 convention is available in which functions that take a fixed number of
3844 arguments pop them but other functions (such as @code{printf}) pop
3845 nothing (the caller pops all). When this convention is in use,
3846 @var{funtype} is examined to determine whether a function takes a fixed
3847 number of arguments.
3850 @defmac CALL_POPS_ARGS (@var{cum})
3851 A C expression that should indicate the number of bytes a call sequence
3852 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3853 when compiling a function call.
3855 @var{cum} is the variable in which all arguments to the called function
3856 have been accumulated.
3858 On certain architectures, such as the SH5, a call trampoline is used
3859 that pops certain registers off the stack, depending on the arguments
3860 that have been passed to the function. Since this is a property of the
3861 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3865 @node Register Arguments
3866 @subsection Passing Arguments in Registers
3867 @cindex arguments in registers
3868 @cindex registers arguments
3870 This section describes the macros which let you control how various
3871 types of arguments are passed in registers or how they are arranged in
3874 @deftypefn {Target Hook} rtx TARGET_FUNCTION_ARG (cumulative_args_t @var{ca}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3875 Return an RTX indicating whether a function argument is passed in a
3876 register and if so, which register.
3878 The arguments are @var{ca}, which summarizes all the previous
3879 arguments; @var{mode}, the machine mode of the argument; @var{type},
3880 the data type of the argument as a tree node or 0 if that is not known
3881 (which happens for C support library functions); and @var{named},
3882 which is @code{true} for an ordinary argument and @code{false} for
3883 nameless arguments that correspond to @samp{@dots{}} in the called
3884 function's prototype. @var{type} can be an incomplete type if a
3885 syntax error has previously occurred.
3887 The return value is usually either a @code{reg} RTX for the hard
3888 register in which to pass the argument, or zero to pass the argument
3891 The return value can be a @code{const_int} which means argument is
3892 passed in a target specific slot with specified number. Target hooks
3893 should be used to store or load argument in such case. See
3894 @code{TARGET_STORE_BOUNDS_FOR_ARG} and @code{TARGET_LOAD_BOUNDS_FOR_ARG}
3895 for more information.
3897 The value of the expression can also be a @code{parallel} RTX@. This is
3898 used when an argument is passed in multiple locations. The mode of the
3899 @code{parallel} should be the mode of the entire argument. The
3900 @code{parallel} holds any number of @code{expr_list} pairs; each one
3901 describes where part of the argument is passed. In each
3902 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
3903 register in which to pass this part of the argument, and the mode of the
3904 register RTX indicates how large this part of the argument is. The
3905 second operand of the @code{expr_list} is a @code{const_int} which gives
3906 the offset in bytes into the entire argument of where this part starts.
3907 As a special exception the first @code{expr_list} in the @code{parallel}
3908 RTX may have a first operand of zero. This indicates that the entire
3909 argument is also stored on the stack.
3911 The last time this hook is called, it is called with @code{MODE ==
3912 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
3913 pattern as operands 2 and 3 respectively.
3915 @cindex @file{stdarg.h} and register arguments
3916 The usual way to make the ISO library @file{stdarg.h} work on a
3917 machine where some arguments are usually passed in registers, is to
3918 cause nameless arguments to be passed on the stack instead. This is
3919 done by making @code{TARGET_FUNCTION_ARG} return 0 whenever
3920 @var{named} is @code{false}.
3922 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{TARGET_FUNCTION_ARG}
3923 @cindex @code{REG_PARM_STACK_SPACE}, and @code{TARGET_FUNCTION_ARG}
3924 You may use the hook @code{targetm.calls.must_pass_in_stack}
3925 in the definition of this macro to determine if this argument is of a
3926 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
3927 is not defined and @code{TARGET_FUNCTION_ARG} returns nonzero for such an
3928 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
3929 defined, the argument will be computed in the stack and then loaded into
3933 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (machine_mode @var{mode}, const_tree @var{type})
3934 This target hook should return @code{true} if we should not pass @var{type}
3935 solely in registers. The file @file{expr.h} defines a
3936 definition that is usually appropriate, refer to @file{expr.h} for additional
3940 @deftypefn {Target Hook} rtx TARGET_FUNCTION_INCOMING_ARG (cumulative_args_t @var{ca}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3941 Define this hook if the caller and callee on the target have different
3942 views of where arguments are passed. Also define this hook if there are
3943 functions that are never directly called, but are invoked by the hardware
3944 and which have nonstandard calling conventions.
3946 In this case @code{TARGET_FUNCTION_ARG} computes the register in
3947 which the caller passes the value, and
3948 @code{TARGET_FUNCTION_INCOMING_ARG} should be defined in a similar
3949 fashion to tell the function being called where the arguments will
3952 @code{TARGET_FUNCTION_INCOMING_ARG} can also return arbitrary address
3953 computation using hard register, which can be forced into a register,
3954 so that it can be used to pass special arguments.
3956 If @code{TARGET_FUNCTION_INCOMING_ARG} is not defined,
3957 @code{TARGET_FUNCTION_ARG} serves both purposes.
3960 @deftypefn {Target Hook} bool TARGET_USE_PSEUDO_PIC_REG (void)
3961 This hook should return 1 in case pseudo register should be created
3962 for pic_offset_table_rtx during function expand.
3965 @deftypefn {Target Hook} void TARGET_INIT_PIC_REG (void)
3966 Perform a target dependent initialization of pic_offset_table_rtx.
3967 This hook is called at the start of register allocation.
3970 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (cumulative_args_t @var{cum}, machine_mode @var{mode}, tree @var{type}, bool @var{named})
3971 This target hook returns the number of bytes at the beginning of an
3972 argument that must be put in registers. The value must be zero for
3973 arguments that are passed entirely in registers or that are entirely
3974 pushed on the stack.
3976 On some machines, certain arguments must be passed partially in
3977 registers and partially in memory. On these machines, typically the
3978 first few words of arguments are passed in registers, and the rest
3979 on the stack. If a multi-word argument (a @code{double} or a
3980 structure) crosses that boundary, its first few words must be passed
3981 in registers and the rest must be pushed. This macro tells the
3982 compiler when this occurs, and how many bytes should go in registers.
3984 @code{TARGET_FUNCTION_ARG} for these arguments should return the first
3985 register to be used by the caller for this argument; likewise
3986 @code{TARGET_FUNCTION_INCOMING_ARG}, for the called function.
3989 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (cumulative_args_t @var{cum}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
3990 This target hook should return @code{true} if an argument at the
3991 position indicated by @var{cum} should be passed by reference. This
3992 predicate is queried after target independent reasons for being
3993 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
3995 If the hook returns true, a copy of that argument is made in memory and a
3996 pointer to the argument is passed instead of the argument itself.
3997 The pointer is passed in whatever way is appropriate for passing a pointer
4001 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (cumulative_args_t @var{cum}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4002 The function argument described by the parameters to this hook is
4003 known to be passed by reference. The hook should return true if the
4004 function argument should be copied by the callee instead of copied
4007 For any argument for which the hook returns true, if it can be
4008 determined that the argument is not modified, then a copy need
4011 The default version of this hook always returns false.
4014 @defmac CUMULATIVE_ARGS
4015 A C type for declaring a variable that is used as the first argument
4016 of @code{TARGET_FUNCTION_ARG} and other related values. For some
4017 target machines, the type @code{int} suffices and can hold the number
4018 of bytes of argument so far.
4020 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4021 arguments that have been passed on the stack. The compiler has other
4022 variables to keep track of that. For target machines on which all
4023 arguments are passed on the stack, there is no need to store anything in
4024 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4025 should not be empty, so use @code{int}.
4028 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4029 If defined, this macro is called before generating any code for a
4030 function, but after the @var{cfun} descriptor for the function has been
4031 created. The back end may use this macro to update @var{cfun} to
4032 reflect an ABI other than that which would normally be used by default.
4033 If the compiler is generating code for a compiler-generated function,
4034 @var{fndecl} may be @code{NULL}.
4037 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4038 A C statement (sans semicolon) for initializing the variable
4039 @var{cum} for the state at the beginning of the argument list. The
4040 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4041 is the tree node for the data type of the function which will receive
4042 the args, or 0 if the args are to a compiler support library function.
4043 For direct calls that are not libcalls, @var{fndecl} contain the
4044 declaration node of the function. @var{fndecl} is also set when
4045 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4046 being compiled. @var{n_named_args} is set to the number of named
4047 arguments, including a structure return address if it is passed as a
4048 parameter, when making a call. When processing incoming arguments,
4049 @var{n_named_args} is set to @minus{}1.
4051 When processing a call to a compiler support library function,
4052 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4053 contains the name of the function, as a string. @var{libname} is 0 when
4054 an ordinary C function call is being processed. Thus, each time this
4055 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4056 never both of them at once.
4059 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4060 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4061 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4062 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4063 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4064 0)} is used instead.
4067 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4068 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4069 finding the arguments for the function being compiled. If this macro is
4070 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4072 The value passed for @var{libname} is always 0, since library routines
4073 with special calling conventions are never compiled with GCC@. The
4074 argument @var{libname} exists for symmetry with
4075 @code{INIT_CUMULATIVE_ARGS}.
4076 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4077 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4080 @deftypefn {Target Hook} void TARGET_FUNCTION_ARG_ADVANCE (cumulative_args_t @var{ca}, machine_mode @var{mode}, const_tree @var{type}, bool @var{named})
4081 This hook updates the summarizer variable pointed to by @var{ca} to
4082 advance past an argument in the argument list. The values @var{mode},
4083 @var{type} and @var{named} describe that argument. Once this is done,
4084 the variable @var{cum} is suitable for analyzing the @emph{following}
4085 argument with @code{TARGET_FUNCTION_ARG}, etc.
4087 This hook need not do anything if the argument in question was passed
4088 on the stack. The compiler knows how to track the amount of stack space
4089 used for arguments without any special help.
4092 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4093 If defined, a C expression that is the number of bytes to add to the
4094 offset of the argument passed in memory. This is needed for the SPU,
4095 which passes @code{char} and @code{short} arguments in the preferred
4096 slot that is in the middle of the quad word instead of starting at the
4100 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4101 If defined, a C expression which determines whether, and in which direction,
4102 to pad out an argument with extra space. The value should be of type
4103 @code{enum direction}: either @code{upward} to pad above the argument,
4104 @code{downward} to pad below, or @code{none} to inhibit padding.
4106 The @emph{amount} of padding is not controlled by this macro, but by the
4107 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
4108 always just enough to reach the next multiple of that boundary.
4110 This macro has a default definition which is right for most systems.
4111 For little-endian machines, the default is to pad upward. For
4112 big-endian machines, the default is to pad downward for an argument of
4113 constant size shorter than an @code{int}, and upward otherwise.
4116 @defmac PAD_VARARGS_DOWN
4117 If defined, a C expression which determines whether the default
4118 implementation of va_arg will attempt to pad down before reading the
4119 next argument, if that argument is smaller than its aligned space as
4120 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4121 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4124 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4125 Specify padding for the last element of a block move between registers and
4126 memory. @var{first} is nonzero if this is the only element. Defining this
4127 macro allows better control of register function parameters on big-endian
4128 machines, without using @code{PARALLEL} rtl. In particular,
4129 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4130 registers, as there is no longer a "wrong" part of a register; For example,
4131 a three byte aggregate may be passed in the high part of a register if so
4135 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_BOUNDARY (machine_mode @var{mode}, const_tree @var{type})
4136 This hook returns the alignment boundary, in bits, of an argument
4137 with the specified mode and type. The default hook returns
4138 @code{PARM_BOUNDARY} for all arguments.
4141 @deftypefn {Target Hook} {unsigned int} TARGET_FUNCTION_ARG_ROUND_BOUNDARY (machine_mode @var{mode}, const_tree @var{type})
4142 Normally, the size of an argument is rounded up to @code{PARM_BOUNDARY},
4143 which is the default value for this hook. You can define this hook to
4144 return a different value if an argument size must be rounded to a larger
4148 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4149 A C expression that is nonzero if @var{regno} is the number of a hard
4150 register in which function arguments are sometimes passed. This does
4151 @emph{not} include implicit arguments such as the static chain and
4152 the structure-value address. On many machines, no registers can be
4153 used for this purpose since all function arguments are pushed on the
4157 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (const_tree @var{type})
4158 This hook should return true if parameter of type @var{type} are passed
4159 as two scalar parameters. By default, GCC will attempt to pack complex
4160 arguments into the target's word size. Some ABIs require complex arguments
4161 to be split and treated as their individual components. For example, on
4162 AIX64, complex floats should be passed in a pair of floating point
4163 registers, even though a complex float would fit in one 64-bit floating
4166 The default value of this hook is @code{NULL}, which is treated as always
4170 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4171 This hook returns a type node for @code{va_list} for the target.
4172 The default version of the hook returns @code{void*}.
4175 @deftypefn {Target Hook} int TARGET_ENUM_VA_LIST_P (int @var{idx}, const char **@var{pname}, tree *@var{ptree})
4176 This target hook is used in function @code{c_common_nodes_and_builtins}
4177 to iterate through the target specific builtin types for va_list. The
4178 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4179 to a @code{const char *} and @var{ptree} a pointer to a @code{tree} typed
4181 The arguments @var{pname} and @var{ptree} are used to store the result of
4182 this macro and are set to the name of the va_list builtin type and its
4184 If the return value of this macro is zero, then there is no more element.
4185 Otherwise the @var{IDX} should be increased for the next call of this
4186 macro to iterate through all types.
4189 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4190 This hook returns the va_list type of the calling convention specified by
4192 The default version of this hook returns @code{va_list_type_node}.
4195 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4196 This hook returns the va_list type of the calling convention specified by the
4197 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4201 @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})
4202 This hook performs target-specific gimplification of
4203 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4204 arguments to @code{va_arg}; the latter two are as in
4205 @code{gimplify.c:gimplify_expr}.
4208 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (machine_mode @var{mode})
4209 Define this to return nonzero if the port can handle pointers
4210 with machine mode @var{mode}. The default version of this
4211 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4214 @deftypefn {Target Hook} bool TARGET_REF_MAY_ALIAS_ERRNO (struct ao_ref *@var{ref})
4215 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.
4218 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (machine_mode @var{mode})
4219 Define this to return nonzero if the port is prepared to handle
4220 insns involving scalar mode @var{mode}. For a scalar mode to be
4221 considered supported, all the basic arithmetic and comparisons
4224 The default version of this hook returns true for any mode
4225 required to handle the basic C types (as defined by the port).
4226 Included here are the double-word arithmetic supported by the
4227 code in @file{optabs.c}.
4230 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (machine_mode @var{mode})
4231 Define this to return nonzero if the port is prepared to handle
4232 insns involving vector mode @var{mode}. At the very least, it
4233 must have move patterns for this mode.
4236 @deftypefn {Target Hook} bool TARGET_ARRAY_MODE_SUPPORTED_P (machine_mode @var{mode}, unsigned HOST_WIDE_INT @var{nelems})
4237 Return true if GCC should try to use a scalar mode to store an array
4238 of @var{nelems} elements, given that each element has mode @var{mode}.
4239 Returning true here overrides the usual @code{MAX_FIXED_MODE} limit
4240 and allows GCC to use any defined integer mode.
4242 One use of this hook is to support vector load and store operations
4243 that operate on several homogeneous vectors. For example, ARM NEON
4244 has operations like:
4247 int8x8x3_t vld3_s8 (const int8_t *)
4250 where the return type is defined as:
4253 typedef struct int8x8x3_t
4259 If this hook allows @code{val} to have a scalar mode, then
4260 @code{int8x8x3_t} can have the same mode. GCC can then store
4261 @code{int8x8x3_t}s in registers rather than forcing them onto the stack.
4264 @deftypefn {Target Hook} bool TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P (machine_mode @var{mode})
4265 Define this to return nonzero if libgcc provides support for the
4266 floating-point mode @var{mode}, which is known to pass
4267 @code{TARGET_SCALAR_MODE_SUPPORTED_P}. The default version of this
4268 hook returns true for all of @code{SFmode}, @code{DFmode},
4269 @code{XFmode} and @code{TFmode}, if such modes exist.
4272 @deftypefn {Target Hook} bool TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P (machine_mode @var{mode})
4273 Define this to return nonzero for machine modes for which the port has
4274 small register classes. If this target hook returns nonzero for a given
4275 @var{mode}, the compiler will try to minimize the lifetime of registers
4276 in @var{mode}. The hook may be called with @code{VOIDmode} as argument.
4277 In this case, the hook is expected to return nonzero if it returns nonzero
4280 On some machines, it is risky to let hard registers live across arbitrary
4281 insns. Typically, these machines have instructions that require values
4282 to be in specific registers (like an accumulator), and reload will fail
4283 if the required hard register is used for another purpose across such an
4286 Passes before reload do not know which hard registers will be used
4287 in an instruction, but the machine modes of the registers set or used in
4288 the instruction are already known. And for some machines, register
4289 classes are small for, say, integer registers but not for floating point
4290 registers. For example, the AMD x86-64 architecture requires specific
4291 registers for the legacy x86 integer instructions, but there are many
4292 SSE registers for floating point operations. On such targets, a good
4293 strategy may be to return nonzero from this hook for @code{INTEGRAL_MODE_P}
4294 machine modes but zero for the SSE register classes.
4296 The default version of this hook returns false for any mode. It is always
4297 safe to redefine this hook to return with a nonzero value. But if you
4298 unnecessarily define it, you will reduce the amount of optimizations
4299 that can be performed in some cases. If you do not define this hook
4300 to return a nonzero value when it is required, the compiler will run out
4301 of spill registers and print a fatal error message.
4305 @subsection How Scalar Function Values Are Returned
4306 @cindex return values in registers
4307 @cindex values, returned by functions
4308 @cindex scalars, returned as values
4310 This section discusses the macros that control returning scalars as
4311 values---values that can fit in registers.
4313 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
4315 Define this to return an RTX representing the place where a function
4316 returns or receives a value of data type @var{ret_type}, a tree node
4317 representing a data type. @var{fn_decl_or_type} is a tree node
4318 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4319 function being called. If @var{outgoing} is false, the hook should
4320 compute the register in which the caller will see the return value.
4321 Otherwise, the hook should return an RTX representing the place where
4322 a function returns a value.
4324 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4325 (Actually, on most machines, scalar values are returned in the same
4326 place regardless of mode.) The value of the expression is usually a
4327 @code{reg} RTX for the hard register where the return value is stored.
4328 The value can also be a @code{parallel} RTX, if the return value is in
4329 multiple places. See @code{TARGET_FUNCTION_ARG} for an explanation of the
4330 @code{parallel} form. Note that the callee will populate every
4331 location specified in the @code{parallel}, but if the first element of
4332 the @code{parallel} contains the whole return value, callers will use
4333 that element as the canonical location and ignore the others. The m68k
4334 port uses this type of @code{parallel} to return pointers in both
4335 @samp{%a0} (the canonical location) and @samp{%d0}.
4337 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4338 the same promotion rules specified in @code{PROMOTE_MODE} if
4339 @var{valtype} is a scalar type.
4341 If the precise function being called is known, @var{func} is a tree
4342 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4343 pointer. This makes it possible to use a different value-returning
4344 convention for specific functions when all their calls are
4347 Some target machines have ``register windows'' so that the register in
4348 which a function returns its value is not the same as the one in which
4349 the caller sees the value. For such machines, you should return
4350 different RTX depending on @var{outgoing}.
4352 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4353 aggregate data types, because these are returned in another way. See
4354 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4357 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4358 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4359 a new target instead.
4362 @defmac LIBCALL_VALUE (@var{mode})
4363 A C expression to create an RTX representing the place where a library
4364 function returns a value of mode @var{mode}.
4366 Note that ``library function'' in this context means a compiler
4367 support routine, used to perform arithmetic, whose name is known
4368 specially by the compiler and was not mentioned in the C code being
4372 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (machine_mode @var{mode}, const_rtx @var{fun})
4373 Define this hook if the back-end needs to know the name of the libcall
4374 function in order to determine where the result should be returned.
4376 The mode of the result is given by @var{mode} and the name of the called
4377 library function is given by @var{fun}. The hook should return an RTX
4378 representing the place where the library function result will be returned.
4380 If this hook is not defined, then LIBCALL_VALUE will be used.
4383 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4384 A C expression that is nonzero if @var{regno} is the number of a hard
4385 register in which the values of called function may come back.
4387 A register whose use for returning values is limited to serving as the
4388 second of a pair (for a value of type @code{double}, say) need not be
4389 recognized by this macro. So for most machines, this definition
4393 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4396 If the machine has register windows, so that the caller and the called
4397 function use different registers for the return value, this macro
4398 should recognize only the caller's register numbers.
4400 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
4401 for a new target instead.
4404 @deftypefn {Target Hook} bool TARGET_FUNCTION_VALUE_REGNO_P (const unsigned int @var{regno})
4405 A target hook that return @code{true} if @var{regno} is the number of a hard
4406 register in which the values of called function may come back.
4408 A register whose use for returning values is limited to serving as the
4409 second of a pair (for a value of type @code{double}, say) need not be
4410 recognized by this target hook.
4412 If the machine has register windows, so that the caller and the called
4413 function use different registers for the return value, this target hook
4414 should recognize only the caller's register numbers.
4416 If this hook is not defined, then FUNCTION_VALUE_REGNO_P will be used.
4419 @defmac APPLY_RESULT_SIZE
4420 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4421 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4422 saving and restoring an arbitrary return value.
4425 @deftypevr {Target Hook} bool TARGET_OMIT_STRUCT_RETURN_REG
4426 Normally, when a function returns a structure by memory, the address
4427 is passed as an invisible pointer argument, but the compiler also
4428 arranges to return the address from the function like it would a normal
4429 pointer return value. Define this to true if that behavior is
4430 undesirable on your target.
4433 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (const_tree @var{type})
4434 This hook should return true if values of type @var{type} are returned
4435 at the most significant end of a register (in other words, if they are
4436 padded at the least significant end). You can assume that @var{type}
4437 is returned in a register; the caller is required to check this.
4439 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4440 be able to hold the complete return value. For example, if a 1-, 2-
4441 or 3-byte structure is returned at the most significant end of a
4442 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4446 @node Aggregate Return
4447 @subsection How Large Values Are Returned
4448 @cindex aggregates as return values
4449 @cindex large return values
4450 @cindex returning aggregate values
4451 @cindex structure value address
4453 When a function value's mode is @code{BLKmode} (and in some other
4454 cases), the value is not returned according to
4455 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4456 caller passes the address of a block of memory in which the value
4457 should be stored. This address is called the @dfn{structure value
4460 This section describes how to control returning structure values in
4463 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (const_tree @var{type}, const_tree @var{fntype})
4464 This target hook should return a nonzero value to say to return the
4465 function value in memory, just as large structures are always returned.
4466 Here @var{type} will be the data type of the value, and @var{fntype}
4467 will be the type of the function doing the returning, or @code{NULL} for
4470 Note that values of mode @code{BLKmode} must be explicitly handled
4471 by this function. Also, the option @option{-fpcc-struct-return}
4472 takes effect regardless of this macro. On most systems, it is
4473 possible to leave the hook undefined; this causes a default
4474 definition to be used, whose value is the constant 1 for @code{BLKmode}
4475 values, and 0 otherwise.
4477 Do not use this hook to indicate that structures and unions should always
4478 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4482 @defmac DEFAULT_PCC_STRUCT_RETURN
4483 Define this macro to be 1 if all structure and union return values must be
4484 in memory. Since this results in slower code, this should be defined
4485 only if needed for compatibility with other compilers or with an ABI@.
4486 If you define this macro to be 0, then the conventions used for structure
4487 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4490 If not defined, this defaults to the value 1.
4493 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4494 This target hook should return the location of the structure value
4495 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4496 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4497 be @code{NULL}, for libcalls. You do not need to define this target
4498 hook if the address is always passed as an ``invisible'' first
4501 On some architectures the place where the structure value address
4502 is found by the called function is not the same place that the
4503 caller put it. This can be due to register windows, or it could
4504 be because the function prologue moves it to a different place.
4505 @var{incoming} is @code{1} or @code{2} when the location is needed in
4506 the context of the called function, and @code{0} in the context of
4509 If @var{incoming} is nonzero and the address is to be found on the
4510 stack, return a @code{mem} which refers to the frame pointer. If
4511 @var{incoming} is @code{2}, the result is being used to fetch the
4512 structure value address at the beginning of a function. If you need
4513 to emit adjusting code, you should do it at this point.
4516 @defmac PCC_STATIC_STRUCT_RETURN
4517 Define this macro if the usual system convention on the target machine
4518 for returning structures and unions is for the called function to return
4519 the address of a static variable containing the value.
4521 Do not define this if the usual system convention is for the caller to
4522 pass an address to the subroutine.
4524 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4525 nothing when you use @option{-freg-struct-return} mode.
4528 @deftypefn {Target Hook} machine_mode TARGET_GET_RAW_RESULT_MODE (int @var{regno})
4529 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.
4532 @deftypefn {Target Hook} machine_mode TARGET_GET_RAW_ARG_MODE (int @var{regno})
4533 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.
4537 @subsection Caller-Saves Register Allocation
4539 If you enable it, GCC can save registers around function calls. This
4540 makes it possible to use call-clobbered registers to hold variables that
4541 must live across calls.
4543 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4544 A C expression specifying which mode is required for saving @var{nregs}
4545 of a pseudo-register in call-clobbered hard register @var{regno}. If
4546 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4547 returned. For most machines this macro need not be defined since GCC
4548 will select the smallest suitable mode.
4551 @node Function Entry
4552 @subsection Function Entry and Exit
4553 @cindex function entry and exit
4557 This section describes the macros that output function entry
4558 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4560 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4561 If defined, a function that outputs the assembler code for entry to a
4562 function. The prologue is responsible for setting up the stack frame,
4563 initializing the frame pointer register, saving registers that must be
4564 saved, and allocating @var{size} additional bytes of storage for the
4565 local variables. @var{size} is an integer. @var{file} is a stdio
4566 stream to which the assembler code should be output.
4568 The label for the beginning of the function need not be output by this
4569 macro. That has already been done when the macro is run.
4571 @findex regs_ever_live
4572 To determine which registers to save, the macro can refer to the array
4573 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4574 @var{r} is used anywhere within the function. This implies the function
4575 prologue should save register @var{r}, provided it is not one of the
4576 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4577 @code{regs_ever_live}.)
4579 On machines that have ``register windows'', the function entry code does
4580 not save on the stack the registers that are in the windows, even if
4581 they are supposed to be preserved by function calls; instead it takes
4582 appropriate steps to ``push'' the register stack, if any non-call-used
4583 registers are used in the function.
4585 @findex frame_pointer_needed
4586 On machines where functions may or may not have frame-pointers, the
4587 function entry code must vary accordingly; it must set up the frame
4588 pointer if one is wanted, and not otherwise. To determine whether a
4589 frame pointer is in wanted, the macro can refer to the variable
4590 @code{frame_pointer_needed}. The variable's value will be 1 at run
4591 time in a function that needs a frame pointer. @xref{Elimination}.
4593 The function entry code is responsible for allocating any stack space
4594 required for the function. This stack space consists of the regions
4595 listed below. In most cases, these regions are allocated in the
4596 order listed, with the last listed region closest to the top of the
4597 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4598 the highest address if it is not defined). You can use a different order
4599 for a machine if doing so is more convenient or required for
4600 compatibility reasons. Except in cases where required by standard
4601 or by a debugger, there is no reason why the stack layout used by GCC
4602 need agree with that used by other compilers for a machine.
4605 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4606 If defined, a function that outputs assembler code at the end of a
4607 prologue. This should be used when the function prologue is being
4608 emitted as RTL, and you have some extra assembler that needs to be
4609 emitted. @xref{prologue instruction pattern}.
4612 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4613 If defined, a function that outputs assembler code at the start of an
4614 epilogue. This should be used when the function epilogue is being
4615 emitted as RTL, and you have some extra assembler that needs to be
4616 emitted. @xref{epilogue instruction pattern}.
4619 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4620 If defined, a function that outputs the assembler code for exit from a
4621 function. The epilogue is responsible for restoring the saved
4622 registers and stack pointer to their values when the function was
4623 called, and returning control to the caller. This macro takes the
4624 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4625 registers to restore are determined from @code{regs_ever_live} and
4626 @code{CALL_USED_REGISTERS} in the same way.
4628 On some machines, there is a single instruction that does all the work
4629 of returning from the function. On these machines, give that
4630 instruction the name @samp{return} and do not define the macro
4631 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4633 Do not define a pattern named @samp{return} if you want the
4634 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4635 switches to control whether return instructions or epilogues are used,
4636 define a @samp{return} pattern with a validity condition that tests the
4637 target switches appropriately. If the @samp{return} pattern's validity
4638 condition is false, epilogues will be used.
4640 On machines where functions may or may not have frame-pointers, the
4641 function exit code must vary accordingly. Sometimes the code for these
4642 two cases is completely different. To determine whether a frame pointer
4643 is wanted, the macro can refer to the variable
4644 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4645 a function that needs a frame pointer.
4647 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4648 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4649 The C variable @code{current_function_is_leaf} is nonzero for such a
4650 function. @xref{Leaf Functions}.
4652 On some machines, some functions pop their arguments on exit while
4653 others leave that for the caller to do. For example, the 68020 when
4654 given @option{-mrtd} pops arguments in functions that take a fixed
4655 number of arguments.
4658 @findex crtl->args.pops_args
4659 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4660 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4661 needs to know what was decided. The number of bytes of the current
4662 function's arguments that this function should pop is available in
4663 @code{crtl->args.pops_args}. @xref{Scalar Return}.
4668 @findex pretend_args_size
4669 @findex crtl->args.pretend_args_size
4670 A region of @code{crtl->args.pretend_args_size} bytes of
4671 uninitialized space just underneath the first argument arriving on the
4672 stack. (This may not be at the very start of the allocated stack region
4673 if the calling sequence has pushed anything else since pushing the stack
4674 arguments. But usually, on such machines, nothing else has been pushed
4675 yet, because the function prologue itself does all the pushing.) This
4676 region is used on machines where an argument may be passed partly in
4677 registers and partly in memory, and, in some cases to support the
4678 features in @code{<stdarg.h>}.
4681 An area of memory used to save certain registers used by the function.
4682 The size of this area, which may also include space for such things as
4683 the return address and pointers to previous stack frames, is
4684 machine-specific and usually depends on which registers have been used
4685 in the function. Machines with register windows often do not require
4689 A region of at least @var{size} bytes, possibly rounded up to an allocation
4690 boundary, to contain the local variables of the function. On some machines,
4691 this region and the save area may occur in the opposite order, with the
4692 save area closer to the top of the stack.
4695 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4696 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4697 @code{crtl->outgoing_args_size} bytes to be used for outgoing
4698 argument lists of the function. @xref{Stack Arguments}.
4701 @defmac EXIT_IGNORE_STACK
4702 Define this macro as a C expression that is nonzero if the return
4703 instruction or the function epilogue ignores the value of the stack
4704 pointer; in other words, if it is safe to delete an instruction to
4705 adjust the stack pointer before a return from the function. The
4708 Note that this macro's value is relevant only for functions for which
4709 frame pointers are maintained. It is never safe to delete a final
4710 stack adjustment in a function that has no frame pointer, and the
4711 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4714 @defmac EPILOGUE_USES (@var{regno})
4715 Define this macro as a C expression that is nonzero for registers that are
4716 used by the epilogue or the @samp{return} pattern. The stack and frame
4717 pointer registers are already assumed to be used as needed.
4720 @defmac EH_USES (@var{regno})
4721 Define this macro as a C expression that is nonzero for registers that are
4722 used by the exception handling mechanism, and so should be considered live
4723 on entry to an exception edge.
4726 @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})
4727 A function that outputs the assembler code for a thunk
4728 function, used to implement C++ virtual function calls with multiple
4729 inheritance. The thunk acts as a wrapper around a virtual function,
4730 adjusting the implicit object parameter before handing control off to
4733 First, emit code to add the integer @var{delta} to the location that
4734 contains the incoming first argument. Assume that this argument
4735 contains a pointer, and is the one used to pass the @code{this} pointer
4736 in C++. This is the incoming argument @emph{before} the function prologue,
4737 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4738 all other incoming arguments.
4740 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4741 made after adding @code{delta}. In particular, if @var{p} is the
4742 adjusted pointer, the following adjustment should be made:
4745 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4748 After the additions, emit code to jump to @var{function}, which is a
4749 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4750 not touch the return address. Hence returning from @var{FUNCTION} will
4751 return to whoever called the current @samp{thunk}.
4753 The effect must be as if @var{function} had been called directly with
4754 the adjusted first argument. This macro is responsible for emitting all
4755 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4756 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4758 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4759 have already been extracted from it.) It might possibly be useful on
4760 some targets, but probably not.
4762 If you do not define this macro, the target-independent code in the C++
4763 front end will generate a less efficient heavyweight thunk that calls
4764 @var{function} instead of jumping to it. The generic approach does
4765 not support varargs.
4768 @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})
4769 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4770 to output the assembler code for the thunk function specified by the
4771 arguments it is passed, and false otherwise. In the latter case, the
4772 generic approach will be used by the C++ front end, with the limitations
4777 @subsection Generating Code for Profiling
4778 @cindex profiling, code generation
4780 These macros will help you generate code for profiling.
4782 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4783 A C statement or compound statement to output to @var{file} some
4784 assembler code to call the profiling subroutine @code{mcount}.
4787 The details of how @code{mcount} expects to be called are determined by
4788 your operating system environment, not by GCC@. To figure them out,
4789 compile a small program for profiling using the system's installed C
4790 compiler and look at the assembler code that results.
4792 Older implementations of @code{mcount} expect the address of a counter
4793 variable to be loaded into some register. The name of this variable is
4794 @samp{LP} followed by the number @var{labelno}, so you would generate
4795 the name using @samp{LP%d} in a @code{fprintf}.
4798 @defmac PROFILE_HOOK
4799 A C statement or compound statement to output to @var{file} some assembly
4800 code to call the profiling subroutine @code{mcount} even the target does
4801 not support profiling.
4804 @defmac NO_PROFILE_COUNTERS
4805 Define this macro to be an expression with a nonzero value if the
4806 @code{mcount} subroutine on your system does not need a counter variable
4807 allocated for each function. This is true for almost all modern
4808 implementations. If you define this macro, you must not use the
4809 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4812 @defmac PROFILE_BEFORE_PROLOGUE
4813 Define this macro if the code for function profiling should come before
4814 the function prologue. Normally, the profiling code comes after.
4817 @deftypefn {Target Hook} bool TARGET_KEEP_LEAF_WHEN_PROFILED (void)
4818 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.
4822 @subsection Permitting tail calls
4825 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4826 True if it is OK to do sibling call optimization for the specified
4827 call expression @var{exp}. @var{decl} will be the called function,
4828 or @code{NULL} if this is an indirect call.
4830 It is not uncommon for limitations of calling conventions to prevent
4831 tail calls to functions outside the current unit of translation, or
4832 during PIC compilation. The hook is used to enforce these restrictions,
4833 as the @code{sibcall} md pattern can not fail, or fall over to a
4834 ``normal'' call. The criteria for successful sibling call optimization
4835 may vary greatly between different architectures.
4838 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap @var{regs})
4839 Add any hard registers to @var{regs} that are live on entry to the
4840 function. This hook only needs to be defined to provide registers that
4841 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4842 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4843 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4844 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4847 @deftypefn {Target Hook} void TARGET_SET_UP_BY_PROLOGUE (struct hard_reg_set_container *@var{})
4848 This hook should add additional registers that are computed by the prologue to the hard regset for shrink-wrapping optimization purposes.
4851 @deftypefn {Target Hook} bool TARGET_WARN_FUNC_RETURN (tree)
4852 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.
4855 @node Stack Smashing Protection
4856 @subsection Stack smashing protection
4857 @cindex stack smashing protection
4859 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4860 This hook returns a @code{DECL} node for the external variable to use
4861 for the stack protection guard. This variable is initialized by the
4862 runtime to some random value and is used to initialize the guard value
4863 that is placed at the top of the local stack frame. The type of this
4864 variable must be @code{ptr_type_node}.
4866 The default version of this hook creates a variable called
4867 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4870 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4871 This hook returns a @code{CALL_EXPR} that alerts the runtime that the
4872 stack protect guard variable has been modified. This expression should
4873 involve a call to a @code{noreturn} function.
4875 The default version of this hook invokes a function called
4876 @samp{__stack_chk_fail}, taking no arguments. This function is
4877 normally defined in @file{libgcc2.c}.
4880 @deftypefn {Common Target Hook} bool TARGET_SUPPORTS_SPLIT_STACK (bool @var{report}, struct gcc_options *@var{opts})
4881 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
4884 @node Miscellaneous Register Hooks
4885 @subsection Miscellaneous register hooks
4886 @cindex miscellaneous register hooks
4888 @deftypevr {Target Hook} bool TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
4889 Set to true if each call that binds to a local definition explicitly
4890 clobbers or sets all non-fixed registers modified by performing the call.
4891 That is, by the call pattern itself, or by code that might be inserted by the
4892 linker (e.g. stubs, veneers, branch islands), but not including those
4893 modifiable by the callee. The affected registers may be mentioned explicitly
4894 in the call pattern, or included as clobbers in CALL_INSN_FUNCTION_USAGE.
4895 The default version of this hook is set to false. The purpose of this hook
4896 is to enable the fipa-ra optimization.
4900 @section Implementing the Varargs Macros
4901 @cindex varargs implementation
4903 GCC comes with an implementation of @code{<varargs.h>} and
4904 @code{<stdarg.h>} that work without change on machines that pass arguments
4905 on the stack. Other machines require their own implementations of
4906 varargs, and the two machine independent header files must have
4907 conditionals to include it.
4909 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4910 the calling convention for @code{va_start}. The traditional
4911 implementation takes just one argument, which is the variable in which
4912 to store the argument pointer. The ISO implementation of
4913 @code{va_start} takes an additional second argument. The user is
4914 supposed to write the last named argument of the function here.
4916 However, @code{va_start} should not use this argument. The way to find
4917 the end of the named arguments is with the built-in functions described
4920 @defmac __builtin_saveregs ()
4921 Use this built-in function to save the argument registers in memory so
4922 that the varargs mechanism can access them. Both ISO and traditional
4923 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4924 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4926 On some machines, @code{__builtin_saveregs} is open-coded under the
4927 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4928 other machines, it calls a routine written in assembler language,
4929 found in @file{libgcc2.c}.
4931 Code generated for the call to @code{__builtin_saveregs} appears at the
4932 beginning of the function, as opposed to where the call to
4933 @code{__builtin_saveregs} is written, regardless of what the code is.
4934 This is because the registers must be saved before the function starts
4935 to use them for its own purposes.
4936 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4940 @defmac __builtin_next_arg (@var{lastarg})
4941 This builtin returns the address of the first anonymous stack
4942 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4943 returns the address of the location above the first anonymous stack
4944 argument. Use it in @code{va_start} to initialize the pointer for
4945 fetching arguments from the stack. Also use it in @code{va_start} to
4946 verify that the second parameter @var{lastarg} is the last named argument
4947 of the current function.
4950 @defmac __builtin_classify_type (@var{object})
4951 Since each machine has its own conventions for which data types are
4952 passed in which kind of register, your implementation of @code{va_arg}
4953 has to embody these conventions. The easiest way to categorize the
4954 specified data type is to use @code{__builtin_classify_type} together
4955 with @code{sizeof} and @code{__alignof__}.
4957 @code{__builtin_classify_type} ignores the value of @var{object},
4958 considering only its data type. It returns an integer describing what
4959 kind of type that is---integer, floating, pointer, structure, and so on.
4961 The file @file{typeclass.h} defines an enumeration that you can use to
4962 interpret the values of @code{__builtin_classify_type}.
4965 These machine description macros help implement varargs:
4967 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
4968 If defined, this hook produces the machine-specific code for a call to
4969 @code{__builtin_saveregs}. This code will be moved to the very
4970 beginning of the function, before any parameter access are made. The
4971 return value of this function should be an RTX that contains the value
4972 to use as the return of @code{__builtin_saveregs}.
4975 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (cumulative_args_t @var{args_so_far}, machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
4976 This target hook offers an alternative to using
4977 @code{__builtin_saveregs} and defining the hook
4978 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
4979 register arguments into the stack so that all the arguments appear to
4980 have been passed consecutively on the stack. Once this is done, you can
4981 use the standard implementation of varargs that works for machines that
4982 pass all their arguments on the stack.
4984 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
4985 structure, containing the values that are obtained after processing the
4986 named arguments. The arguments @var{mode} and @var{type} describe the
4987 last named argument---its machine mode and its data type as a tree node.
4989 The target hook should do two things: first, push onto the stack all the
4990 argument registers @emph{not} used for the named arguments, and second,
4991 store the size of the data thus pushed into the @code{int}-valued
4992 variable pointed to by @var{pretend_args_size}. The value that you
4993 store here will serve as additional offset for setting up the stack
4996 Because you must generate code to push the anonymous arguments at
4997 compile time without knowing their data types,
4998 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
4999 have just a single category of argument register and use it uniformly
5002 If the argument @var{second_time} is nonzero, it means that the
5003 arguments of the function are being analyzed for the second time. This
5004 happens for an inline function, which is not actually compiled until the
5005 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5006 not generate any instructions in this case.
5009 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (cumulative_args_t @var{ca})
5010 Define this hook to return @code{true} if the location where a function
5011 argument is passed depends on whether or not it is a named argument.
5013 This hook controls how the @var{named} argument to @code{TARGET_FUNCTION_ARG}
5014 is set for varargs and stdarg functions. If this hook returns
5015 @code{true}, the @var{named} argument is always true for named
5016 arguments, and false for unnamed arguments. If it returns @code{false},
5017 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5018 then all arguments are treated as named. Otherwise, all named arguments
5019 except the last are treated as named.
5021 You need not define this hook if it always returns @code{false}.
5024 @deftypefn {Target Hook} void TARGET_CALL_ARGS (rtx, @var{tree})
5025 While generating RTL for a function call, this target hook is invoked once
5026 for each argument passed to the function, either a register returned by
5027 @code{TARGET_FUNCTION_ARG} or a memory location. It is called just
5028 before the point where argument registers are stored. The type of the
5029 function to be called is also passed as the second argument; it is
5030 @code{NULL_TREE} for libcalls. The @code{TARGET_END_CALL_ARGS} hook is
5031 invoked just after the code to copy the return reg has been emitted.
5032 This functionality can be used to perform special setup of call argument
5033 registers if a target needs it.
5034 For functions without arguments, the hook is called once with @code{pc_rtx}
5035 passed instead of an argument register.
5036 Most ports do not need to implement anything for this hook.
5039 @deftypefn {Target Hook} void TARGET_END_CALL_ARGS (void)
5040 This target hook is invoked while generating RTL for a function call,
5041 just after the point where the return reg is copied into a pseudo. It
5042 signals that all the call argument and return registers for the just
5043 emitted call are now no longer in use.
5044 Most ports do not need to implement anything for this hook.
5047 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED (cumulative_args_t @var{ca})
5048 If you need to conditionally change ABIs so that one works with
5049 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5050 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5051 defined, then define this hook to return @code{true} if
5052 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5053 Otherwise, you should not define this hook.
5056 @deftypefn {Target Hook} rtx TARGET_LOAD_BOUNDS_FOR_ARG (rtx @var{slot}, rtx @var{arg}, rtx @var{slot_no})
5057 This hook is used by expand pass to emit insn to load bounds of
5058 @var{arg} passed in @var{slot}. Expand pass uses this hook in case
5059 bounds of @var{arg} are not passed in register. If @var{slot} is a
5060 memory, then bounds are loaded as for regular pointer loaded from
5061 memory. If @var{slot} is not a memory then @var{slot_no} is an integer
5062 constant holding number of the target dependent special slot which
5063 should be used to obtain bounds. Hook returns RTX holding loaded bounds.
5066 @deftypefn {Target Hook} void TARGET_STORE_BOUNDS_FOR_ARG (rtx @var{arg}, rtx @var{slot}, rtx @var{bounds}, rtx @var{slot_no})
5067 This hook is used by expand pass to emit insns to store @var{bounds} of
5068 @var{arg} passed in @var{slot}. Expand pass uses this hook in case
5069 @var{bounds} of @var{arg} are not passed in register. If @var{slot} is a
5070 memory, then @var{bounds} are stored as for regular pointer stored in
5071 memory. If @var{slot} is not a memory then @var{slot_no} is an integer
5072 constant holding number of the target dependent special slot which
5073 should be used to store @var{bounds}.
5076 @deftypefn {Target Hook} rtx TARGET_LOAD_RETURNED_BOUNDS (rtx @var{slot})
5077 This hook is used by expand pass to emit insn to load bounds
5078 returned by function call in @var{slot}. Hook returns RTX holding
5082 @deftypefn {Target Hook} void TARGET_STORE_RETURNED_BOUNDS (rtx @var{slot}, rtx @var{bounds})
5083 This hook is used by expand pass to emit insn to store @var{bounds}
5084 returned by function call into @var{slot}.
5087 @deftypefn {Target Hook} rtx TARGET_CHKP_FUNCTION_VALUE_BOUNDS (const_tree @var{ret_type}, const_tree @var{fn_decl_or_type}, bool @var{outgoing})
5088 Define this to return an RTX representing the place where a function
5089 returns bounds for returned pointers. Arguments meaning is similar to
5090 @code{TARGET_FUNCTION_VALUE}.
5093 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARG_BOUNDS (cumulative_args_t @var{args_so_far}, enum machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
5094 Use it to store bounds for anonymous register arguments stored
5095 into the stack. Arguments meaning is similar to
5096 @code{TARGET_SETUP_INCOMING_VARARGS}.
5100 @section Trampolines for Nested Functions
5101 @cindex trampolines for nested functions
5102 @cindex nested functions, trampolines for
5104 A @dfn{trampoline} is a small piece of code that is created at run time
5105 when the address of a nested function is taken. It normally resides on
5106 the stack, in the stack frame of the containing function. These macros
5107 tell GCC how to generate code to allocate and initialize a
5110 The instructions in the trampoline must do two things: load a constant
5111 address into the static chain register, and jump to the real address of
5112 the nested function. On CISC machines such as the m68k, this requires
5113 two instructions, a move immediate and a jump. Then the two addresses
5114 exist in the trampoline as word-long immediate operands. On RISC
5115 machines, it is often necessary to load each address into a register in
5116 two parts. Then pieces of each address form separate immediate
5119 The code generated to initialize the trampoline must store the variable
5120 parts---the static chain value and the function address---into the
5121 immediate operands of the instructions. On a CISC machine, this is
5122 simply a matter of copying each address to a memory reference at the
5123 proper offset from the start of the trampoline. On a RISC machine, it
5124 may be necessary to take out pieces of the address and store them
5127 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5128 This hook is called by @code{assemble_trampoline_template} to output,
5129 on the stream @var{f}, assembler code for a block of data that contains
5130 the constant parts of a trampoline. This code should not include a
5131 label---the label is taken care of automatically.
5133 If you do not define this hook, it means no template is needed
5134 for the target. Do not define this hook on systems where the block move
5135 code to copy the trampoline into place would be larger than the code
5136 to generate it on the spot.
5139 @defmac TRAMPOLINE_SECTION
5140 Return the section into which the trampoline template is to be placed
5141 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5144 @defmac TRAMPOLINE_SIZE
5145 A C expression for the size in bytes of the trampoline, as an integer.
5148 @defmac TRAMPOLINE_ALIGNMENT
5149 Alignment required for trampolines, in bits.
5151 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5152 is used for aligning trampolines.
5155 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5156 This hook is called to initialize a trampoline.
5157 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5158 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5159 RTX for the static chain value that should be passed to the function
5162 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5163 first thing this hook should do is emit a block move into @var{m_tramp}
5164 from the memory block returned by @code{assemble_trampoline_template}.
5165 Note that the block move need only cover the constant parts of the
5166 trampoline. If the target isolates the variable parts of the trampoline
5167 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5169 If the target requires any other actions, such as flushing caches or
5170 enabling stack execution, these actions should be performed after
5171 initializing the trampoline proper.
5174 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5175 This hook should perform any machine-specific adjustment in
5176 the address of the trampoline. Its argument contains the address of the
5177 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5178 the address to be used for a function call should be different from the
5179 address at which the template was stored, the different address should
5180 be returned; otherwise @var{addr} should be returned unchanged.
5181 If this hook is not defined, @var{addr} will be used for function calls.
5184 Implementing trampolines is difficult on many machines because they have
5185 separate instruction and data caches. Writing into a stack location
5186 fails to clear the memory in the instruction cache, so when the program
5187 jumps to that location, it executes the old contents.
5189 Here are two possible solutions. One is to clear the relevant parts of
5190 the instruction cache whenever a trampoline is set up. The other is to
5191 make all trampolines identical, by having them jump to a standard
5192 subroutine. The former technique makes trampoline execution faster; the
5193 latter makes initialization faster.
5195 To clear the instruction cache when a trampoline is initialized, define
5196 the following macro.
5198 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5199 If defined, expands to a C expression clearing the @emph{instruction
5200 cache} in the specified interval. The definition of this macro would
5201 typically be a series of @code{asm} statements. Both @var{beg} and
5202 @var{end} are both pointer expressions.
5205 To use a standard subroutine, define the following macro. In addition,
5206 you must make sure that the instructions in a trampoline fill an entire
5207 cache line with identical instructions, or else ensure that the
5208 beginning of the trampoline code is always aligned at the same point in
5209 its cache line. Look in @file{m68k.h} as a guide.
5211 @defmac TRANSFER_FROM_TRAMPOLINE
5212 Define this macro if trampolines need a special subroutine to do their
5213 work. The macro should expand to a series of @code{asm} statements
5214 which will be compiled with GCC@. They go in a library function named
5215 @code{__transfer_from_trampoline}.
5217 If you need to avoid executing the ordinary prologue code of a compiled
5218 C function when you jump to the subroutine, you can do so by placing a
5219 special label of your own in the assembler code. Use one @code{asm}
5220 statement to generate an assembler label, and another to make the label
5221 global. Then trampolines can use that label to jump directly to your
5222 special assembler code.
5226 @section Implicit Calls to Library Routines
5227 @cindex library subroutine names
5228 @cindex @file{libgcc.a}
5230 @c prevent bad page break with this line
5231 Here is an explanation of implicit calls to library routines.
5233 @defmac DECLARE_LIBRARY_RENAMES
5234 This macro, if defined, should expand to a piece of C code that will get
5235 expanded when compiling functions for libgcc.a. It can be used to
5236 provide alternate names for GCC's internal library functions if there
5237 are ABI-mandated names that the compiler should provide.
5240 @findex set_optab_libfunc
5241 @findex init_one_libfunc
5242 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5243 This hook should declare additional library routines or rename
5244 existing ones, using the functions @code{set_optab_libfunc} and
5245 @code{init_one_libfunc} defined in @file{optabs.c}.
5246 @code{init_optabs} calls this macro after initializing all the normal
5249 The default is to do nothing. Most ports don't need to define this hook.
5252 @deftypevr {Target Hook} bool TARGET_LIBFUNC_GNU_PREFIX
5253 If false (the default), internal library routines start with two
5254 underscores. If set to true, these routines start with @code{__gnu_}
5255 instead. E.g., @code{__muldi3} changes to @code{__gnu_muldi3}. This
5256 currently only affects functions defined in @file{libgcc2.c}. If this
5257 is set to true, the @file{tm.h} file must also
5258 @code{#define LIBGCC2_GNU_PREFIX}.
5261 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5262 This macro should return @code{true} if the library routine that
5263 implements the floating point comparison operator @var{comparison} in
5264 mode @var{mode} will return a boolean, and @var{false} if it will
5267 GCC's own floating point libraries return tristates from the
5268 comparison operators, so the default returns false always. Most ports
5269 don't need to define this macro.
5272 @defmac TARGET_LIB_INT_CMP_BIASED
5273 This macro should evaluate to @code{true} if the integer comparison
5274 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5275 operand is smaller than the second, 1 to indicate that they are equal,
5276 and 2 to indicate that the first operand is greater than the second.
5277 If this macro evaluates to @code{false} the comparison functions return
5278 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5279 in @file{libgcc.a}, you do not need to define this macro.
5282 @defmac TARGET_HAS_NO_HW_DIVIDE
5283 This macro should be defined if the target has no hardware divide
5284 instructions. If this macro is defined, GCC will use an algorithm which
5285 make use of simple logical and arithmetic operations for 64-bit
5286 division. If the macro is not defined, GCC will use an algorithm which
5287 make use of a 64-bit by 32-bit divide primitive.
5290 @cindex @code{EDOM}, implicit usage
5293 The value of @code{EDOM} on the target machine, as a C integer constant
5294 expression. If you don't define this macro, GCC does not attempt to
5295 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5296 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5299 If you do not define @code{TARGET_EDOM}, then compiled code reports
5300 domain errors by calling the library function and letting it report the
5301 error. If mathematical functions on your system use @code{matherr} when
5302 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5303 that @code{matherr} is used normally.
5306 @cindex @code{errno}, implicit usage
5307 @defmac GEN_ERRNO_RTX
5308 Define this macro as a C expression to create an rtl expression that
5309 refers to the global ``variable'' @code{errno}. (On certain systems,
5310 @code{errno} may not actually be a variable.) If you don't define this
5311 macro, a reasonable default is used.
5314 @deftypefn {Target Hook} bool TARGET_LIBC_HAS_FUNCTION (enum function_class @var{fn_class})
5315 This hook determines whether a function from a class of functions
5316 @var{fn_class} is present at the runtime.
5319 @defmac NEXT_OBJC_RUNTIME
5320 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
5321 by default. This calling convention involves passing the object, the selector
5322 and the method arguments all at once to the method-lookup library function.
5323 This is the usual setting when targeting Darwin/Mac OS X systems, which have
5324 the NeXT runtime installed.
5326 If the macro is set to 0, the "GNU" Objective-C message sending convention
5327 will be used by default. This convention passes just the object and the
5328 selector to the method-lookup function, which returns a pointer to the method.
5330 In either case, it remains possible to select code-generation for the alternate
5331 scheme, by means of compiler command line switches.
5334 @node Addressing Modes
5335 @section Addressing Modes
5336 @cindex addressing modes
5338 @c prevent bad page break with this line
5339 This is about addressing modes.
5341 @defmac HAVE_PRE_INCREMENT
5342 @defmacx HAVE_PRE_DECREMENT
5343 @defmacx HAVE_POST_INCREMENT
5344 @defmacx HAVE_POST_DECREMENT
5345 A C expression that is nonzero if the machine supports pre-increment,
5346 pre-decrement, post-increment, or post-decrement addressing respectively.
5349 @defmac HAVE_PRE_MODIFY_DISP
5350 @defmacx HAVE_POST_MODIFY_DISP
5351 A C expression that is nonzero if the machine supports pre- or
5352 post-address side-effect generation involving constants other than
5353 the size of the memory operand.
5356 @defmac HAVE_PRE_MODIFY_REG
5357 @defmacx HAVE_POST_MODIFY_REG
5358 A C expression that is nonzero if the machine supports pre- or
5359 post-address side-effect generation involving a register displacement.
5362 @defmac CONSTANT_ADDRESS_P (@var{x})
5363 A C expression that is 1 if the RTX @var{x} is a constant which
5364 is a valid address. On most machines the default definition of
5365 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
5366 is acceptable, but a few machines are more restrictive as to which
5367 constant addresses are supported.
5370 @defmac CONSTANT_P (@var{x})
5371 @code{CONSTANT_P}, which is defined by target-independent code,
5372 accepts integer-values expressions whose values are not explicitly
5373 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5374 expressions and @code{const} arithmetic expressions, in addition to
5375 @code{const_int} and @code{const_double} expressions.
5378 @defmac MAX_REGS_PER_ADDRESS
5379 A number, the maximum number of registers that can appear in a valid
5380 memory address. Note that it is up to you to specify a value equal to
5381 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5385 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_ADDRESS_P (machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5386 A function that returns whether @var{x} (an RTX) is a legitimate memory
5387 address on the target machine for a memory operand of mode @var{mode}.
5389 Legitimate addresses are defined in two variants: a strict variant and a
5390 non-strict one. The @var{strict} parameter chooses which variant is
5391 desired by the caller.
5393 The strict variant is used in the reload pass. It must be defined so
5394 that any pseudo-register that has not been allocated a hard register is
5395 considered a memory reference. This is because in contexts where some
5396 kind of register is required, a pseudo-register with no hard register
5397 must be rejected. For non-hard registers, the strict variant should look
5398 up the @code{reg_renumber} array; it should then proceed using the hard
5399 register number in the array, or treat the pseudo as a memory reference
5400 if the array holds @code{-1}.
5402 The non-strict variant is used in other passes. It must be defined to
5403 accept all pseudo-registers in every context where some kind of
5404 register is required.
5406 Normally, constant addresses which are the sum of a @code{symbol_ref}
5407 and an integer are stored inside a @code{const} RTX to mark them as
5408 constant. Therefore, there is no need to recognize such sums
5409 specifically as legitimate addresses. Normally you would simply
5410 recognize any @code{const} as legitimate.
5412 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5413 sums that are not marked with @code{const}. It assumes that a naked
5414 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5415 naked constant sums as illegitimate addresses, so that none of them will
5416 be given to @code{PRINT_OPERAND_ADDRESS}.
5418 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5419 On some machines, whether a symbolic address is legitimate depends on
5420 the section that the address refers to. On these machines, define the
5421 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5422 into the @code{symbol_ref}, and then check for it here. When you see a
5423 @code{const}, you will have to look inside it to find the
5424 @code{symbol_ref} in order to determine the section. @xref{Assembler
5427 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5428 Some ports are still using a deprecated legacy substitute for
5429 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5433 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5437 and should @code{goto @var{label}} if the address @var{x} is a valid
5438 address on the target machine for a memory operand of mode @var{mode}.
5440 @findex REG_OK_STRICT
5441 Compiler source files that want to use the strict variant of this
5442 macro define the macro @code{REG_OK_STRICT}. You should use an
5443 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
5444 that case and the non-strict variant otherwise.
5446 Using the hook is usually simpler because it limits the number of
5447 files that are recompiled when changes are made.
5450 @defmac TARGET_MEM_CONSTRAINT
5451 A single character to be used instead of the default @code{'m'}
5452 character for general memory addresses. This defines the constraint
5453 letter which matches the memory addresses accepted by
5454 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5455 support new address formats in your back end without changing the
5456 semantics of the @code{'m'} constraint. This is necessary in order to
5457 preserve functionality of inline assembly constructs using the
5458 @code{'m'} constraint.
5461 @defmac FIND_BASE_TERM (@var{x})
5462 A C expression to determine the base term of address @var{x},
5463 or to provide a simplified version of @var{x} from which @file{alias.c}
5464 can easily find the base term. This macro is used in only two places:
5465 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5467 It is always safe for this macro to not be defined. It exists so
5468 that alias analysis can understand machine-dependent addresses.
5470 The typical use of this macro is to handle addresses containing
5471 a label_ref or symbol_ref within an UNSPEC@.
5474 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode})
5475 This hook is given an invalid memory address @var{x} for an
5476 operand of mode @var{mode} and should try to return a valid memory
5479 @findex break_out_memory_refs
5480 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5481 and @var{oldx} will be the operand that was given to that function to produce
5484 The code of the hook should not alter the substructure of
5485 @var{x}. If it transforms @var{x} into a more legitimate form, it
5486 should return the new @var{x}.
5488 It is not necessary for this hook to come up with a legitimate address,
5489 with the exception of native TLS addresses (@pxref{Emulated TLS}).
5490 The compiler has standard ways of doing so in all cases. In fact, if
5491 the target supports only emulated TLS, it
5492 is safe to omit this hook or make it return @var{x} if it cannot find
5493 a valid way to legitimize the address. But often a machine-dependent
5494 strategy can generate better code.
5497 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5498 A C compound statement that attempts to replace @var{x}, which is an address
5499 that needs reloading, with a valid memory address for an operand of mode
5500 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5501 It is not necessary to define this macro, but it might be useful for
5502 performance reasons.
5504 For example, on the i386, it is sometimes possible to use a single
5505 reload register instead of two by reloading a sum of two pseudo
5506 registers into a register. On the other hand, for number of RISC
5507 processors offsets are limited so that often an intermediate address
5508 needs to be generated in order to address a stack slot. By defining
5509 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5510 generated for adjacent some stack slots can be made identical, and thus
5513 @emph{Note}: This macro should be used with caution. It is necessary
5514 to know something of how reload works in order to effectively use this,
5515 and it is quite easy to produce macros that build in too much knowledge
5516 of reload internals.
5518 @emph{Note}: This macro must be able to reload an address created by a
5519 previous invocation of this macro. If it fails to handle such addresses
5520 then the compiler may generate incorrect code or abort.
5523 The macro definition should use @code{push_reload} to indicate parts that
5524 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5525 suitable to be passed unaltered to @code{push_reload}.
5527 The code generated by this macro must not alter the substructure of
5528 @var{x}. If it transforms @var{x} into a more legitimate form, it
5529 should assign @var{x} (which will always be a C variable) a new value.
5530 This also applies to parts that you change indirectly by calling
5533 @findex strict_memory_address_p
5534 The macro definition may use @code{strict_memory_address_p} to test if
5535 the address has become legitimate.
5538 If you want to change only a part of @var{x}, one standard way of doing
5539 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5540 single level of rtl. Thus, if the part to be changed is not at the
5541 top level, you'll need to replace first the top level.
5542 It is not necessary for this macro to come up with a legitimate
5543 address; but often a machine-dependent strategy can generate better code.
5546 @deftypefn {Target Hook} bool TARGET_MODE_DEPENDENT_ADDRESS_P (const_rtx @var{addr}, addr_space_t @var{addrspace})
5547 This hook returns @code{true} if memory address @var{addr} in address
5548 space @var{addrspace} can have
5549 different meanings depending on the machine mode of the memory
5550 reference it is used for or if the address is valid for some modes
5553 Autoincrement and autodecrement addresses typically have mode-dependent
5554 effects because the amount of the increment or decrement is the size
5555 of the operand being addressed. Some machines have other mode-dependent
5556 addresses. Many RISC machines have no mode-dependent addresses.
5558 You may assume that @var{addr} is a valid address for the machine.
5560 The default version of this hook returns @code{false}.
5563 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_CONSTANT_P (machine_mode @var{mode}, rtx @var{x})
5564 This hook returns true if @var{x} is a legitimate constant for a
5565 @var{mode}-mode immediate operand on the target machine. You can assume that
5566 @var{x} satisfies @code{CONSTANT_P}, so you need not check this.
5568 The default definition returns true.
5571 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5572 This hook is used to undo the possibly obfuscating effects of the
5573 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5574 macros. Some backend implementations of these macros wrap symbol
5575 references inside an @code{UNSPEC} rtx to represent PIC or similar
5576 addressing modes. This target hook allows GCC's optimizers to understand
5577 the semantics of these opaque @code{UNSPEC}s by converting them back
5578 into their original form.
5581 @deftypefn {Target Hook} bool TARGET_CONST_NOT_OK_FOR_DEBUG_P (rtx @var{x})
5582 This hook should return true if @var{x} should not be emitted into
5586 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (machine_mode @var{mode}, rtx @var{x})
5587 This hook should return true if @var{x} is of a form that cannot (or
5588 should not) be spilled to the constant pool. @var{mode} is the mode
5591 The default version of this hook returns false.
5593 The primary reason to define this hook is to prevent reload from
5594 deciding that a non-legitimate constant would be better reloaded
5595 from the constant pool instead of spilling and reloading a register
5596 holding the constant. This restriction is often true of addresses
5597 of TLS symbols for various targets.
5600 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (machine_mode @var{mode}, const_rtx @var{x})
5601 This hook should return true if pool entries for constant @var{x} can
5602 be placed in an @code{object_block} structure. @var{mode} is the mode
5605 The default version returns false for all constants.
5608 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_DECL_P (const_tree @var{decl})
5609 This hook should return true if pool entries for @var{decl} should
5610 be placed in an @code{object_block} structure.
5612 The default version returns true for all decls.
5615 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (tree @var{fndecl})
5616 This hook should return the DECL of a function that implements the
5617 reciprocal of the machine-specific builtin function @var{fndecl}, or
5618 @code{NULL_TREE} if such a function is not available.
5621 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5622 This hook should return the DECL of a function @var{f} that given an
5623 address @var{addr} as an argument returns a mask @var{m} that can be
5624 used to extract from two vectors the relevant data that resides in
5625 @var{addr} in case @var{addr} is not properly aligned.
5627 The autovectorizer, when vectorizing a load operation from an address
5628 @var{addr} that may be unaligned, will generate two vector loads from
5629 the two aligned addresses around @var{addr}. It then generates a
5630 @code{REALIGN_LOAD} operation to extract the relevant data from the
5631 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5632 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5633 the third argument, @var{OFF}, defines how the data will be extracted
5634 from these two vectors: if @var{OFF} is 0, then the returned vector is
5635 @var{v2}; otherwise, the returned vector is composed from the last
5636 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5637 @var{OFF} elements of @var{v2}.
5639 If this hook is defined, the autovectorizer will generate a call
5640 to @var{f} (using the DECL tree that this hook returns) and will
5641 use the return value of @var{f} as the argument @var{OFF} to
5642 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5643 should comply with the semantics expected by @code{REALIGN_LOAD}
5645 If this hook is not defined, then @var{addr} will be used as
5646 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5647 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5650 @deftypefn {Target Hook} int TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST (enum vect_cost_for_stmt @var{type_of_cost}, tree @var{vectype}, int @var{misalign})
5651 Returns cost of different scalar or vector statements for vectorization cost model.
5652 For vector memory operations the cost may depend on type (@var{vectype}) and
5653 misalignment value (@var{misalign}).
5656 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE (const_tree @var{type}, bool @var{is_packed})
5657 Return true if vector alignment is reachable (by peeling N iterations) for the given type.
5660 @deftypefn {Target Hook} bool TARGET_VECTORIZE_VEC_PERM_CONST_OK (machine_mode, const unsigned char *@var{sel})
5661 Return true if a vector created for @code{vec_perm_const} is valid.
5664 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (unsigned @var{code}, tree @var{dest_type}, tree @var{src_type})
5665 This hook should return the DECL of a function that implements conversion of the
5666 input vector of type @var{src_type} to type @var{dest_type}.
5667 The value of @var{code} is one of the enumerators in @code{enum tree_code} and
5668 specifies how the conversion is to be applied
5669 (truncation, rounding, etc.).
5671 If this hook is defined, the autovectorizer will use the
5672 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5673 conversion. Otherwise, it will return @code{NULL_TREE}.
5676 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (unsigned @var{code}, tree @var{vec_type_out}, tree @var{vec_type_in})
5677 This hook should return the decl of a function that implements the
5678 vectorized variant of the function with the @code{combined_fn} code
5679 @var{code} or @code{NULL_TREE} if such a function is not available.
5680 The return type of the vectorized function shall be of vector type
5681 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5684 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION (tree @var{fndecl}, tree @var{vec_type_out}, tree @var{vec_type_in})
5685 This hook should return the decl of a function that implements the
5686 vectorized variant of target built-in function @code{fndecl}. The
5687 return type of the vectorized function shall be of vector type
5688 @var{vec_type_out} and the argument types should be @var{vec_type_in}.
5691 @deftypefn {Target Hook} bool TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT (machine_mode @var{mode}, const_tree @var{type}, int @var{misalignment}, bool @var{is_packed})
5692 This hook should return true if the target supports misaligned vector
5693 store/load of a specific factor denoted in the @var{misalignment}
5694 parameter. The vector store/load should be of machine mode @var{mode} and
5695 the elements in the vectors should be of type @var{type}. @var{is_packed}
5696 parameter is true if the memory access is defined in a packed struct.
5699 @deftypefn {Target Hook} machine_mode TARGET_VECTORIZE_PREFERRED_SIMD_MODE (machine_mode @var{mode})
5700 This hook should return the preferred mode for vectorizing scalar
5701 mode @var{mode}. The default is
5702 equal to @code{word_mode}, because the vectorizer can do some
5703 transformations even in absence of specialized @acronym{SIMD} hardware.
5706 @deftypefn {Target Hook} {unsigned int} TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES (void)
5707 This hook should return a mask of sizes that should be iterated over
5708 after trying to autovectorize using the vector size derived from the
5709 mode returned by @code{TARGET_VECTORIZE_PREFERRED_SIMD_MODE}.
5710 The default is zero which means to not iterate over other vector sizes.
5713 @deftypefn {Target Hook} machine_mode TARGET_VECTORIZE_GET_MASK_MODE (unsigned @var{nunits}, unsigned @var{length})
5714 This hook returns mode to be used for a mask to be used for a vector
5715 of specified @var{length} with @var{nunits} elements. By default an integer
5716 vector mode of a proper size is returned.
5719 @deftypefn {Target Hook} {void *} TARGET_VECTORIZE_INIT_COST (struct loop *@var{loop_info})
5720 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.
5723 @deftypefn {Target Hook} unsigned TARGET_VECTORIZE_ADD_STMT_COST (void *@var{data}, int @var{count}, enum vect_cost_for_stmt @var{kind}, struct _stmt_vec_info *@var{stmt_info}, int @var{misalign}, enum vect_cost_model_location @var{where})
5724 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.
5727 @deftypefn {Target Hook} void TARGET_VECTORIZE_FINISH_COST (void *@var{data}, unsigned *@var{prologue_cost}, unsigned *@var{body_cost}, unsigned *@var{epilogue_cost})
5728 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.
5731 @deftypefn {Target Hook} void TARGET_VECTORIZE_DESTROY_COST_DATA (void *@var{data})
5732 This hook should release @var{data} and any related data structures allocated by TARGET_VECTORIZE_INIT_COST. The default releases the accumulator.
5735 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_GATHER (const_tree @var{mem_vectype}, const_tree @var{index_type}, int @var{scale})
5736 Target builtin that implements vector gather operation. @var{mem_vectype}
5737 is the vector type of the load and @var{index_type} is scalar type of
5738 the index, scaled by @var{scale}.
5739 The default is @code{NULL_TREE} which means to not vectorize gather
5743 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_SCATTER (const_tree @var{vectype}, const_tree @var{index_type}, int @var{scale})
5744 Target builtin that implements vector scatter operation. @var{vectype}
5745 is the vector type of the store and @var{index_type} is scalar type of
5746 the index, scaled by @var{scale}.
5747 The default is @code{NULL_TREE} which means to not vectorize scatter
5751 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN (struct cgraph_node *@var{}, struct cgraph_simd_clone *@var{}, @var{tree}, @var{int})
5752 This hook should set @var{vecsize_mangle}, @var{vecsize_int}, @var{vecsize_float}
5753 fields in @var{simd_clone} structure pointed by @var{clone_info} argument and also
5754 @var{simdlen} field if it was previously 0.
5755 The hook should return 0 if SIMD clones shouldn't be emitted,
5756 or number of @var{vecsize_mangle} variants that should be emitted.
5759 @deftypefn {Target Hook} void TARGET_SIMD_CLONE_ADJUST (struct cgraph_node *@var{})
5760 This hook should add implicit @code{attribute(target("..."))} attribute
5761 to SIMD clone @var{node} if needed.
5764 @deftypefn {Target Hook} int TARGET_SIMD_CLONE_USABLE (struct cgraph_node *@var{})
5765 This hook should return -1 if SIMD clone @var{node} shouldn't be used
5766 in vectorized loops in current function, or non-negative number if it is
5767 usable. In that case, the smaller the number is, the more desirable it is
5771 @deftypefn {Target Hook} bool TARGET_GOACC_VALIDATE_DIMS (tree @var{decl}, int *@var{dims}, int @var{fn_level})
5772 This hook should check the launch dimensions provided for an OpenACC
5773 compute region, or routine. Defaulted values are represented as -1
5774 and non-constant values as 0. The @var{fn_level} is negative for the
5775 function corresponding to the compute region. For a routine is is the
5776 outermost level at which partitioned execution may be spawned. The hook
5777 should verify non-default values. If DECL is NULL, global defaults
5778 are being validated and unspecified defaults should be filled in.
5779 Diagnostics should be issued as appropriate. Return
5780 true, if changes have been made. You must override this hook to
5781 provide dimensions larger than 1.
5784 @deftypefn {Target Hook} int TARGET_GOACC_DIM_LIMIT (int @var{axis})
5785 This hook should return the maximum size of a particular dimension,
5786 or zero if unbounded.
5789 @deftypefn {Target Hook} bool TARGET_GOACC_FORK_JOIN (gcall *@var{call}, const int *@var{dims}, bool @var{is_fork})
5790 This hook can be used to convert IFN_GOACC_FORK and IFN_GOACC_JOIN
5791 function calls to target-specific gimple, or indicate whether they
5792 should be retained. It is executed during the oacc_device_lower pass.
5793 It should return true, if the call should be retained. It should
5794 return false, if it is to be deleted (either because target-specific
5795 gimple has been inserted before it, or there is no need for it).
5796 The default hook returns false, if there are no RTL expanders for them.
5799 @deftypefn {Target Hook} void TARGET_GOACC_REDUCTION (gcall *@var{call})
5800 This hook is used by the oacc_transform pass to expand calls to the
5801 @var{GOACC_REDUCTION} internal function, into a sequence of gimple
5802 instructions. @var{call} is gimple statement containing the call to
5803 the function. This hook removes statement @var{call} after the
5804 expanded sequence has been inserted. This hook is also responsible
5805 for allocating any storage for reductions when necessary.
5808 @node Anchored Addresses
5809 @section Anchored Addresses
5810 @cindex anchored addresses
5811 @cindex @option{-fsection-anchors}
5813 GCC usually addresses every static object as a separate entity.
5814 For example, if we have:
5818 int foo (void) @{ return a + b + c; @}
5821 the code for @code{foo} will usually calculate three separate symbolic
5822 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5823 it would be better to calculate just one symbolic address and access
5824 the three variables relative to it. The equivalent pseudocode would
5830 register int *xr = &x;
5831 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5835 (which isn't valid C). We refer to shared addresses like @code{x} as
5836 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5838 The hooks below describe the target properties that GCC needs to know
5839 in order to make effective use of section anchors. It won't use
5840 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5841 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5843 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5844 The minimum offset that should be applied to a section anchor.
5845 On most targets, it should be the smallest offset that can be
5846 applied to a base register while still giving a legitimate address
5847 for every mode. The default value is 0.
5850 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5851 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5852 offset that should be applied to section anchors. The default
5856 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5857 Write the assembly code to define section anchor @var{x}, which is a
5858 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5859 The hook is called with the assembly output position set to the beginning
5860 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5862 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5863 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5864 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5865 is @code{NULL}, which disables the use of section anchors altogether.
5868 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (const_rtx @var{x})
5869 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5870 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5871 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5873 The default version is correct for most targets, but you might need to
5874 intercept this hook to handle things like target-specific attributes
5875 or target-specific sections.
5878 @node Condition Code
5879 @section Condition Code Status
5880 @cindex condition code status
5882 The macros in this section can be split in two families, according to the
5883 two ways of representing condition codes in GCC.
5885 The first representation is the so called @code{(cc0)} representation
5886 (@pxref{Jump Patterns}), where all instructions can have an implicit
5887 clobber of the condition codes. The second is the condition code
5888 register representation, which provides better schedulability for
5889 architectures that do have a condition code register, but on which
5890 most instructions do not affect it. The latter category includes
5893 The implicit clobbering poses a strong restriction on the placement of
5894 the definition and use of the condition code. In the past the definition
5895 and use were always adjacent. However, recent changes to support trapping
5896 arithmatic may result in the definition and user being in different blocks.
5897 Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally,
5898 the definition may be the source of exception handling edges.
5900 These restrictions can prevent important
5901 optimizations on some machines. For example, on the IBM RS/6000, there
5902 is a delay for taken branches unless the condition code register is set
5903 three instructions earlier than the conditional branch. The instruction
5904 scheduler cannot perform this optimization if it is not permitted to
5905 separate the definition and use of the condition code register.
5907 For this reason, it is possible and suggested to use a register to
5908 represent the condition code for new ports. If there is a specific
5909 condition code register in the machine, use a hard register. If the
5910 condition code or comparison result can be placed in any general register,
5911 or if there are multiple condition registers, use a pseudo register.
5912 Registers used to store the condition code value will usually have a mode
5913 that is in class @code{MODE_CC}.
5915 Alternatively, you can use @code{BImode} if the comparison operator is
5916 specified already in the compare instruction. In this case, you are not
5917 interested in most macros in this section.
5920 * CC0 Condition Codes:: Old style representation of condition codes.
5921 * MODE_CC Condition Codes:: Modern representation of condition codes.
5924 @node CC0 Condition Codes
5925 @subsection Representation of condition codes using @code{(cc0)}
5929 The file @file{conditions.h} defines a variable @code{cc_status} to
5930 describe how the condition code was computed (in case the interpretation of
5931 the condition code depends on the instruction that it was set by). This
5932 variable contains the RTL expressions on which the condition code is
5933 currently based, and several standard flags.
5935 Sometimes additional machine-specific flags must be defined in the machine
5936 description header file. It can also add additional machine-specific
5937 information by defining @code{CC_STATUS_MDEP}.
5939 @defmac CC_STATUS_MDEP
5940 C code for a data type which is used for declaring the @code{mdep}
5941 component of @code{cc_status}. It defaults to @code{int}.
5943 This macro is not used on machines that do not use @code{cc0}.
5946 @defmac CC_STATUS_MDEP_INIT
5947 A C expression to initialize the @code{mdep} field to ``empty''.
5948 The default definition does nothing, since most machines don't use
5949 the field anyway. If you want to use the field, you should probably
5950 define this macro to initialize it.
5952 This macro is not used on machines that do not use @code{cc0}.
5955 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5956 A C compound statement to set the components of @code{cc_status}
5957 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5958 this macro's responsibility to recognize insns that set the condition
5959 code as a byproduct of other activity as well as those that explicitly
5962 This macro is not used on machines that do not use @code{cc0}.
5964 If there are insns that do not set the condition code but do alter
5965 other machine registers, this macro must check to see whether they
5966 invalidate the expressions that the condition code is recorded as
5967 reflecting. For example, on the 68000, insns that store in address
5968 registers do not set the condition code, which means that usually
5969 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5970 insns. But suppose that the previous insn set the condition code
5971 based on location @samp{a4@@(102)} and the current insn stores a new
5972 value in @samp{a4}. Although the condition code is not changed by
5973 this, it will no longer be true that it reflects the contents of
5974 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5975 @code{cc_status} in this case to say that nothing is known about the
5976 condition code value.
5978 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5979 with the results of peephole optimization: insns whose patterns are
5980 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5981 constants which are just the operands. The RTL structure of these
5982 insns is not sufficient to indicate what the insns actually do. What
5983 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5984 @code{CC_STATUS_INIT}.
5986 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5987 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5988 @samp{cc}. This avoids having detailed information about patterns in
5989 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5992 @node MODE_CC Condition Codes
5993 @subsection Representation of condition codes using registers
5997 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5998 On many machines, the condition code may be produced by other instructions
5999 than compares, for example the branch can use directly the condition
6000 code set by a subtract instruction. However, on some machines
6001 when the condition code is set this way some bits (such as the overflow
6002 bit) are not set in the same way as a test instruction, so that a different
6003 branch instruction must be used for some conditional branches. When
6004 this happens, use the machine mode of the condition code register to
6005 record different formats of the condition code register. Modes can
6006 also be used to record which compare instruction (e.g. a signed or an
6007 unsigned comparison) produced the condition codes.
6009 If other modes than @code{CCmode} are required, add them to
6010 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
6011 a mode given an operand of a compare. This is needed because the modes
6012 have to be chosen not only during RTL generation but also, for example,
6013 by instruction combination. The result of @code{SELECT_CC_MODE} should
6014 be consistent with the mode used in the patterns; for example to support
6015 the case of the add on the SPARC discussed above, we have the pattern
6019 [(set (reg:CC_NOOV 0)
6021 (plus:SI (match_operand:SI 0 "register_operand" "%r")
6022 (match_operand:SI 1 "arith_operand" "rI"))
6029 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
6030 for comparisons whose argument is a @code{plus}:
6033 #define SELECT_CC_MODE(OP,X,Y) \
6034 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
6035 ? ((OP == LT || OP == LE || OP == GT || OP == GE) \
6036 ? CCFPEmode : CCFPmode) \
6037 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
6038 || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \
6039 ? CC_NOOVmode : CCmode))
6042 Another reason to use modes is to retain information on which operands
6043 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
6046 You should define this macro if and only if you define extra CC modes
6047 in @file{@var{machine}-modes.def}.
6050 @deftypefn {Target Hook} void TARGET_CANONICALIZE_COMPARISON (int *@var{code}, rtx *@var{op0}, rtx *@var{op1}, bool @var{op0_preserve_value})
6051 On some machines not all possible comparisons are defined, but you can
6052 convert an invalid comparison into a valid one. For example, the Alpha
6053 does not have a @code{GT} comparison, but you can use an @code{LT}
6054 comparison instead and swap the order of the operands.
6056 On such machines, implement this hook to do any required conversions.
6057 @var{code} is the initial comparison code and @var{op0} and @var{op1}
6058 are the left and right operands of the comparison, respectively. If
6059 @var{op0_preserve_value} is @code{true} the implementation is not
6060 allowed to change the value of @var{op0} since the value might be used
6061 in RTXs which aren't comparisons. E.g. the implementation is not
6062 allowed to swap operands in that case.
6064 GCC will not assume that the comparison resulting from this macro is
6065 valid but will see if the resulting insn matches a pattern in the
6068 You need not to implement this hook if it would never change the
6069 comparison code or operands.
6072 @defmac REVERSIBLE_CC_MODE (@var{mode})
6073 A C expression whose value is one if it is always safe to reverse a
6074 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
6075 can ever return @var{mode} for a floating-point inequality comparison,
6076 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
6078 You need not define this macro if it would always returns zero or if the
6079 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
6080 For example, here is the definition used on the SPARC, where floating-point
6081 inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}:
6084 #define REVERSIBLE_CC_MODE(MODE) \
6085 ((MODE) != CCFPEmode && (MODE) != CCFPmode)
6089 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
6090 A C expression whose value is reversed condition code of the @var{code} for
6091 comparison done in CC_MODE @var{mode}. The macro is used only in case
6092 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
6093 machine has some non-standard way how to reverse certain conditionals. For
6094 instance in case all floating point conditions are non-trapping, compiler may
6095 freely convert unordered compares to ordered ones. Then definition may look
6099 #define REVERSE_CONDITION(CODE, MODE) \
6100 ((MODE) != CCFPmode ? reverse_condition (CODE) \
6101 : reverse_condition_maybe_unordered (CODE))
6105 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *@var{p1}, unsigned int *@var{p2})
6106 On targets which do not use @code{(cc0)}, and which use a hard
6107 register rather than a pseudo-register to hold condition codes, the
6108 regular CSE passes are often not able to identify cases in which the
6109 hard register is set to a common value. Use this hook to enable a
6110 small pass which optimizes such cases. This hook should return true
6111 to enable this pass, and it should set the integers to which its
6112 arguments point to the hard register numbers used for condition codes.
6113 When there is only one such register, as is true on most systems, the
6114 integer pointed to by @var{p2} should be set to
6115 @code{INVALID_REGNUM}.
6117 The default version of this hook returns false.
6120 @deftypefn {Target Hook} machine_mode TARGET_CC_MODES_COMPATIBLE (machine_mode @var{m1}, machine_mode @var{m2})
6121 On targets which use multiple condition code modes in class
6122 @code{MODE_CC}, it is sometimes the case that a comparison can be
6123 validly done in more than one mode. On such a system, define this
6124 target hook to take two mode arguments and to return a mode in which
6125 both comparisons may be validly done. If there is no such mode,
6126 return @code{VOIDmode}.
6128 The default version of this hook checks whether the modes are the
6129 same. If they are, it returns that mode. If they are different, it
6130 returns @code{VOIDmode}.
6133 @deftypevr {Target Hook} {unsigned int} TARGET_FLAGS_REGNUM
6134 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.
6138 @section Describing Relative Costs of Operations
6139 @cindex costs of instructions
6140 @cindex relative costs
6141 @cindex speed of instructions
6143 These macros let you describe the relative speed of various operations
6144 on the target machine.
6146 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6147 A C expression for the cost of moving data of mode @var{mode} from a
6148 register in class @var{from} to one in class @var{to}. The classes are
6149 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6150 value of 2 is the default; other values are interpreted relative to
6153 It is not required that the cost always equal 2 when @var{from} is the
6154 same as @var{to}; on some machines it is expensive to move between
6155 registers if they are not general registers.
6157 If reload sees an insn consisting of a single @code{set} between two
6158 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6159 classes returns a value of 2, reload does not check to ensure that the
6160 constraints of the insn are met. Setting a cost of other than 2 will
6161 allow reload to verify that the constraints are met. You should do this
6162 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6164 These macros are obsolete, new ports should use the target hook
6165 @code{TARGET_REGISTER_MOVE_COST} instead.
6168 @deftypefn {Target Hook} int TARGET_REGISTER_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{from}, reg_class_t @var{to})
6169 This target hook should return the cost of moving data of mode @var{mode}
6170 from a register in class @var{from} to one in class @var{to}. The classes
6171 are expressed using the enumeration values such as @code{GENERAL_REGS}.
6172 A value of 2 is the default; other values are interpreted relative to
6175 It is not required that the cost always equal 2 when @var{from} is the
6176 same as @var{to}; on some machines it is expensive to move between
6177 registers if they are not general registers.
6179 If reload sees an insn consisting of a single @code{set} between two
6180 hard registers, and if @code{TARGET_REGISTER_MOVE_COST} applied to their
6181 classes returns a value of 2, reload does not check to ensure that the
6182 constraints of the insn are met. Setting a cost of other than 2 will
6183 allow reload to verify that the constraints are met. You should do this
6184 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6186 The default version of this function returns 2.
6189 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6190 A C expression for the cost of moving data of mode @var{mode} between a
6191 register of class @var{class} and memory; @var{in} is zero if the value
6192 is to be written to memory, nonzero if it is to be read in. This cost
6193 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6194 registers and memory is more expensive than between two registers, you
6195 should define this macro to express the relative cost.
6197 If you do not define this macro, GCC uses a default cost of 4 plus
6198 the cost of copying via a secondary reload register, if one is
6199 needed. If your machine requires a secondary reload register to copy
6200 between memory and a register of @var{class} but the reload mechanism is
6201 more complex than copying via an intermediate, define this macro to
6202 reflect the actual cost of the move.
6204 GCC defines the function @code{memory_move_secondary_cost} if
6205 secondary reloads are needed. It computes the costs due to copying via
6206 a secondary register. If your machine copies from memory using a
6207 secondary register in the conventional way but the default base value of
6208 4 is not correct for your machine, define this macro to add some other
6209 value to the result of that function. The arguments to that function
6210 are the same as to this macro.
6212 These macros are obsolete, new ports should use the target hook
6213 @code{TARGET_MEMORY_MOVE_COST} instead.
6216 @deftypefn {Target Hook} int TARGET_MEMORY_MOVE_COST (machine_mode @var{mode}, reg_class_t @var{rclass}, bool @var{in})
6217 This target hook should return the cost of moving data of mode @var{mode}
6218 between a register of class @var{rclass} and memory; @var{in} is @code{false}
6219 if the value is to be written to memory, @code{true} if it is to be read in.
6220 This cost is relative to those in @code{TARGET_REGISTER_MOVE_COST}.
6221 If moving between registers and memory is more expensive than between two
6222 registers, you should add this target hook to express the relative cost.
6224 If you do not add this target hook, GCC uses a default cost of 4 plus
6225 the cost of copying via a secondary reload register, if one is
6226 needed. If your machine requires a secondary reload register to copy
6227 between memory and a register of @var{rclass} but the reload mechanism is
6228 more complex than copying via an intermediate, use this target hook to
6229 reflect the actual cost of the move.
6231 GCC defines the function @code{memory_move_secondary_cost} if
6232 secondary reloads are needed. It computes the costs due to copying via
6233 a secondary register. If your machine copies from memory using a
6234 secondary register in the conventional way but the default base value of
6235 4 is not correct for your machine, use this target hook to add some other
6236 value to the result of that function. The arguments to that function
6237 are the same as to this target hook.
6240 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6241 A C expression for the cost of a branch instruction. A value of 1 is
6242 the default; other values are interpreted relative to that. Parameter
6243 @var{speed_p} is true when the branch in question should be optimized
6244 for speed. When it is false, @code{BRANCH_COST} should return a value
6245 optimal for code size rather than performance. @var{predictable_p} is
6246 true for well-predicted branches. On many architectures the
6247 @code{BRANCH_COST} can be reduced then.
6250 Here are additional macros which do not specify precise relative costs,
6251 but only that certain actions are more expensive than GCC would
6254 @defmac SLOW_BYTE_ACCESS
6255 Define this macro as a C expression which is nonzero if accessing less
6256 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6257 faster than accessing a word of memory, i.e., if such access
6258 require more than one instruction or if there is no difference in cost
6259 between byte and (aligned) word loads.
6261 When this macro is not defined, the compiler will access a field by
6262 finding the smallest containing object; when it is defined, a fullword
6263 load will be used if alignment permits. Unless bytes accesses are
6264 faster than word accesses, using word accesses is preferable since it
6265 may eliminate subsequent memory access if subsequent accesses occur to
6266 other fields in the same word of the structure, but to different bytes.
6269 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6270 Define this macro to be the value 1 if memory accesses described by the
6271 @var{mode} and @var{alignment} parameters have a cost many times greater
6272 than aligned accesses, for example if they are emulated in a trap
6275 When this macro is nonzero, the compiler will act as if
6276 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6277 moves. This can cause significantly more instructions to be produced.
6278 Therefore, do not set this macro nonzero if unaligned accesses only add a
6279 cycle or two to the time for a memory access.
6281 If the value of this macro is always zero, it need not be defined. If
6282 this macro is defined, it should produce a nonzero value when
6283 @code{STRICT_ALIGNMENT} is nonzero.
6286 @defmac MOVE_RATIO (@var{speed})
6287 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6288 which a sequence of insns should be generated instead of a
6289 string move insn or a library call. Increasing the value will always
6290 make code faster, but eventually incurs high cost in increased code size.
6292 Note that on machines where the corresponding move insn is a
6293 @code{define_expand} that emits a sequence of insns, this macro counts
6294 the number of such sequences.
6296 The parameter @var{speed} is true if the code is currently being
6297 optimized for speed rather than size.
6299 If you don't define this, a reasonable default is used.
6302 @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})
6303 GCC will attempt several strategies when asked to copy between
6304 two areas of memory, or to set, clear or store to memory, for example
6305 when copying a @code{struct}. The @code{by_pieces} infrastructure
6306 implements such memory operations as a sequence of load, store or move
6307 insns. Alternate strategies are to expand the
6308 @code{movmem} or @code{setmem} optabs, to emit a library call, or to emit
6309 unit-by-unit, loop-based operations.
6311 This target hook should return true if, for a memory operation with a
6312 given @var{size} and @var{alignment}, using the @code{by_pieces}
6313 infrastructure is expected to result in better code generation.
6314 Both @var{size} and @var{alignment} are measured in terms of storage
6317 The parameter @var{op} is one of: @code{CLEAR_BY_PIECES},
6318 @code{MOVE_BY_PIECES}, @code{SET_BY_PIECES}, @code{STORE_BY_PIECES}.
6319 These describe the type of memory operation under consideration.
6321 The parameter @var{speed_p} is true if the code is currently being
6322 optimized for speed rather than size.
6324 Returning true for higher values of @var{size} can improve code generation
6325 for speed if the target does not provide an implementation of the
6326 @code{movmem} or @code{setmem} standard names, if the @code{movmem} or
6327 @code{setmem} implementation would be more expensive than a sequence of
6328 insns, or if the overhead of a library call would dominate that of
6329 the body of the memory operation.
6331 Returning true for higher values of @code{size} may also cause an increase
6332 in code size, for example where the number of insns emitted to perform a
6333 move would be greater than that of a library call.
6336 @defmac MOVE_MAX_PIECES
6337 A C expression used by @code{move_by_pieces} to determine the largest unit
6338 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6341 @defmac CLEAR_RATIO (@var{speed})
6342 The threshold of number of scalar move insns, @emph{below} which a sequence
6343 of insns should be generated to clear memory instead of a string clear insn
6344 or a library call. Increasing the value will always make code faster, but
6345 eventually incurs high cost in increased code size.
6347 The parameter @var{speed} is true if the code is currently being
6348 optimized for speed rather than size.
6350 If you don't define this, a reasonable default is used.
6353 @defmac SET_RATIO (@var{speed})
6354 The threshold of number of scalar move insns, @emph{below} which a sequence
6355 of insns should be generated to set memory to a constant value, instead of
6356 a block set insn or a library call.
6357 Increasing the value will always make code faster, but
6358 eventually incurs high cost in increased code size.
6360 The parameter @var{speed} is true if the code is currently being
6361 optimized for speed rather than size.
6363 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6366 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6367 A C expression used to determine whether a load postincrement is a good
6368 thing to use for a given mode. Defaults to the value of
6369 @code{HAVE_POST_INCREMENT}.
6372 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6373 A C expression used to determine whether a load postdecrement is a good
6374 thing to use for a given mode. Defaults to the value of
6375 @code{HAVE_POST_DECREMENT}.
6378 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6379 A C expression used to determine whether a load preincrement is a good
6380 thing to use for a given mode. Defaults to the value of
6381 @code{HAVE_PRE_INCREMENT}.
6384 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6385 A C expression used to determine whether a load predecrement is a good
6386 thing to use for a given mode. Defaults to the value of
6387 @code{HAVE_PRE_DECREMENT}.
6390 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6391 A C expression used to determine whether a store postincrement is a good
6392 thing to use for a given mode. Defaults to the value of
6393 @code{HAVE_POST_INCREMENT}.
6396 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6397 A C expression used to determine whether a store postdecrement is a good
6398 thing to use for a given mode. Defaults to the value of
6399 @code{HAVE_POST_DECREMENT}.
6402 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6403 This macro is used to determine whether a store preincrement is a good
6404 thing to use for a given mode. Defaults to the value of
6405 @code{HAVE_PRE_INCREMENT}.
6408 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6409 This macro is used to determine whether a store predecrement is a good
6410 thing to use for a given mode. Defaults to the value of
6411 @code{HAVE_PRE_DECREMENT}.
6414 @defmac NO_FUNCTION_CSE
6415 Define this macro to be true if it is as good or better to call a constant
6416 function address than to call an address kept in a register.
6419 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
6420 Define this macro if a non-short-circuit operation produced by
6421 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6422 @code{BRANCH_COST} is greater than or equal to the value 2.
6425 @deftypefn {Target Hook} bool TARGET_OPTAB_SUPPORTED_P (int @var{op}, machine_mode @var{mode1}, machine_mode @var{mode2}, optimization_type @var{opt_type})
6426 Return true if the optimizers should use optab @var{op} with
6427 modes @var{mode1} and @var{mode2} for optimization type @var{opt_type}.
6428 The optab is known to have an associated @file{.md} instruction
6429 whose C condition is true. @var{mode2} is only meaningful for conversion
6430 optabs; for direct optabs it is a copy of @var{mode1}.
6432 For example, when called with @var{op} equal to @code{rint_optab} and
6433 @var{mode1} equal to @code{DFmode}, the hook should say whether the
6434 optimizers should use optab @code{rintdf2}.
6436 The default hook returns true for all inputs.
6439 @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})
6440 This target hook describes the relative costs of RTL expressions.
6442 The cost may depend on the precise form of the expression, which is
6443 available for examination in @var{x}, and the fact that @var{x} appears
6444 as operand @var{opno} of an expression with rtx code @var{outer_code}.
6445 That is, the hook can assume that there is some rtx @var{y} such
6446 that @samp{GET_CODE (@var{y}) == @var{outer_code}} and such that
6447 either (a) @samp{XEXP (@var{y}, @var{opno}) == @var{x}} or
6448 (b) @samp{XVEC (@var{y}, @var{opno})} contains @var{x}.
6450 @var{mode} is @var{x}'s machine mode, or for cases like @code{const_int} that
6451 do not have a mode, the mode in which @var{x} is used.
6453 In implementing this hook, you can use the construct
6454 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6457 On entry to the hook, @code{*@var{total}} contains a default estimate
6458 for the cost of the expression. The hook should modify this value as
6459 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6460 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6461 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6463 When optimizing for code size, i.e.@: when @code{speed} is
6464 false, this target hook should be used to estimate the relative
6465 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6467 The hook returns true when all subexpressions of @var{x} have been
6468 processed, and false when @code{rtx_cost} should recurse.
6471 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address}, machine_mode @var{mode}, addr_space_t @var{as}, bool @var{speed})
6472 This hook computes the cost of an addressing mode that contains
6473 @var{address}. If not defined, the cost is computed from
6474 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6476 For most CISC machines, the default cost is a good approximation of the
6477 true cost of the addressing mode. However, on RISC machines, all
6478 instructions normally have the same length and execution time. Hence
6479 all addresses will have equal costs.
6481 In cases where more than one form of an address is known, the form with
6482 the lowest cost will be used. If multiple forms have the same, lowest,
6483 cost, the one that is the most complex will be used.
6485 For example, suppose an address that is equal to the sum of a register
6486 and a constant is used twice in the same basic block. When this macro
6487 is not defined, the address will be computed in a register and memory
6488 references will be indirect through that register. On machines where
6489 the cost of the addressing mode containing the sum is no higher than
6490 that of a simple indirect reference, this will produce an additional
6491 instruction and possibly require an additional register. Proper
6492 specification of this macro eliminates this overhead for such machines.
6494 This hook is never called with an invalid address.
6496 On machines where an address involving more than one register is as
6497 cheap as an address computation involving only one register, defining
6498 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6499 be live over a region of code where only one would have been if
6500 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6501 should be considered in the definition of this macro. Equivalent costs
6502 should probably only be given to addresses with different numbers of
6503 registers on machines with lots of registers.
6506 @deftypefn {Target Hook} bool TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P (void)
6507 This predicate controls the use of the eager delay slot filler to disallow
6508 speculatively executed instructions being placed in delay slots. Targets
6509 such as certain MIPS architectures possess both branches with and without
6510 delay slots. As the eager delay slot filler can decrease performance,
6511 disabling it is beneficial when ordinary branches are available. Use of
6512 delay slot branches filled using the basic filler is often still desirable
6513 as the delay slot can hide a pipeline bubble.
6517 @section Adjusting the Instruction Scheduler
6519 The instruction scheduler may need a fair amount of machine-specific
6520 adjustment in order to produce good code. GCC provides several target
6521 hooks for this purpose. It is usually enough to define just a few of
6522 them: try the first ones in this list first.
6524 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6525 This hook returns the maximum number of instructions that can ever
6526 issue at the same time on the target machine. The default is one.
6527 Although the insn scheduler can define itself the possibility of issue
6528 an insn on the same cycle, the value can serve as an additional
6529 constraint to issue insns on the same simulated processor cycle (see
6530 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6531 This value must be constant over the entire compilation. If you need
6532 it to vary depending on what the instructions are, you must use
6533 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6536 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx_insn *@var{insn}, int @var{more})
6537 This hook is executed by the scheduler after it has scheduled an insn
6538 from the ready list. It should return the number of insns which can
6539 still be issued in the current cycle. The default is
6540 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6541 @code{USE}, which normally are not counted against the issue rate.
6542 You should define this hook if some insns take more machine resources
6543 than others, so that fewer insns can follow them in the same cycle.
6544 @var{file} is either a null pointer, or a stdio stream to write any
6545 debug output to. @var{verbose} is the verbose level provided by
6546 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6550 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx_insn *@var{insn}, rtx @var{link}, rtx_insn *@var{dep_insn}, int @var{cost})
6551 This function corrects the value of @var{cost} based on the
6552 relationship between @var{insn} and @var{dep_insn} through the
6553 dependence @var{link}. It should return the new value. The default
6554 is to make no adjustment to @var{cost}. This can be used for example
6555 to specify to the scheduler using the traditional pipeline description
6556 that an output- or anti-dependence does not incur the same cost as a
6557 data-dependence. If the scheduler using the automaton based pipeline
6558 description, the cost of anti-dependence is zero and the cost of
6559 output-dependence is maximum of one and the difference of latency
6560 times of the first and the second insns. If these values are not
6561 acceptable, you could use the hook to modify them too. See also
6562 @pxref{Processor pipeline description}.
6565 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx_insn *@var{insn}, int @var{priority})
6566 This hook adjusts the integer scheduling priority @var{priority} of
6567 @var{insn}. It should return the new priority. Increase the priority to
6568 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6569 later. Do not define this hook if you do not need to adjust the
6570 scheduling priorities of insns.
6573 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock})
6574 This hook is executed by the scheduler after it has scheduled the ready
6575 list, to allow the machine description to reorder it (for example to
6576 combine two small instructions together on @samp{VLIW} machines).
6577 @var{file} is either a null pointer, or a stdio stream to write any
6578 debug output to. @var{verbose} is the verbose level provided by
6579 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6580 list of instructions that are ready to be scheduled. @var{n_readyp} is
6581 a pointer to the number of elements in the ready list. The scheduler
6582 reads the ready list in reverse order, starting with
6583 @var{ready}[@var{*n_readyp} @minus{} 1] and going to @var{ready}[0]. @var{clock}
6584 is the timer tick of the scheduler. You may modify the ready list and
6585 the number of ready insns. The return value is the number of insns that
6586 can issue this cycle; normally this is just @code{issue_rate}. See also
6587 @samp{TARGET_SCHED_REORDER2}.
6590 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx_insn **@var{ready}, int *@var{n_readyp}, int @var{clock})
6591 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6592 function is called whenever the scheduler starts a new cycle. This one
6593 is called once per iteration over a cycle, immediately after
6594 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6595 return the number of insns to be scheduled in the same cycle. Defining
6596 this hook can be useful if there are frequent situations where
6597 scheduling one insn causes other insns to become ready in the same
6598 cycle. These other insns can then be taken into account properly.
6601 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_P (void)
6602 This hook is used to check whether target platform supports macro fusion.
6605 @deftypefn {Target Hook} bool TARGET_SCHED_MACRO_FUSION_PAIR_P (rtx_insn *@var{prev}, rtx_insn *@var{curr})
6606 This hook is used to check whether two insns should be macro fused for
6607 a target microarchitecture. If this hook returns true for the given insn pair
6608 (@var{prev} and @var{curr}), the scheduler will put them into a sched
6609 group, and they will not be scheduled apart. The two insns will be either
6610 two SET insns or a compare and a conditional jump and this hook should
6611 validate any dependencies needed to fuse the two insns together.
6614 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx_insn *@var{head}, rtx_insn *@var{tail})
6615 This hook is called after evaluation forward dependencies of insns in
6616 chain given by two parameter values (@var{head} and @var{tail}
6617 correspondingly) but before insns scheduling of the insn chain. For
6618 example, it can be used for better insn classification if it requires
6619 analysis of dependencies. This hook can use backward and forward
6620 dependencies of the insn scheduler because they are already
6624 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6625 This hook is executed by the scheduler at the beginning of each block of
6626 instructions that are to be scheduled. @var{file} is either a null
6627 pointer, or a stdio stream to write any debug output to. @var{verbose}
6628 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6629 @var{max_ready} is the maximum number of insns in the current scheduling
6630 region that can be live at the same time. This can be used to allocate
6631 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6634 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6635 This hook is executed by the scheduler at the end of each block of
6636 instructions that are to be scheduled. It can be used to perform
6637 cleanup of any actions done by the other scheduling hooks. @var{file}
6638 is either a null pointer, or a stdio stream to write any debug output
6639 to. @var{verbose} is the verbose level provided by
6640 @option{-fsched-verbose-@var{n}}.
6643 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6644 This hook is executed by the scheduler after function level initializations.
6645 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6646 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6647 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6650 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6651 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6652 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6653 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6656 @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6657 The hook returns an RTL insn. The automaton state used in the
6658 pipeline hazard recognizer is changed as if the insn were scheduled
6659 when the new simulated processor cycle starts. Usage of the hook may
6660 simplify the automaton pipeline description for some @acronym{VLIW}
6661 processors. If the hook is defined, it is used only for the automaton
6662 based pipeline description. The default is not to change the state
6663 when the new simulated processor cycle starts.
6666 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6667 The hook can be used to initialize data used by the previous hook.
6670 @deftypefn {Target Hook} {rtx_insn *} TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6671 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6672 to changed the state as if the insn were scheduled when the new
6673 simulated processor cycle finishes.
6676 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6677 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6678 used to initialize data used by the previous hook.
6681 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE (void)
6682 The hook to notify target that the current simulated cycle is about to finish.
6683 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6684 to change the state in more complicated situations - e.g., when advancing
6685 state on a single insn is not enough.
6688 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_ADVANCE_CYCLE (void)
6689 The hook to notify target that new simulated cycle has just started.
6690 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6691 to change the state in more complicated situations - e.g., when advancing
6692 state on a single insn is not enough.
6695 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6696 This hook controls better choosing an insn from the ready insn queue
6697 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6698 chooses the first insn from the queue. If the hook returns a positive
6699 value, an additional scheduler code tries all permutations of
6700 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6701 subsequent ready insns to choose an insn whose issue will result in
6702 maximal number of issued insns on the same cycle. For the
6703 @acronym{VLIW} processor, the code could actually solve the problem of
6704 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6705 rules of @acronym{VLIW} packing are described in the automaton.
6707 This code also could be used for superscalar @acronym{RISC}
6708 processors. Let us consider a superscalar @acronym{RISC} processor
6709 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6710 @var{B}, some insns can be executed only in pipelines @var{B} or
6711 @var{C}, and one insn can be executed in pipeline @var{B}. The
6712 processor may issue the 1st insn into @var{A} and the 2nd one into
6713 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6714 until the next cycle. If the scheduler issues the 3rd insn the first,
6715 the processor could issue all 3 insns per cycle.
6717 Actually this code demonstrates advantages of the automaton based
6718 pipeline hazard recognizer. We try quickly and easy many insn
6719 schedules to choose the best one.
6721 The default is no multipass scheduling.
6724 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx_insn *@var{insn}, int @var{ready_index})
6726 This hook controls what insns from the ready insn queue will be
6727 considered for the multipass insn scheduling. If the hook returns
6728 zero for @var{insn}, the insn will be considered in multipass scheduling.
6729 Positive return values will remove @var{insn} from consideration on
6730 the current round of multipass scheduling.
6731 Negative return values will remove @var{insn} from consideration for given
6733 Backends should be careful about returning non-zero for highest priority
6734 instruction at position 0 in the ready list. @var{ready_index} is passed
6735 to allow backends make correct judgements.
6737 The default is that any ready insns can be chosen to be issued.
6740 @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})
6741 This hook prepares the target backend for a new round of multipass
6745 @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})
6746 This hook is called when multipass scheduling evaluates instruction INSN.
6749 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK (const void *@var{data}, signed char *@var{ready_try}, int @var{n_ready})
6750 This is called when multipass scheduling backtracks from evaluation of
6754 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END (const void *@var{data})
6755 This hook notifies the target about the result of the concluded current
6756 round of multipass scheduling.
6759 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT (void *@var{data})
6760 This hook initializes target-specific data used in multipass scheduling.
6763 @deftypefn {Target Hook} void TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI (void *@var{data})
6764 This hook finalizes target-specific data used in multipass scheduling.
6767 @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})
6768 This hook is called by the insn scheduler before issuing @var{insn}
6769 on cycle @var{clock}. If the hook returns nonzero,
6770 @var{insn} is not issued on this processor cycle. Instead,
6771 the processor cycle is advanced. If *@var{sort_p}
6772 is zero, the insn ready queue is not sorted on the new cycle
6773 start as usually. @var{dump} and @var{verbose} specify the file and
6774 verbosity level to use for debugging output.
6775 @var{last_clock} and @var{clock} are, respectively, the
6776 processor cycle on which the previous insn has been issued,
6777 and the current processor cycle.
6780 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct _dep *@var{_dep}, int @var{cost}, int @var{distance})
6781 This hook is used to define which dependences are considered costly by
6782 the target, so costly that it is not advisable to schedule the insns that
6783 are involved in the dependence too close to one another. The parameters
6784 to this hook are as follows: The first parameter @var{_dep} is the dependence
6785 being evaluated. The second parameter @var{cost} is the cost of the
6786 dependence as estimated by the scheduler, and the third
6787 parameter @var{distance} is the distance in cycles between the two insns.
6788 The hook returns @code{true} if considering the distance between the two
6789 insns the dependence between them is considered costly by the target,
6790 and @code{false} otherwise.
6792 Defining this hook can be useful in multiple-issue out-of-order machines,
6793 where (a) it's practically hopeless to predict the actual data/resource
6794 delays, however: (b) there's a better chance to predict the actual grouping
6795 that will be formed, and (c) correctly emulating the grouping can be very
6796 important. In such targets one may want to allow issuing dependent insns
6797 closer to one another---i.e., closer than the dependence distance; however,
6798 not in cases of ``costly dependences'', which this hooks allows to define.
6801 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6802 This hook is called by the insn scheduler after emitting a new instruction to
6803 the instruction stream. The hook notifies a target backend to extend its
6804 per instruction data structures.
6807 @deftypefn {Target Hook} {void *} TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6808 Return a pointer to a store large enough to hold target scheduling context.
6811 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6812 Initialize store pointed to by @var{tc} to hold target scheduling context.
6813 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6814 beginning of the block. Otherwise, copy the current context into @var{tc}.
6817 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6818 Copy target scheduling context pointed to by @var{tc} to the current context.
6821 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6822 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6825 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6826 Deallocate a store for target scheduling context pointed to by @var{tc}.
6829 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx_insn *@var{insn}, unsigned int @var{dep_status}, rtx *@var{new_pat})
6830 This hook is called by the insn scheduler when @var{insn} has only
6831 speculative dependencies and therefore can be scheduled speculatively.
6832 The hook is used to check if the pattern of @var{insn} has a speculative
6833 version and, in case of successful check, to generate that speculative
6834 pattern. The hook should return 1, if the instruction has a speculative form,
6835 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6836 speculation. If the return value equals 1 then @var{new_pat} is assigned
6837 the generated speculative pattern.
6840 @deftypefn {Target Hook} bool TARGET_SCHED_NEEDS_BLOCK_P (unsigned int @var{dep_status})
6841 This hook is called by the insn scheduler during generation of recovery code
6842 for @var{insn}. It should return @code{true}, if the corresponding check
6843 instruction should branch to recovery code, or @code{false} otherwise.
6846 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_SPEC_CHECK (rtx_insn *@var{insn}, rtx_insn *@var{label}, unsigned int @var{ds})
6847 This hook is called by the insn scheduler to generate a pattern for recovery
6848 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6849 speculative instruction for which the check should be generated.
6850 @var{label} is either a label of a basic block, where recovery code should
6851 be emitted, or a null pointer, when requested check doesn't branch to
6852 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6853 a pattern for a branchy check corresponding to a simple check denoted by
6854 @var{insn} should be generated. In this case @var{label} can't be null.
6857 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (struct spec_info_def *@var{spec_info})
6858 This hook is used by the insn scheduler to find out what features should be
6860 The structure *@var{spec_info} should be filled in by the target.
6861 The structure describes speculation types that can be used in the scheduler.
6864 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6865 This hook is called by the swing modulo scheduler to calculate a
6866 resource-based lower bound which is based on the resources available in
6867 the machine and the resources required by each instruction. The target
6868 backend can use @var{g} to calculate such bound. A very simple lower
6869 bound will be used in case this hook is not implemented: the total number
6870 of instructions divided by the issue rate.
6873 @deftypefn {Target Hook} bool TARGET_SCHED_DISPATCH (rtx_insn *@var{insn}, int @var{x})
6874 This hook is called by Haifa Scheduler. It returns true if dispatch scheduling
6875 is supported in hardware and the condition specified in the parameter is true.
6878 @deftypefn {Target Hook} void TARGET_SCHED_DISPATCH_DO (rtx_insn *@var{insn}, int @var{x})
6879 This hook is called by Haifa Scheduler. It performs the operation specified
6880 in its second parameter.
6883 @deftypevr {Target Hook} bool TARGET_SCHED_EXPOSED_PIPELINE
6884 True if the processor has an exposed pipeline, which means that not just
6885 the order of instructions is important for correctness when scheduling, but
6886 also the latencies of operations.
6889 @deftypefn {Target Hook} int TARGET_SCHED_REASSOCIATION_WIDTH (unsigned int @var{opc}, machine_mode @var{mode})
6890 This hook is called by tree reassociator to determine a level of
6891 parallelism required in output calculations chain.
6894 @deftypefn {Target Hook} void TARGET_SCHED_FUSION_PRIORITY (rtx_insn *@var{insn}, int @var{max_pri}, int *@var{fusion_pri}, int *@var{pri})
6895 This hook is called by scheduling fusion pass. It calculates fusion
6896 priorities for each instruction passed in by parameter. The priorities
6897 are returned via pointer parameters.
6899 @var{insn} is the instruction whose priorities need to be calculated.
6900 @var{max_pri} is the maximum priority can be returned in any cases.
6901 @var{fusion_pri} is the pointer parameter through which @var{insn}'s
6902 fusion priority should be calculated and returned.
6903 @var{pri} is the pointer parameter through which @var{insn}'s priority
6904 should be calculated and returned.
6906 Same @var{fusion_pri} should be returned for instructions which should
6907 be scheduled together. Different @var{pri} should be returned for
6908 instructions with same @var{fusion_pri}. @var{fusion_pri} is the major
6909 sort key, @var{pri} is the minor sort key. All instructions will be
6910 scheduled according to the two priorities. All priorities calculated
6911 should be between 0 (exclusive) and @var{max_pri} (inclusive). To avoid
6912 false dependencies, @var{fusion_pri} of instructions which need to be
6913 scheduled together should be smaller than @var{fusion_pri} of irrelevant
6916 Given below example:
6929 On targets like ARM/AArch64, the two pairs of consecutive loads should be
6930 merged. Since peephole2 pass can't help in this case unless consecutive
6931 loads are actually next to each other in instruction flow. That's where
6932 this scheduling fusion pass works. This hook calculates priority for each
6933 instruction based on its fustion type, like:
6936 ldr r10, [r1, 4] ; fusion_pri=99, pri=96
6937 add r4, r4, r10 ; fusion_pri=100, pri=100
6938 ldr r15, [r2, 8] ; fusion_pri=98, pri=92
6939 sub r5, r5, r15 ; fusion_pri=100, pri=100
6940 ldr r11, [r1, 0] ; fusion_pri=99, pri=100
6941 add r4, r4, r11 ; fusion_pri=100, pri=100
6942 ldr r16, [r2, 12] ; fusion_pri=98, pri=88
6943 sub r5, r5, r16 ; fusion_pri=100, pri=100
6946 Scheduling fusion pass then sorts all ready to issue instructions according
6947 to the priorities. As a result, instructions of same fusion type will be
6948 pushed together in instruction flow, like:
6961 Now peephole2 pass can simply merge the two pairs of loads.
6963 Since scheduling fusion pass relies on peephole2 to do real fusion
6964 work, it is only enabled by default when peephole2 is in effect.
6966 This is firstly introduced on ARM/AArch64 targets, please refer to
6967 the hook implementation for how different fusion types are supported.
6971 @section Dividing the Output into Sections (Texts, Data, @dots{})
6972 @c the above section title is WAY too long. maybe cut the part between
6973 @c the (...)? --mew 10feb93
6975 An object file is divided into sections containing different types of
6976 data. In the most common case, there are three sections: the @dfn{text
6977 section}, which holds instructions and read-only data; the @dfn{data
6978 section}, which holds initialized writable data; and the @dfn{bss
6979 section}, which holds uninitialized data. Some systems have other kinds
6982 @file{varasm.c} provides several well-known sections, such as
6983 @code{text_section}, @code{data_section} and @code{bss_section}.
6984 The normal way of controlling a @code{@var{foo}_section} variable
6985 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6986 as described below. The macros are only read once, when @file{varasm.c}
6987 initializes itself, so their values must be run-time constants.
6988 They may however depend on command-line flags.
6990 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6991 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6992 to be string literals.
6994 Some assemblers require a different string to be written every time a
6995 section is selected. If your assembler falls into this category, you
6996 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6997 @code{get_unnamed_section} to set up the sections.
6999 You must always create a @code{text_section}, either by defining
7000 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
7001 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
7002 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
7003 create a distinct @code{readonly_data_section}, the default is to
7004 reuse @code{text_section}.
7006 All the other @file{varasm.c} sections are optional, and are null
7007 if the target does not provide them.
7009 @defmac TEXT_SECTION_ASM_OP
7010 A C expression whose value is a string, including spacing, containing the
7011 assembler operation that should precede instructions and read-only data.
7012 Normally @code{"\t.text"} is right.
7015 @defmac HOT_TEXT_SECTION_NAME
7016 If defined, a C string constant for the name of the section containing most
7017 frequently executed functions of the program. If not defined, GCC will provide
7018 a default definition if the target supports named sections.
7021 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
7022 If defined, a C string constant for the name of the section containing unlikely
7023 executed functions in the program.
7026 @defmac DATA_SECTION_ASM_OP
7027 A C expression whose value is a string, including spacing, containing the
7028 assembler operation to identify the following data as writable initialized
7029 data. Normally @code{"\t.data"} is right.
7032 @defmac SDATA_SECTION_ASM_OP
7033 If defined, a C expression whose value is a string, including spacing,
7034 containing the assembler operation to identify the following data as
7035 initialized, writable small data.
7038 @defmac READONLY_DATA_SECTION_ASM_OP
7039 A C expression whose value is a string, including spacing, containing the
7040 assembler operation to identify the following data as read-only initialized
7044 @defmac BSS_SECTION_ASM_OP
7045 If defined, a C expression whose value is a string, including spacing,
7046 containing the assembler operation to identify the following data as
7047 uninitialized global data. If not defined, and
7048 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
7049 uninitialized global data will be output in the data section if
7050 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
7054 @defmac SBSS_SECTION_ASM_OP
7055 If defined, a C expression whose value is a string, including spacing,
7056 containing the assembler operation to identify the following data as
7057 uninitialized, writable small data.
7060 @defmac TLS_COMMON_ASM_OP
7061 If defined, a C expression whose value is a string containing the
7062 assembler operation to identify the following data as thread-local
7063 common data. The default is @code{".tls_common"}.
7066 @defmac TLS_SECTION_ASM_FLAG
7067 If defined, a C expression whose value is a character constant
7068 containing the flag used to mark a section as a TLS section. The
7069 default is @code{'T'}.
7072 @defmac INIT_SECTION_ASM_OP
7073 If defined, a C expression whose value is a string, including spacing,
7074 containing the assembler operation to identify the following data as
7075 initialization code. If not defined, GCC will assume such a section does
7076 not exist. This section has no corresponding @code{init_section}
7077 variable; it is used entirely in runtime code.
7080 @defmac FINI_SECTION_ASM_OP
7081 If defined, a C expression whose value is a string, including spacing,
7082 containing the assembler operation to identify the following data as
7083 finalization code. If not defined, GCC will assume such a section does
7084 not exist. This section has no corresponding @code{fini_section}
7085 variable; it is used entirely in runtime code.
7088 @defmac INIT_ARRAY_SECTION_ASM_OP
7089 If defined, a C expression whose value is a string, including spacing,
7090 containing the assembler operation to identify the following data as
7091 part of the @code{.init_array} (or equivalent) section. If not
7092 defined, GCC will assume such a section does not exist. Do not define
7093 both this macro and @code{INIT_SECTION_ASM_OP}.
7096 @defmac FINI_ARRAY_SECTION_ASM_OP
7097 If defined, a C expression whose value is a string, including spacing,
7098 containing the assembler operation to identify the following data as
7099 part of the @code{.fini_array} (or equivalent) section. If not
7100 defined, GCC will assume such a section does not exist. Do not define
7101 both this macro and @code{FINI_SECTION_ASM_OP}.
7104 @defmac MACH_DEP_SECTION_ASM_FLAG
7105 If defined, a C expression whose value is a character constant
7106 containing the flag used to mark a machine-dependent section. This
7107 corresponds to the @code{SECTION_MACH_DEP} section flag.
7110 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
7111 If defined, an ASM statement that switches to a different section
7112 via @var{section_op}, calls @var{function}, and switches back to
7113 the text section. This is used in @file{crtstuff.c} if
7114 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
7115 to initialization and finalization functions from the init and fini
7116 sections. By default, this macro uses a simple function call. Some
7117 ports need hand-crafted assembly code to avoid dependencies on
7118 registers initialized in the function prologue or to ensure that
7119 constant pools don't end up too far way in the text section.
7122 @defmac TARGET_LIBGCC_SDATA_SECTION
7123 If defined, a string which names the section into which small
7124 variables defined in crtstuff and libgcc should go. This is useful
7125 when the target has options for optimizing access to small data, and
7126 you want the crtstuff and libgcc routines to be conservative in what
7127 they expect of your application yet liberal in what your application
7128 expects. For example, for targets with a @code{.sdata} section (like
7129 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
7130 require small data support from your application, but use this macro
7131 to put small data into @code{.sdata} so that your application can
7132 access these variables whether it uses small data or not.
7135 @defmac FORCE_CODE_SECTION_ALIGN
7136 If defined, an ASM statement that aligns a code section to some
7137 arbitrary boundary. This is used to force all fragments of the
7138 @code{.init} and @code{.fini} sections to have to same alignment
7139 and thus prevent the linker from having to add any padding.
7142 @defmac JUMP_TABLES_IN_TEXT_SECTION
7143 Define this macro to be an expression with a nonzero value if jump
7144 tables (for @code{tablejump} insns) should be output in the text
7145 section, along with the assembler instructions. Otherwise, the
7146 readonly data section is used.
7148 This macro is irrelevant if there is no separate readonly data section.
7151 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
7152 Define this hook if you need to do something special to set up the
7153 @file{varasm.c} sections, or if your target has some special sections
7154 of its own that you need to create.
7156 GCC calls this hook after processing the command line, but before writing
7157 any assembly code, and before calling any of the section-returning hooks
7161 @deftypefn {Target Hook} int TARGET_ASM_RELOC_RW_MASK (void)
7162 Return a mask describing how relocations should be treated when
7163 selecting sections. Bit 1 should be set if global relocations
7164 should be placed in a read-write section; bit 0 should be set if
7165 local relocations should be placed in a read-write section.
7167 The default version of this function returns 3 when @option{-fpic}
7168 is in effect, and 0 otherwise. The hook is typically redefined
7169 when the target cannot support (some kinds of) dynamic relocations
7170 in read-only sections even in executables.
7173 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
7174 Return the section into which @var{exp} should be placed. You can
7175 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
7176 some sort. @var{reloc} indicates whether the initial value of @var{exp}
7177 requires link-time relocations. Bit 0 is set when variable contains
7178 local relocations only, while bit 1 is set for global relocations.
7179 @var{align} is the constant alignment in bits.
7181 The default version of this function takes care of putting read-only
7182 variables in @code{readonly_data_section}.
7184 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
7187 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
7188 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
7189 for @code{FUNCTION_DECL}s as well as for variables and constants.
7191 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
7192 function has been determined to be likely to be called, and nonzero if
7193 it is unlikely to be called.
7196 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
7197 Build up a unique section name, expressed as a @code{STRING_CST} node,
7198 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
7199 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
7200 the initial value of @var{exp} requires link-time relocations.
7202 The default version of this function appends the symbol name to the
7203 ELF section name that would normally be used for the symbol. For
7204 example, the function @code{foo} would be placed in @code{.text.foo}.
7205 Whatever the actual target object format, this is often good enough.
7208 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
7209 Return the readonly data section associated with
7210 @samp{DECL_SECTION_NAME (@var{decl})}.
7211 The default version of this function selects @code{.gnu.linkonce.r.name} if
7212 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
7213 if function is in @code{.text.name}, and the normal readonly-data section
7217 @deftypevr {Target Hook} {const char *} TARGET_ASM_MERGEABLE_RODATA_PREFIX
7218 Usually, the compiler uses the prefix @code{".rodata"} to construct
7219 section names for mergeable constant data. Define this macro to override
7220 the string if a different section name should be used.
7223 @deftypefn {Target Hook} {section *} TARGET_ASM_TM_CLONE_TABLE_SECTION (void)
7224 Return the section that should be used for transactional memory clone tables.
7227 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
7228 Return the section into which a constant @var{x}, of mode @var{mode},
7229 should be placed. You can assume that @var{x} is some kind of
7230 constant in RTL@. The argument @var{mode} is redundant except in the
7231 case of a @code{const_int} rtx. @var{align} is the constant alignment
7234 The default version of this function takes care of putting symbolic
7235 constants in @code{flag_pic} mode in @code{data_section} and everything
7236 else in @code{readonly_data_section}.
7239 @deftypefn {Target Hook} tree TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
7240 Define this hook if you need to postprocess the assembler name generated
7241 by target-independent code. The @var{id} provided to this hook will be
7242 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
7243 or the mangled name of the @var{decl} in C++). The return value of the
7244 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
7245 your target system. The default implementation of this hook just
7246 returns the @var{id} provided.
7249 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
7250 Define this hook if references to a symbol or a constant must be
7251 treated differently depending on something about the variable or
7252 function named by the symbol (such as what section it is in).
7254 The hook is executed immediately after rtl has been created for
7255 @var{decl}, which may be a variable or function declaration or
7256 an entry in the constant pool. In either case, @var{rtl} is the
7257 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
7258 in this hook; that field may not have been initialized yet.
7260 In the case of a constant, it is safe to assume that the rtl is
7261 a @code{mem} whose address is a @code{symbol_ref}. Most decls
7262 will also have this form, but that is not guaranteed. Global
7263 register variables, for instance, will have a @code{reg} for their
7264 rtl. (Normally the right thing to do with such unusual rtl is
7267 The @var{new_decl_p} argument will be true if this is the first time
7268 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
7269 be false for subsequent invocations, which will happen for duplicate
7270 declarations. Whether or not anything must be done for the duplicate
7271 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
7272 @var{new_decl_p} is always true when the hook is called for a constant.
7274 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
7275 The usual thing for this hook to do is to record flags in the
7276 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
7277 Historically, the name string was modified if it was necessary to
7278 encode more than one bit of information, but this practice is now
7279 discouraged; use @code{SYMBOL_REF_FLAGS}.
7281 The default definition of this hook, @code{default_encode_section_info}
7282 in @file{varasm.c}, sets a number of commonly-useful bits in
7283 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
7284 before overriding it.
7287 @deftypefn {Target Hook} {const char *} TARGET_STRIP_NAME_ENCODING (const char *@var{name})
7288 Decode @var{name} and return the real name part, sans
7289 the characters that @code{TARGET_ENCODE_SECTION_INFO}
7293 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (const_tree @var{exp})
7294 Returns true if @var{exp} should be placed into a ``small data'' section.
7295 The default version of this hook always returns false.
7298 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
7299 Contains the value true if the target places read-only
7300 ``small data'' into a separate section. The default value is false.
7303 @deftypefn {Target Hook} bool TARGET_PROFILE_BEFORE_PROLOGUE (void)
7304 It returns true if target wants profile code emitted before prologue.
7306 The default version of this hook use the target macro
7307 @code{PROFILE_BEFORE_PROLOGUE}.
7310 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (const_tree @var{exp})
7311 Returns true if @var{exp} names an object for which name resolution
7312 rules must resolve to the current ``module'' (dynamic shared library
7313 or executable image).
7315 The default version of this hook implements the name resolution rules
7316 for ELF, which has a looser model of global name binding than other
7317 currently supported object file formats.
7320 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
7321 Contains the value true if the target supports thread-local storage.
7322 The default value is false.
7327 @section Position Independent Code
7328 @cindex position independent code
7331 This section describes macros that help implement generation of position
7332 independent code. Simply defining these macros is not enough to
7333 generate valid PIC; you must also add support to the hook
7334 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7335 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7336 must modify the definition of @samp{movsi} to do something appropriate
7337 when the source operand contains a symbolic address. You may also
7338 need to alter the handling of switch statements so that they use
7340 @c i rearranged the order of the macros above to try to force one of
7341 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7343 @defmac PIC_OFFSET_TABLE_REGNUM
7344 The register number of the register used to address a table of static
7345 data addresses in memory. In some cases this register is defined by a
7346 processor's ``application binary interface'' (ABI)@. When this macro
7347 is defined, RTL is generated for this register once, as with the stack
7348 pointer and frame pointer registers. If this macro is not defined, it
7349 is up to the machine-dependent files to allocate such a register (if
7350 necessary). Note that this register must be fixed when in use (e.g.@:
7351 when @code{flag_pic} is true).
7354 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7355 A C expression that is nonzero if the register defined by
7356 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
7357 the default is zero. Do not define
7358 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7361 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7362 A C expression that is nonzero if @var{x} is a legitimate immediate
7363 operand on the target machine when generating position independent code.
7364 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7365 check this. You can also assume @var{flag_pic} is true, so you need not
7366 check it either. You need not define this macro if all constants
7367 (including @code{SYMBOL_REF}) can be immediate operands when generating
7368 position independent code.
7371 @node Assembler Format
7372 @section Defining the Output Assembler Language
7374 This section describes macros whose principal purpose is to describe how
7375 to write instructions in assembler language---rather than what the
7379 * File Framework:: Structural information for the assembler file.
7380 * Data Output:: Output of constants (numbers, strings, addresses).
7381 * Uninitialized Data:: Output of uninitialized variables.
7382 * Label Output:: Output and generation of labels.
7383 * Initialization:: General principles of initialization
7384 and termination routines.
7385 * Macros for Initialization::
7386 Specific macros that control the handling of
7387 initialization and termination routines.
7388 * Instruction Output:: Output of actual instructions.
7389 * Dispatch Tables:: Output of jump tables.
7390 * Exception Region Output:: Output of exception region code.
7391 * Alignment Output:: Pseudo ops for alignment and skipping data.
7394 @node File Framework
7395 @subsection The Overall Framework of an Assembler File
7396 @cindex assembler format
7397 @cindex output of assembler code
7399 @c prevent bad page break with this line
7400 This describes the overall framework of an assembly file.
7402 @findex default_file_start
7403 @deftypefn {Target Hook} void TARGET_ASM_FILE_START (void)
7404 Output to @code{asm_out_file} any text which the assembler expects to
7405 find at the beginning of a file. The default behavior is controlled
7406 by two flags, documented below. Unless your target's assembler is
7407 quite unusual, if you override the default, you should call
7408 @code{default_file_start} at some point in your target hook. This
7409 lets other target files rely on these variables.
7412 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7413 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7414 printed as the very first line in the assembly file, unless
7415 @option{-fverbose-asm} is in effect. (If that macro has been defined
7416 to the empty string, this variable has no effect.) With the normal
7417 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7418 assembler that it need not bother stripping comments or extra
7419 whitespace from its input. This allows it to work a bit faster.
7421 The default is false. You should not set it to true unless you have
7422 verified that your port does not generate any extra whitespace or
7423 comments that will cause GAS to issue errors in NO_APP mode.
7426 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7427 If this flag is true, @code{output_file_directive} will be called
7428 for the primary source file, immediately after printing
7429 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7430 this to be done. The default is false.
7433 @deftypefn {Target Hook} void TARGET_ASM_FILE_END (void)
7434 Output to @code{asm_out_file} any text which the assembler expects
7435 to find at the end of a file. The default is to output nothing.
7438 @deftypefun void file_end_indicate_exec_stack ()
7439 Some systems use a common convention, the @samp{.note.GNU-stack}
7440 special section, to indicate whether or not an object file relies on
7441 the stack being executable. If your system uses this convention, you
7442 should define @code{TARGET_ASM_FILE_END} to this function. If you
7443 need to do other things in that hook, have your hook function call
7447 @deftypefn {Target Hook} void TARGET_ASM_LTO_START (void)
7448 Output to @code{asm_out_file} any text which the assembler expects
7449 to find at the start of an LTO section. The default is to output
7453 @deftypefn {Target Hook} void TARGET_ASM_LTO_END (void)
7454 Output to @code{asm_out_file} any text which the assembler expects
7455 to find at the end of an LTO section. The default is to output
7459 @deftypefn {Target Hook} void TARGET_ASM_CODE_END (void)
7460 Output to @code{asm_out_file} any text which is needed before emitting
7461 unwind info and debug info at the end of a file. Some targets emit
7462 here PIC setup thunks that cannot be emitted at the end of file,
7463 because they couldn't have unwind info then. The default is to output
7467 @defmac ASM_COMMENT_START
7468 A C string constant describing how to begin a comment in the target
7469 assembler language. The compiler assumes that the comment will end at
7470 the end of the line.
7474 A C string constant for text to be output before each @code{asm}
7475 statement or group of consecutive ones. Normally this is
7476 @code{"#APP"}, which is a comment that has no effect on most
7477 assemblers but tells the GNU assembler that it must check the lines
7478 that follow for all valid assembler constructs.
7482 A C string constant for text to be output after each @code{asm}
7483 statement or group of consecutive ones. Normally this is
7484 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7485 time-saving assumptions that are valid for ordinary compiler output.
7488 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7489 A C statement to output COFF information or DWARF debugging information
7490 which indicates that filename @var{name} is the current source file to
7491 the stdio stream @var{stream}.
7493 This macro need not be defined if the standard form of output
7494 for the file format in use is appropriate.
7497 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_SOURCE_FILENAME (FILE *@var{file}, const char *@var{name})
7498 Output COFF information or DWARF debugging information which indicates that filename @var{name} is the current source file to the stdio stream @var{file}.
7500 This target hook need not be defined if the standard form of output for the file format in use is appropriate.
7503 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_IDENT (const char *@var{name})
7504 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.
7507 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7508 A C statement to output the string @var{string} to the stdio stream
7509 @var{stream}. If you do not call the function @code{output_quoted_string}
7510 in your config files, GCC will only call it to output filenames to
7511 the assembler source. So you can use it to canonicalize the format
7512 of the filename using this macro.
7515 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, tree @var{decl})
7516 Output assembly directives to switch to section @var{name}. The section
7517 should have attributes as specified by @var{flags}, which is a bit mask
7518 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{decl}
7519 is non-NULL, it is the @code{VAR_DECL} or @code{FUNCTION_DECL} with which
7520 this section is associated.
7523 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_SECTION (tree @var{decl}, enum node_frequency @var{freq}, bool @var{startup}, bool @var{exit})
7524 Return preferred text (sub)section for function @var{decl}.
7525 Main purpose of this function is to separate cold, normal and hot
7526 functions. @var{startup} is true when function is known to be used only
7527 at startup (from static constructors or it is @code{main()}).
7528 @var{exit} is true when function is known to be used only at exit
7529 (from static destructors).
7530 Return NULL if function should go to default text section.
7533 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS (FILE *@var{file}, tree @var{decl}, bool @var{new_is_cold})
7534 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}.
7537 @deftypevr {Common Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7538 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7539 It must not be modified by command-line option processing.
7542 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7543 @deftypevr {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7544 This flag is true if we can create zeroed data by switching to a BSS
7545 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7546 This is true on most ELF targets.
7549 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7550 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7551 based on a variable or function decl, a section name, and whether or not the
7552 declaration's initializer may contain runtime relocations. @var{decl} may be
7553 null, in which case read-write data should be assumed.
7555 The default version of this function handles choosing code vs data,
7556 read-only vs read-write data, and @code{flag_pic}. You should only
7557 need to override this if your target has special flags that might be
7558 set via @code{__attribute__}.
7561 @deftypefn {Target Hook} int TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char *@var{text})
7562 Provides the target with the ability to record the gcc command line
7563 switches that have been passed to the compiler, and options that are
7564 enabled. The @var{type} argument specifies what is being recorded.
7565 It can take the following values:
7568 @item SWITCH_TYPE_PASSED
7569 @var{text} is a command line switch that has been set by the user.
7571 @item SWITCH_TYPE_ENABLED
7572 @var{text} is an option which has been enabled. This might be as a
7573 direct result of a command line switch, or because it is enabled by
7574 default or because it has been enabled as a side effect of a different
7575 command line switch. For example, the @option{-O2} switch enables
7576 various different individual optimization passes.
7578 @item SWITCH_TYPE_DESCRIPTIVE
7579 @var{text} is either NULL or some descriptive text which should be
7580 ignored. If @var{text} is NULL then it is being used to warn the
7581 target hook that either recording is starting or ending. The first
7582 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7583 warning is for start up and the second time the warning is for
7584 wind down. This feature is to allow the target hook to make any
7585 necessary preparations before it starts to record switches and to
7586 perform any necessary tidying up after it has finished recording
7589 @item SWITCH_TYPE_LINE_START
7590 This option can be ignored by this target hook.
7592 @item SWITCH_TYPE_LINE_END
7593 This option can be ignored by this target hook.
7596 The hook's return value must be zero. Other return values may be
7597 supported in the future.
7599 By default this hook is set to NULL, but an example implementation is
7600 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7601 it records the switches as ASCII text inside a new, string mergeable
7602 section in the assembler output file. The name of the new section is
7603 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7607 @deftypevr {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7608 This is the name of the section that will be created by the example
7609 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7615 @subsection Output of Data
7618 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7619 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7620 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7621 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7622 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7623 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7624 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7625 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7626 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7627 These hooks specify assembly directives for creating certain kinds
7628 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7629 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7630 aligned two-byte object, and so on. Any of the hooks may be
7631 @code{NULL}, indicating that no suitable directive is available.
7633 The compiler will print these strings at the start of a new line,
7634 followed immediately by the object's initial value. In most cases,
7635 the string should contain a tab, a pseudo-op, and then another tab.
7638 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7639 The @code{assemble_integer} function uses this hook to output an
7640 integer object. @var{x} is the object's value, @var{size} is its size
7641 in bytes and @var{aligned_p} indicates whether it is aligned. The
7642 function should return @code{true} if it was able to output the
7643 object. If it returns false, @code{assemble_integer} will try to
7644 split the object into smaller parts.
7646 The default implementation of this hook will use the
7647 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7648 when the relevant string is @code{NULL}.
7651 @deftypefn {Target Hook} void TARGET_ASM_DECL_END (void)
7652 Define this hook if the target assembler requires a special marker to
7653 terminate an initialized variable declaration.
7656 @deftypefn {Target Hook} bool TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA (FILE *@var{file}, rtx @var{x})
7657 A target hook to recognize @var{rtx} patterns that @code{output_addr_const}
7658 can't deal with, and output assembly code to @var{file} corresponding to
7659 the pattern @var{x}. This may be used to allow machine-dependent
7660 @code{UNSPEC}s to appear within constants.
7662 If target hook fails to recognize a pattern, it must return @code{false},
7663 so that a standard error message is printed. If it prints an error message
7664 itself, by calling, for example, @code{output_operand_lossage}, it may just
7668 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7669 A C statement to output to the stdio stream @var{stream} an assembler
7670 instruction to assemble a string constant containing the @var{len}
7671 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7672 @code{char *} and @var{len} a C expression of type @code{int}.
7674 If the assembler has a @code{.ascii} pseudo-op as found in the
7675 Berkeley Unix assembler, do not define the macro
7676 @code{ASM_OUTPUT_ASCII}.
7679 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7680 A C statement to output word @var{n} of a function descriptor for
7681 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7682 is defined, and is otherwise unused.
7685 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7686 You may define this macro as a C expression. You should define the
7687 expression to have a nonzero value if GCC should output the constant
7688 pool for a function before the code for the function, or a zero value if
7689 GCC should output the constant pool after the function. If you do
7690 not define this macro, the usual case, GCC will output the constant
7691 pool before the function.
7694 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7695 A C statement to output assembler commands to define the start of the
7696 constant pool for a function. @var{funname} is a string giving
7697 the name of the function. Should the return type of the function
7698 be required, it can be obtained via @var{fundecl}. @var{size}
7699 is the size, in bytes, of the constant pool that will be written
7700 immediately after this call.
7702 If no constant-pool prefix is required, the usual case, this macro need
7706 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7707 A C statement (with or without semicolon) to output a constant in the
7708 constant pool, if it needs special treatment. (This macro need not do
7709 anything for RTL expressions that can be output normally.)
7711 The argument @var{file} is the standard I/O stream to output the
7712 assembler code on. @var{x} is the RTL expression for the constant to
7713 output, and @var{mode} is the machine mode (in case @var{x} is a
7714 @samp{const_int}). @var{align} is the required alignment for the value
7715 @var{x}; you should output an assembler directive to force this much
7718 The argument @var{labelno} is a number to use in an internal label for
7719 the address of this pool entry. The definition of this macro is
7720 responsible for outputting the label definition at the proper place.
7721 Here is how to do this:
7724 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7727 When you output a pool entry specially, you should end with a
7728 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7729 entry from being output a second time in the usual manner.
7731 You need not define this macro if it would do nothing.
7734 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7735 A C statement to output assembler commands to at the end of the constant
7736 pool for a function. @var{funname} is a string giving the name of the
7737 function. Should the return type of the function be required, you can
7738 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7739 constant pool that GCC wrote immediately before this call.
7741 If no constant-pool epilogue is required, the usual case, you need not
7745 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7746 Define this macro as a C expression which is nonzero if @var{C} is
7747 used as a logical line separator by the assembler. @var{STR} points
7748 to the position in the string where @var{C} was found; this can be used if
7749 a line separator uses multiple characters.
7751 If you do not define this macro, the default is that only
7752 the character @samp{;} is treated as a logical line separator.
7755 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7756 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7757 These target hooks are C string constants, describing the syntax in the
7758 assembler for grouping arithmetic expressions. If not overridden, they
7759 default to normal parentheses, which is correct for most assemblers.
7762 These macros are provided by @file{real.h} for writing the definitions
7763 of @code{ASM_OUTPUT_DOUBLE} and the like:
7765 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7766 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7767 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7768 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7769 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7770 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7771 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7772 target's floating point representation, and store its bit pattern in
7773 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7774 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7775 simple @code{long int}. For the others, it should be an array of
7776 @code{long int}. The number of elements in this array is determined
7777 by the size of the desired target floating point data type: 32 bits of
7778 it go in each @code{long int} array element. Each array element holds
7779 32 bits of the result, even if @code{long int} is wider than 32 bits
7780 on the host machine.
7782 The array element values are designed so that you can print them out
7783 using @code{fprintf} in the order they should appear in the target
7787 @node Uninitialized Data
7788 @subsection Output of Uninitialized Variables
7790 Each of the macros in this section is used to do the whole job of
7791 outputting a single uninitialized variable.
7793 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7794 A C statement (sans semicolon) to output to the stdio stream
7795 @var{stream} the assembler definition of a common-label named
7796 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7797 is the size rounded up to whatever alignment the caller wants. It is
7798 possible that @var{size} may be zero, for instance if a struct with no
7799 other member than a zero-length array is defined. In this case, the
7800 backend must output a symbol definition that allocates at least one
7801 byte, both so that the address of the resulting object does not compare
7802 equal to any other, and because some object formats cannot even express
7803 the concept of a zero-sized common symbol, as that is how they represent
7804 an ordinary undefined external.
7806 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7807 output the name itself; before and after that, output the additional
7808 assembler syntax for defining the name, and a newline.
7810 This macro controls how the assembler definitions of uninitialized
7811 common global variables are output.
7814 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7815 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7816 separate, explicit argument. If you define this macro, it is used in
7817 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7818 handling the required alignment of the variable. The alignment is specified
7819 as the number of bits.
7822 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7823 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7824 variable to be output, if there is one, or @code{NULL_TREE} if there
7825 is no corresponding variable. If you define this macro, GCC will use it
7826 in place of both @code{ASM_OUTPUT_COMMON} and
7827 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7828 the variable's decl in order to chose what to output.
7831 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7832 A C statement (sans semicolon) to output to the stdio stream
7833 @var{stream} the assembler definition of uninitialized global @var{decl} named
7834 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
7835 is the alignment specified as the number of bits.
7837 Try to use function @code{asm_output_aligned_bss} defined in file
7838 @file{varasm.c} when defining this macro. If unable, use the expression
7839 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7840 before and after that, output the additional assembler syntax for defining
7841 the name, and a newline.
7843 There are two ways of handling global BSS@. One is to define this macro.
7844 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7845 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7846 You do not need to do both.
7848 Some languages do not have @code{common} data, and require a
7849 non-common form of global BSS in order to handle uninitialized globals
7850 efficiently. C++ is one example of this. However, if the target does
7851 not support global BSS, the front end may choose to make globals
7852 common in order to save space in the object file.
7855 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7856 A C statement (sans semicolon) to output to the stdio stream
7857 @var{stream} the assembler definition of a local-common-label named
7858 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7859 is the size rounded up to whatever alignment the caller wants.
7861 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7862 output the name itself; before and after that, output the additional
7863 assembler syntax for defining the name, and a newline.
7865 This macro controls how the assembler definitions of uninitialized
7866 static variables are output.
7869 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7870 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7871 separate, explicit argument. If you define this macro, it is used in
7872 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7873 handling the required alignment of the variable. The alignment is specified
7874 as the number of bits.
7877 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7878 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7879 variable to be output, if there is one, or @code{NULL_TREE} if there
7880 is no corresponding variable. If you define this macro, GCC will use it
7881 in place of both @code{ASM_OUTPUT_DECL} and
7882 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7883 the variable's decl in order to chose what to output.
7887 @subsection Output and Generation of Labels
7889 @c prevent bad page break with this line
7890 This is about outputting labels.
7892 @findex assemble_name
7893 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7894 A C statement (sans semicolon) to output to the stdio stream
7895 @var{stream} the assembler definition of a label named @var{name}.
7896 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7897 output the name itself; before and after that, output the additional
7898 assembler syntax for defining the name, and a newline. A default
7899 definition of this macro is provided which is correct for most systems.
7902 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
7903 A C statement (sans semicolon) to output to the stdio stream
7904 @var{stream} the assembler definition of a label named @var{name} of
7906 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7907 output the name itself; before and after that, output the additional
7908 assembler syntax for defining the name, and a newline. A default
7909 definition of this macro is provided which is correct for most systems.
7911 If this macro is not defined, then the function name is defined in the
7912 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7915 @findex assemble_name_raw
7916 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7917 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7918 to refer to a compiler-generated label. The default definition uses
7919 @code{assemble_name_raw}, which is like @code{assemble_name} except
7920 that it is more efficient.
7924 A C string containing the appropriate assembler directive to specify the
7925 size of a symbol, without any arguments. On systems that use ELF, the
7926 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7927 systems, the default is not to define this macro.
7929 Define this macro only if it is correct to use the default definitions
7930 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7931 for your system. If you need your own custom definitions of those
7932 macros, or if you do not need explicit symbol sizes at all, do not
7936 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7937 A C statement (sans semicolon) to output to the stdio stream
7938 @var{stream} a directive telling the assembler that the size of the
7939 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7940 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7944 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7945 A C statement (sans semicolon) to output to the stdio stream
7946 @var{stream} a directive telling the assembler to calculate the size of
7947 the symbol @var{name} by subtracting its address from the current
7950 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7951 provided. The default assumes that the assembler recognizes a special
7952 @samp{.} symbol as referring to the current address, and can calculate
7953 the difference between this and another symbol. If your assembler does
7954 not recognize @samp{.} or cannot do calculations with it, you will need
7955 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7958 @defmac NO_DOLLAR_IN_LABEL
7959 Define this macro if the assembler does not accept the character
7960 @samp{$} in label names. By default constructors and destructors in
7961 G++ have @samp{$} in the identifiers. If this macro is defined,
7962 @samp{.} is used instead.
7965 @defmac NO_DOT_IN_LABEL
7966 Define this macro if the assembler does not accept the character
7967 @samp{.} in label names. By default constructors and destructors in G++
7968 have names that use @samp{.}. If this macro is defined, these names
7969 are rewritten to avoid @samp{.}.
7973 A C string containing the appropriate assembler directive to specify the
7974 type of a symbol, without any arguments. On systems that use ELF, the
7975 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7976 systems, the default is not to define this macro.
7978 Define this macro only if it is correct to use the default definition of
7979 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7980 custom definition of this macro, or if you do not need explicit symbol
7981 types at all, do not define this macro.
7984 @defmac TYPE_OPERAND_FMT
7985 A C string which specifies (using @code{printf} syntax) the format of
7986 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7987 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7988 the default is not to define this macro.
7990 Define this macro only if it is correct to use the default definition of
7991 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7992 custom definition of this macro, or if you do not need explicit symbol
7993 types at all, do not define this macro.
7996 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7997 A C statement (sans semicolon) to output to the stdio stream
7998 @var{stream} a directive telling the assembler that the type of the
7999 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
8000 that string is always either @samp{"function"} or @samp{"object"}, but
8001 you should not count on this.
8003 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
8004 definition of this macro is provided.
8007 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
8008 A C statement (sans semicolon) to output to the stdio stream
8009 @var{stream} any text necessary for declaring the name @var{name} of a
8010 function which is being defined. This macro is responsible for
8011 outputting the label definition (perhaps using
8012 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
8013 @code{FUNCTION_DECL} tree node representing the function.
8015 If this macro is not defined, then the function name is defined in the
8016 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
8018 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
8022 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
8023 A C statement (sans semicolon) to output to the stdio stream
8024 @var{stream} any text necessary for declaring the size of a function
8025 which is being defined. The argument @var{name} is the name of the
8026 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
8027 representing the function.
8029 If this macro is not defined, then the function size is not defined.
8031 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
8035 @defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
8036 A C statement (sans semicolon) to output to the stdio stream
8037 @var{stream} any text necessary for declaring the name @var{name} of a
8038 cold function partition which is being defined. This macro is responsible
8039 for outputting the label definition (perhaps using
8040 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
8041 @code{FUNCTION_DECL} tree node representing the function.
8043 If this macro is not defined, then the cold partition name is defined in the
8044 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8046 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
8050 @defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
8051 A C statement (sans semicolon) to output to the stdio stream
8052 @var{stream} any text necessary for declaring the size of a cold function
8053 partition which is being defined. The argument @var{name} is the name of the
8054 cold partition of the function. The argument @var{decl} is the
8055 @code{FUNCTION_DECL} tree node representing the function.
8057 If this macro is not defined, then the partition size is not defined.
8059 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
8063 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
8064 A C statement (sans semicolon) to output to the stdio stream
8065 @var{stream} any text necessary for declaring the name @var{name} of an
8066 initialized variable which is being defined. This macro must output the
8067 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
8068 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
8070 If this macro is not defined, then the variable name is defined in the
8071 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
8073 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
8074 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
8077 @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})
8078 A target hook to output to the stdio stream @var{file} any text necessary
8079 for declaring the name @var{name} of a constant which is being defined. This
8080 target hook is responsible for outputting the label definition (perhaps using
8081 @code{assemble_label}). The argument @var{exp} is the value of the constant,
8082 and @var{size} is the size of the constant in bytes. The @var{name}
8083 will be an internal label.
8085 The default version of this target hook, define the @var{name} in the
8086 usual manner as a label (by means of @code{assemble_label}).
8088 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in this target hook.
8091 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
8092 A C statement (sans semicolon) to output to the stdio stream
8093 @var{stream} any text necessary for claiming a register @var{regno}
8094 for a global variable @var{decl} with name @var{name}.
8096 If you don't define this macro, that is equivalent to defining it to do
8100 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
8101 A C statement (sans semicolon) to finish up declaring a variable name
8102 once the compiler has processed its initializer fully and thus has had a
8103 chance to determine the size of an array when controlled by an
8104 initializer. This is used on systems where it's necessary to declare
8105 something about the size of the object.
8107 If you don't define this macro, that is equivalent to defining it to do
8110 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
8111 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
8114 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
8115 This target hook is a function to output to the stdio stream
8116 @var{stream} some commands that will make the label @var{name} global;
8117 that is, available for reference from other files.
8119 The default implementation relies on a proper definition of
8120 @code{GLOBAL_ASM_OP}.
8123 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
8124 This target hook is a function to output to the stdio stream
8125 @var{stream} some commands that will make the name associated with @var{decl}
8126 global; that is, available for reference from other files.
8128 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
8131 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_UNDEFINED_DECL (FILE *@var{stream}, const char *@var{name}, const_tree @var{decl})
8132 This target hook is a function to output to the stdio stream
8133 @var{stream} some commands that will declare the name associated with
8134 @var{decl} which is not defined in the current translation unit. Most
8135 assemblers do not require anything to be output in this case.
8138 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
8139 A C statement (sans semicolon) to output to the stdio stream
8140 @var{stream} some commands that will make the label @var{name} weak;
8141 that is, available for reference from other files but only used if
8142 no other definition is available. Use the expression
8143 @code{assemble_name (@var{stream}, @var{name})} to output the name
8144 itself; before and after that, output the additional assembler syntax
8145 for making that name weak, and a newline.
8147 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
8148 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
8152 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
8153 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
8154 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
8155 or variable decl. If @var{value} is not @code{NULL}, this C statement
8156 should output to the stdio stream @var{stream} assembler code which
8157 defines (equates) the weak symbol @var{name} to have the value
8158 @var{value}. If @var{value} is @code{NULL}, it should output commands
8159 to make @var{name} weak.
8162 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
8163 Outputs a directive that enables @var{name} to be used to refer to
8164 symbol @var{value} with weak-symbol semantics. @code{decl} is the
8165 declaration of @code{name}.
8168 @defmac SUPPORTS_WEAK
8169 A preprocessor constant expression which evaluates to true if the target
8170 supports weak symbols.
8172 If you don't define this macro, @file{defaults.h} provides a default
8173 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
8174 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
8177 @defmac TARGET_SUPPORTS_WEAK
8178 A C expression which evaluates to true if the target supports weak symbols.
8180 If you don't define this macro, @file{defaults.h} provides a default
8181 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
8182 this macro if you want to control weak symbol support with a compiler
8183 flag such as @option{-melf}.
8186 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
8187 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
8188 public symbol such that extra copies in multiple translation units will
8189 be discarded by the linker. Define this macro if your object file
8190 format provides support for this concept, such as the @samp{COMDAT}
8191 section flags in the Microsoft Windows PE/COFF format, and this support
8192 requires changes to @var{decl}, such as putting it in a separate section.
8195 @defmac SUPPORTS_ONE_ONLY
8196 A C expression which evaluates to true if the target supports one-only
8199 If you don't define this macro, @file{varasm.c} provides a default
8200 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
8201 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
8202 you want to control one-only symbol support with a compiler flag, or if
8203 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
8204 be emitted as one-only.
8207 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, int @var{visibility})
8208 This target hook is a function to output to @var{asm_out_file} some
8209 commands that will make the symbol(s) associated with @var{decl} have
8210 hidden, protected or internal visibility as specified by @var{visibility}.
8213 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
8214 A C expression that evaluates to true if the target's linker expects
8215 that weak symbols do not appear in a static archive's table of contents.
8216 The default is @code{0}.
8218 Leaving weak symbols out of an archive's table of contents means that,
8219 if a symbol will only have a definition in one translation unit and
8220 will have undefined references from other translation units, that
8221 symbol should not be weak. Defining this macro to be nonzero will
8222 thus have the effect that certain symbols that would normally be weak
8223 (explicit template instantiations, and vtables for polymorphic classes
8224 with noninline key methods) will instead be nonweak.
8226 The C++ ABI requires this macro to be zero. Define this macro for
8227 targets where full C++ ABI compliance is impossible and where linker
8228 restrictions require weak symbols to be left out of a static archive's
8232 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
8233 A C statement (sans semicolon) to output to the stdio stream
8234 @var{stream} any text necessary for declaring the name of an external
8235 symbol named @var{name} which is referenced in this compilation but
8236 not defined. The value of @var{decl} is the tree node for the
8239 This macro need not be defined if it does not need to output anything.
8240 The GNU assembler and most Unix assemblers don't require anything.
8243 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
8244 This target hook is a function to output to @var{asm_out_file} an assembler
8245 pseudo-op to declare a library function name external. The name of the
8246 library function is given by @var{symref}, which is a @code{symbol_ref}.
8249 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (const char *@var{symbol})
8250 This target hook is a function to output to @var{asm_out_file} an assembler
8251 directive to annotate @var{symbol} as used. The Darwin target uses the
8252 .no_dead_code_strip directive.
8255 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
8256 A C statement (sans semicolon) to output to the stdio stream
8257 @var{stream} a reference in assembler syntax to a label named
8258 @var{name}. This should add @samp{_} to the front of the name, if that
8259 is customary on your operating system, as it is in most Berkeley Unix
8260 systems. This macro is used in @code{assemble_name}.
8263 @deftypefn {Target Hook} tree TARGET_MANGLE_ASSEMBLER_NAME (const char *@var{name})
8264 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.
8267 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
8268 A C statement (sans semicolon) to output a reference to
8269 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
8270 will be used to output the name of the symbol. This macro may be used
8271 to modify the way a symbol is referenced depending on information
8272 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
8275 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
8276 A C statement (sans semicolon) to output a reference to @var{buf}, the
8277 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
8278 @code{assemble_name} will be used to output the name of the symbol.
8279 This macro is not used by @code{output_asm_label}, or the @code{%l}
8280 specifier that calls it; the intention is that this macro should be set
8281 when it is necessary to output a label differently when its address is
8285 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
8286 A function to output to the stdio stream @var{stream} a label whose
8287 name is made from the string @var{prefix} and the number @var{labelno}.
8289 It is absolutely essential that these labels be distinct from the labels
8290 used for user-level functions and variables. Otherwise, certain programs
8291 will have name conflicts with internal labels.
8293 It is desirable to exclude internal labels from the symbol table of the
8294 object file. Most assemblers have a naming convention for labels that
8295 should be excluded; on many systems, the letter @samp{L} at the
8296 beginning of a label has this effect. You should find out what
8297 convention your system uses, and follow it.
8299 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
8302 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
8303 A C statement to output to the stdio stream @var{stream} a debug info
8304 label whose name is made from the string @var{prefix} and the number
8305 @var{num}. This is useful for VLIW targets, where debug info labels
8306 may need to be treated differently than branch target labels. On some
8307 systems, branch target labels must be at the beginning of instruction
8308 bundles, but debug info labels can occur in the middle of instruction
8311 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
8315 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
8316 A C statement to store into the string @var{string} a label whose name
8317 is made from the string @var{prefix} and the number @var{num}.
8319 This string, when output subsequently by @code{assemble_name}, should
8320 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
8321 with the same @var{prefix} and @var{num}.
8323 If the string begins with @samp{*}, then @code{assemble_name} will
8324 output the rest of the string unchanged. It is often convenient for
8325 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
8326 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
8327 to output the string, and may change it. (Of course,
8328 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
8329 you should know what it does on your machine.)
8332 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
8333 A C expression to assign to @var{outvar} (which is a variable of type
8334 @code{char *}) a newly allocated string made from the string
8335 @var{name} and the number @var{number}, with some suitable punctuation
8336 added. Use @code{alloca} to get space for the string.
8338 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
8339 produce an assembler label for an internal static variable whose name is
8340 @var{name}. Therefore, the string must be such as to result in valid
8341 assembler code. The argument @var{number} is different each time this
8342 macro is executed; it prevents conflicts between similarly-named
8343 internal static variables in different scopes.
8345 Ideally this string should not be a valid C identifier, to prevent any
8346 conflict with the user's own symbols. Most assemblers allow periods
8347 or percent signs in assembler symbols; putting at least one of these
8348 between the name and the number will suffice.
8350 If this macro is not defined, a default definition will be provided
8351 which is correct for most systems.
8354 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
8355 A C statement to output to the stdio stream @var{stream} assembler code
8356 which defines (equates) the symbol @var{name} to have the value @var{value}.
8359 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8360 correct for most systems.
8363 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
8364 A C statement to output to the stdio stream @var{stream} assembler code
8365 which defines (equates) the symbol whose tree node is @var{decl_of_name}
8366 to have the value of the tree node @var{decl_of_value}. This macro will
8367 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
8368 the tree nodes are available.
8371 If @code{SET_ASM_OP} is defined, a default definition is provided which is
8372 correct for most systems.
8375 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
8376 A C statement that evaluates to true if the assembler code which defines
8377 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
8378 of the tree node @var{decl_of_value} should be emitted near the end of the
8379 current compilation unit. The default is to not defer output of defines.
8380 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
8381 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
8384 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
8385 A C statement to output to the stdio stream @var{stream} assembler code
8386 which defines (equates) the weak symbol @var{name} to have the value
8387 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
8388 an undefined weak symbol.
8390 Define this macro if the target only supports weak aliases; define
8391 @code{ASM_OUTPUT_DEF} instead if possible.
8394 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
8395 Define this macro to override the default assembler names used for
8396 Objective-C methods.
8398 The default name is a unique method number followed by the name of the
8399 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
8400 the category is also included in the assembler name (e.g.@:
8403 These names are safe on most systems, but make debugging difficult since
8404 the method's selector is not present in the name. Therefore, particular
8405 systems define other ways of computing names.
8407 @var{buf} is an expression of type @code{char *} which gives you a
8408 buffer in which to store the name; its length is as long as
8409 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
8410 50 characters extra.
8412 The argument @var{is_inst} specifies whether the method is an instance
8413 method or a class method; @var{class_name} is the name of the class;
8414 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
8415 in a category); and @var{sel_name} is the name of the selector.
8417 On systems where the assembler can handle quoted names, you can use this
8418 macro to provide more human-readable names.
8421 @node Initialization
8422 @subsection How Initialization Functions Are Handled
8423 @cindex initialization routines
8424 @cindex termination routines
8425 @cindex constructors, output of
8426 @cindex destructors, output of
8428 The compiled code for certain languages includes @dfn{constructors}
8429 (also called @dfn{initialization routines})---functions to initialize
8430 data in the program when the program is started. These functions need
8431 to be called before the program is ``started''---that is to say, before
8432 @code{main} is called.
8434 Compiling some languages generates @dfn{destructors} (also called
8435 @dfn{termination routines}) that should be called when the program
8438 To make the initialization and termination functions work, the compiler
8439 must output something in the assembler code to cause those functions to
8440 be called at the appropriate time. When you port the compiler to a new
8441 system, you need to specify how to do this.
8443 There are two major ways that GCC currently supports the execution of
8444 initialization and termination functions. Each way has two variants.
8445 Much of the structure is common to all four variations.
8447 @findex __CTOR_LIST__
8448 @findex __DTOR_LIST__
8449 The linker must build two lists of these functions---a list of
8450 initialization functions, called @code{__CTOR_LIST__}, and a list of
8451 termination functions, called @code{__DTOR_LIST__}.
8453 Each list always begins with an ignored function pointer (which may hold
8454 0, @minus{}1, or a count of the function pointers after it, depending on
8455 the environment). This is followed by a series of zero or more function
8456 pointers to constructors (or destructors), followed by a function
8457 pointer containing zero.
8459 Depending on the operating system and its executable file format, either
8460 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8461 time and exit time. Constructors are called in reverse order of the
8462 list; destructors in forward order.
8464 The best way to handle static constructors works only for object file
8465 formats which provide arbitrarily-named sections. A section is set
8466 aside for a list of constructors, and another for a list of destructors.
8467 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8468 object file that defines an initialization function also puts a word in
8469 the constructor section to point to that function. The linker
8470 accumulates all these words into one contiguous @samp{.ctors} section.
8471 Termination functions are handled similarly.
8473 This method will be chosen as the default by @file{target-def.h} if
8474 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8475 support arbitrary sections, but does support special designated
8476 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8477 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8479 When arbitrary sections are available, there are two variants, depending
8480 upon how the code in @file{crtstuff.c} is called. On systems that
8481 support a @dfn{.init} section which is executed at program startup,
8482 parts of @file{crtstuff.c} are compiled into that section. The
8483 program is linked by the @command{gcc} driver like this:
8486 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8489 The prologue of a function (@code{__init}) appears in the @code{.init}
8490 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8491 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8492 files are provided by the operating system or by the GNU C library, but
8493 are provided by GCC for a few targets.
8495 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8496 compiled from @file{crtstuff.c}. They contain, among other things, code
8497 fragments within the @code{.init} and @code{.fini} sections that branch
8498 to routines in the @code{.text} section. The linker will pull all parts
8499 of a section together, which results in a complete @code{__init} function
8500 that invokes the routines we need at startup.
8502 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8505 If no init section is available, when GCC compiles any function called
8506 @code{main} (or more accurately, any function designated as a program
8507 entry point by the language front end calling @code{expand_main_function}),
8508 it inserts a procedure call to @code{__main} as the first executable code
8509 after the function prologue. The @code{__main} function is defined
8510 in @file{libgcc2.c} and runs the global constructors.
8512 In file formats that don't support arbitrary sections, there are again
8513 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8514 and an `a.out' format must be used. In this case,
8515 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8516 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8517 and with the address of the void function containing the initialization
8518 code as its value. The GNU linker recognizes this as a request to add
8519 the value to a @dfn{set}; the values are accumulated, and are eventually
8520 placed in the executable as a vector in the format described above, with
8521 a leading (ignored) count and a trailing zero element.
8522 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8523 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8524 the compilation of @code{main} to call @code{__main} as above, starting
8525 the initialization process.
8527 The last variant uses neither arbitrary sections nor the GNU linker.
8528 This is preferable when you want to do dynamic linking and when using
8529 file formats which the GNU linker does not support, such as `ECOFF'@. In
8530 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8531 termination functions are recognized simply by their names. This requires
8532 an extra program in the linkage step, called @command{collect2}. This program
8533 pretends to be the linker, for use with GCC; it does its job by running
8534 the ordinary linker, but also arranges to include the vectors of
8535 initialization and termination functions. These functions are called
8536 via @code{__main} as described above. In order to use this method,
8537 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8540 The following section describes the specific macros that control and
8541 customize the handling of initialization and termination functions.
8544 @node Macros for Initialization
8545 @subsection Macros Controlling Initialization Routines
8547 Here are the macros that control how the compiler handles initialization
8548 and termination functions:
8550 @defmac INIT_SECTION_ASM_OP
8551 If defined, a C string constant, including spacing, for the assembler
8552 operation to identify the following data as initialization code. If not
8553 defined, GCC will assume such a section does not exist. When you are
8554 using special sections for initialization and termination functions, this
8555 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8556 run the initialization functions.
8559 @defmac HAS_INIT_SECTION
8560 If defined, @code{main} will not call @code{__main} as described above.
8561 This macro should be defined for systems that control start-up code
8562 on a symbol-by-symbol basis, such as OSF/1, and should not
8563 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8566 @defmac LD_INIT_SWITCH
8567 If defined, a C string constant for a switch that tells the linker that
8568 the following symbol is an initialization routine.
8571 @defmac LD_FINI_SWITCH
8572 If defined, a C string constant for a switch that tells the linker that
8573 the following symbol is a finalization routine.
8576 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8577 If defined, a C statement that will write a function that can be
8578 automatically called when a shared library is loaded. The function
8579 should call @var{func}, which takes no arguments. If not defined, and
8580 the object format requires an explicit initialization function, then a
8581 function called @code{_GLOBAL__DI} will be generated.
8583 This function and the following one are used by collect2 when linking a
8584 shared library that needs constructors or destructors, or has DWARF2
8585 exception tables embedded in the code.
8588 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8589 If defined, a C statement that will write a function that can be
8590 automatically called when a shared library is unloaded. The function
8591 should call @var{func}, which takes no arguments. If not defined, and
8592 the object format requires an explicit finalization function, then a
8593 function called @code{_GLOBAL__DD} will be generated.
8596 @defmac INVOKE__main
8597 If defined, @code{main} will call @code{__main} despite the presence of
8598 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8599 where the init section is not actually run automatically, but is still
8600 useful for collecting the lists of constructors and destructors.
8603 @defmac SUPPORTS_INIT_PRIORITY
8604 If nonzero, the C++ @code{init_priority} attribute is supported and the
8605 compiler should emit instructions to control the order of initialization
8606 of objects. If zero, the compiler will issue an error message upon
8607 encountering an @code{init_priority} attribute.
8610 @deftypevr {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8611 This value is true if the target supports some ``native'' method of
8612 collecting constructors and destructors to be run at startup and exit.
8613 It is false if we must use @command{collect2}.
8616 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8617 If defined, a function that outputs assembler code to arrange to call
8618 the function referenced by @var{symbol} at initialization time.
8620 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8621 no arguments and with no return value. If the target supports initialization
8622 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8623 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8625 If this macro is not defined by the target, a suitable default will
8626 be chosen if (1) the target supports arbitrary section names, (2) the
8627 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8631 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8632 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8633 functions rather than initialization functions.
8636 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8637 generated for the generated object file will have static linkage.
8639 If your system uses @command{collect2} as the means of processing
8640 constructors, then that program normally uses @command{nm} to scan
8641 an object file for constructor functions to be called.
8643 On certain kinds of systems, you can define this macro to make
8644 @command{collect2} work faster (and, in some cases, make it work at all):
8646 @defmac OBJECT_FORMAT_COFF
8647 Define this macro if the system uses COFF (Common Object File Format)
8648 object files, so that @command{collect2} can assume this format and scan
8649 object files directly for dynamic constructor/destructor functions.
8651 This macro is effective only in a native compiler; @command{collect2} as
8652 part of a cross compiler always uses @command{nm} for the target machine.
8655 @defmac REAL_NM_FILE_NAME
8656 Define this macro as a C string constant containing the file name to use
8657 to execute @command{nm}. The default is to search the path normally for
8662 @command{collect2} calls @command{nm} to scan object files for static
8663 constructors and destructors and LTO info. By default, @option{-n} is
8664 passed. Define @code{NM_FLAGS} to a C string constant if other options
8665 are needed to get the same output format as GNU @command{nm -n}
8669 If your system supports shared libraries and has a program to list the
8670 dynamic dependencies of a given library or executable, you can define
8671 these macros to enable support for running initialization and
8672 termination functions in shared libraries:
8675 Define this macro to a C string constant containing the name of the program
8676 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
8679 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8680 Define this macro to be C code that extracts filenames from the output
8681 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8682 of type @code{char *} that points to the beginning of a line of output
8683 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8684 code must advance @var{ptr} to the beginning of the filename on that
8685 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8688 @defmac SHLIB_SUFFIX
8689 Define this macro to a C string constant containing the default shared
8690 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8691 strips version information after this suffix when generating global
8692 constructor and destructor names. This define is only needed on targets
8693 that use @command{collect2} to process constructors and destructors.
8696 @node Instruction Output
8697 @subsection Output of Assembler Instructions
8699 @c prevent bad page break with this line
8700 This describes assembler instruction output.
8702 @defmac REGISTER_NAMES
8703 A C initializer containing the assembler's names for the machine
8704 registers, each one as a C string constant. This is what translates
8705 register numbers in the compiler into assembler language.
8708 @defmac ADDITIONAL_REGISTER_NAMES
8709 If defined, a C initializer for an array of structures containing a name
8710 and a register number. This macro defines additional names for hard
8711 registers, thus allowing the @code{asm} option in declarations to refer
8712 to registers using alternate names.
8715 @defmac OVERLAPPING_REGISTER_NAMES
8716 If defined, a C initializer for an array of structures containing a
8717 name, a register number and a count of the number of consecutive
8718 machine registers the name overlaps. This macro defines additional
8719 names for hard registers, thus allowing the @code{asm} option in
8720 declarations to refer to registers using alternate names. Unlike
8721 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
8722 register name implies multiple underlying registers.
8724 This macro should be used when it is important that a clobber in an
8725 @code{asm} statement clobbers all the underlying values implied by the
8726 register name. For example, on ARM, clobbering the double-precision
8727 VFP register ``d0'' implies clobbering both single-precision registers
8731 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8732 Define this macro if you are using an unusual assembler that
8733 requires different names for the machine instructions.
8735 The definition is a C statement or statements which output an
8736 assembler instruction opcode to the stdio stream @var{stream}. The
8737 macro-operand @var{ptr} is a variable of type @code{char *} which
8738 points to the opcode name in its ``internal'' form---the form that is
8739 written in the machine description. The definition should output the
8740 opcode name to @var{stream}, performing any translation you desire, and
8741 increment the variable @var{ptr} to point at the end of the opcode
8742 so that it will not be output twice.
8744 In fact, your macro definition may process less than the entire opcode
8745 name, or more than the opcode name; but if you want to process text
8746 that includes @samp{%}-sequences to substitute operands, you must take
8747 care of the substitution yourself. Just be sure to increment
8748 @var{ptr} over whatever text should not be output normally.
8750 @findex recog_data.operand
8751 If you need to look at the operand values, they can be found as the
8752 elements of @code{recog_data.operand}.
8754 If the macro definition does nothing, the instruction is output
8758 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8759 If defined, a C statement to be executed just prior to the output of
8760 assembler code for @var{insn}, to modify the extracted operands so
8761 they will be output differently.
8763 Here the argument @var{opvec} is the vector containing the operands
8764 extracted from @var{insn}, and @var{noperands} is the number of
8765 elements of the vector which contain meaningful data for this insn.
8766 The contents of this vector are what will be used to convert the insn
8767 template into assembler code, so you can change the assembler output
8768 by changing the contents of the vector.
8770 This macro is useful when various assembler syntaxes share a single
8771 file of instruction patterns; by defining this macro differently, you
8772 can cause a large class of instructions to be output differently (such
8773 as with rearranged operands). Naturally, variations in assembler
8774 syntax affecting individual insn patterns ought to be handled by
8775 writing conditional output routines in those patterns.
8777 If this macro is not defined, it is equivalent to a null statement.
8780 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{file}, rtx_insn *@var{insn}, rtx *@var{opvec}, int @var{noperands})
8781 If defined, this target hook is a function which is executed just after the
8782 output of assembler code for @var{insn}, to change the mode of the assembler
8785 Here the argument @var{opvec} is the vector containing the operands
8786 extracted from @var{insn}, and @var{noperands} is the number of
8787 elements of the vector which contain meaningful data for this insn.
8788 The contents of this vector are what was used to convert the insn
8789 template into assembler code, so you can change the assembler mode
8790 by checking the contents of the vector.
8793 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8794 A C compound statement to output to stdio stream @var{stream} the
8795 assembler syntax for an instruction operand @var{x}. @var{x} is an
8798 @var{code} is a value that can be used to specify one of several ways
8799 of printing the operand. It is used when identical operands must be
8800 printed differently depending on the context. @var{code} comes from
8801 the @samp{%} specification that was used to request printing of the
8802 operand. If the specification was just @samp{%@var{digit}} then
8803 @var{code} is 0; if the specification was @samp{%@var{ltr}
8804 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8807 If @var{x} is a register, this macro should print the register's name.
8808 The names can be found in an array @code{reg_names} whose type is
8809 @code{char *[]}. @code{reg_names} is initialized from
8810 @code{REGISTER_NAMES}.
8812 When the machine description has a specification @samp{%@var{punct}}
8813 (a @samp{%} followed by a punctuation character), this macro is called
8814 with a null pointer for @var{x} and the punctuation character for
8818 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8819 A C expression which evaluates to true if @var{code} is a valid
8820 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8821 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8822 punctuation characters (except for the standard one, @samp{%}) are used
8826 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8827 A C compound statement to output to stdio stream @var{stream} the
8828 assembler syntax for an instruction operand that is a memory reference
8829 whose address is @var{x}. @var{x} is an RTL expression.
8831 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8832 On some machines, the syntax for a symbolic address depends on the
8833 section that the address refers to. On these machines, define the hook
8834 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8835 @code{symbol_ref}, and then check for it here. @xref{Assembler
8839 @findex dbr_sequence_length
8840 @defmac DBR_OUTPUT_SEQEND (@var{file})
8841 A C statement, to be executed after all slot-filler instructions have
8842 been output. If necessary, call @code{dbr_sequence_length} to
8843 determine the number of slots filled in a sequence (zero if not
8844 currently outputting a sequence), to decide how many no-ops to output,
8847 Don't define this macro if it has nothing to do, but it is helpful in
8848 reading assembly output if the extent of the delay sequence is made
8849 explicit (e.g.@: with white space).
8852 @findex final_sequence
8853 Note that output routines for instructions with delay slots must be
8854 prepared to deal with not being output as part of a sequence
8855 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8856 found.) The variable @code{final_sequence} is null when not
8857 processing a sequence, otherwise it contains the @code{sequence} rtx
8861 @defmac REGISTER_PREFIX
8862 @defmacx LOCAL_LABEL_PREFIX
8863 @defmacx USER_LABEL_PREFIX
8864 @defmacx IMMEDIATE_PREFIX
8865 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8866 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8867 @file{final.c}). These are useful when a single @file{md} file must
8868 support multiple assembler formats. In that case, the various @file{tm.h}
8869 files can define these macros differently.
8872 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8873 If defined this macro should expand to a series of @code{case}
8874 statements which will be parsed inside the @code{switch} statement of
8875 the @code{asm_fprintf} function. This allows targets to define extra
8876 printf formats which may useful when generating their assembler
8877 statements. Note that uppercase letters are reserved for future
8878 generic extensions to asm_fprintf, and so are not available to target
8879 specific code. The output file is given by the parameter @var{file}.
8880 The varargs input pointer is @var{argptr} and the rest of the format
8881 string, starting the character after the one that is being switched
8882 upon, is pointed to by @var{format}.
8885 @defmac ASSEMBLER_DIALECT
8886 If your target supports multiple dialects of assembler language (such as
8887 different opcodes), define this macro as a C expression that gives the
8888 numeric index of the assembler language dialect to use, with zero as the
8891 If this macro is defined, you may use constructs of the form
8893 @samp{@{option0|option1|option2@dots{}@}}
8896 in the output templates of patterns (@pxref{Output Template}) or in the
8897 first argument of @code{asm_fprintf}. This construct outputs
8898 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8899 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8900 within these strings retain their usual meaning. If there are fewer
8901 alternatives within the braces than the value of
8902 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
8903 to print curly braces or @samp{|} character in assembler output directly,
8904 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
8906 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8907 @samp{@}} do not have any special meaning when used in templates or
8908 operands to @code{asm_fprintf}.
8910 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8911 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8912 the variations in assembler language syntax with that mechanism. Define
8913 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8914 if the syntax variant are larger and involve such things as different
8915 opcodes or operand order.
8918 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8919 A C expression to output to @var{stream} some assembler code
8920 which will push hard register number @var{regno} onto the stack.
8921 The code need not be optimal, since this macro is used only when
8925 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8926 A C expression to output to @var{stream} some assembler code
8927 which will pop hard register number @var{regno} off of the stack.
8928 The code need not be optimal, since this macro is used only when
8932 @node Dispatch Tables
8933 @subsection Output of Dispatch Tables
8935 @c prevent bad page break with this line
8936 This concerns dispatch tables.
8938 @cindex dispatch table
8939 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8940 A C statement to output to the stdio stream @var{stream} an assembler
8941 pseudo-instruction to generate a difference between two labels.
8942 @var{value} and @var{rel} are the numbers of two internal labels. The
8943 definitions of these labels are output using
8944 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8945 way here. For example,
8948 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8949 @var{value}, @var{rel})
8952 You must provide this macro on machines where the addresses in a
8953 dispatch table are relative to the table's own address. If defined, GCC
8954 will also use this macro on all machines when producing PIC@.
8955 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8956 mode and flags can be read.
8959 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8960 This macro should be provided on machines where the addresses
8961 in a dispatch table are absolute.
8963 The definition should be a C statement to output to the stdio stream
8964 @var{stream} an assembler pseudo-instruction to generate a reference to
8965 a label. @var{value} is the number of an internal label whose
8966 definition is output using @code{(*targetm.asm_out.internal_label)}.
8970 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8974 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8975 Define this if the label before a jump-table needs to be output
8976 specially. The first three arguments are the same as for
8977 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8978 jump-table which follows (a @code{jump_table_data} containing an
8979 @code{addr_vec} or @code{addr_diff_vec}).
8981 This feature is used on system V to output a @code{swbeg} statement
8984 If this macro is not defined, these labels are output with
8985 @code{(*targetm.asm_out.internal_label)}.
8988 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8989 Define this if something special must be output at the end of a
8990 jump-table. The definition should be a C statement to be executed
8991 after the assembler code for the table is written. It should write
8992 the appropriate code to stdio stream @var{stream}. The argument
8993 @var{table} is the jump-table insn, and @var{num} is the label-number
8994 of the preceding label.
8996 If this macro is not defined, nothing special is output at the end of
9000 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (FILE *@var{stream}, tree @var{decl}, int @var{for_eh}, int @var{empty})
9001 This target hook emits a label at the beginning of each FDE@. It
9002 should be defined on targets where FDEs need special labels, and it
9003 should write the appropriate label, for the FDE associated with the
9004 function declaration @var{decl}, to the stdio stream @var{stream}.
9005 The third argument, @var{for_eh}, is a boolean: true if this is for an
9006 exception table. The fourth argument, @var{empty}, is a boolean:
9007 true if this is a placeholder label for an omitted FDE@.
9009 The default is that FDEs are not given nonlocal labels.
9012 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (FILE *@var{stream})
9013 This target hook emits a label at the beginning of the exception table.
9014 It should be defined on targets where it is desirable for the table
9015 to be broken up according to function.
9017 The default is that no label is emitted.
9020 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_PERSONALITY (rtx @var{personality})
9021 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.
9024 @deftypefn {Target Hook} void TARGET_ASM_UNWIND_EMIT (FILE *@var{stream}, rtx_insn *@var{insn})
9025 This target hook emits assembly directives required to unwind the
9026 given instruction. This is only used when @code{TARGET_EXCEPT_UNWIND_INFO}
9027 returns @code{UI_TARGET}.
9030 @deftypevr {Target Hook} bool TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
9031 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.
9034 @node Exception Region Output
9035 @subsection Assembler Commands for Exception Regions
9037 @c prevent bad page break with this line
9039 This describes commands marking the start and the end of an exception
9042 @defmac EH_FRAME_SECTION_NAME
9043 If defined, a C string constant for the name of the section containing
9044 exception handling frame unwind information. If not defined, GCC will
9045 provide a default definition if the target supports named sections.
9046 @file{crtstuff.c} uses this macro to switch to the appropriate section.
9048 You should define this symbol if your target supports DWARF 2 frame
9049 unwind information and the default definition does not work.
9052 @defmac EH_FRAME_THROUGH_COLLECT2
9053 If defined, DWARF 2 frame unwind information will identified by
9054 specially named labels. The collect2 process will locate these
9055 labels and generate code to register the frames.
9057 This might be necessary, for instance, if the system linker will not
9058 place the eh_frames in-between the sentinals from @file{crtstuff.c},
9059 or if the system linker does garbage collection and sections cannot
9060 be marked as not to be collected.
9063 @defmac EH_TABLES_CAN_BE_READ_ONLY
9064 Define this macro to 1 if your target is such that no frame unwind
9065 information encoding used with non-PIC code will ever require a
9066 runtime relocation, but the linker may not support merging read-only
9067 and read-write sections into a single read-write section.
9070 @defmac MASK_RETURN_ADDR
9071 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
9072 that it does not contain any extraneous set bits in it.
9075 @defmac DWARF2_UNWIND_INFO
9076 Define this macro to 0 if your target supports DWARF 2 frame unwind
9077 information, but it does not yet work with exception handling.
9078 Otherwise, if your target supports this information (if it defines
9079 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
9080 GCC will provide a default definition of 1.
9083 @deftypefn {Common Target Hook} {enum unwind_info_type} TARGET_EXCEPT_UNWIND_INFO (struct gcc_options *@var{opts})
9084 This hook defines the mechanism that will be used for exception handling
9085 by the target. If the target has ABI specified unwind tables, the hook
9086 should return @code{UI_TARGET}. If the target is to use the
9087 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
9088 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
9089 information, the hook should return @code{UI_DWARF2}.
9091 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
9092 This may end up simplifying other parts of target-specific code. The
9093 default implementation of this hook never returns @code{UI_NONE}.
9095 Note that the value returned by this hook should be constant. It should
9096 not depend on anything except the command-line switches described by
9097 @var{opts}. In particular, the
9098 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
9099 macros and builtin functions related to exception handling are set up
9100 depending on this setting.
9102 The default implementation of the hook first honors the
9103 @option{--enable-sjlj-exceptions} configure option, then
9104 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
9105 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
9106 must define this hook so that @var{opts} is used correctly.
9109 @deftypevr {Common Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
9110 This variable should be set to @code{true} if the target ABI requires unwinding
9111 tables even when exceptions are not used. It must not be modified by
9112 command-line option processing.
9115 @defmac DONT_USE_BUILTIN_SETJMP
9116 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
9117 should use the @code{setjmp}/@code{longjmp} functions from the C library
9118 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
9121 @defmac JMP_BUF_SIZE
9122 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
9123 defined. Define this macro if the default size of @code{jmp_buf} buffer
9124 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
9125 is not large enough, or if it is much too large.
9126 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
9129 @defmac DWARF_CIE_DATA_ALIGNMENT
9130 This macro need only be defined if the target might save registers in the
9131 function prologue at an offset to the stack pointer that is not aligned to
9132 @code{UNITS_PER_WORD}. The definition should be the negative minimum
9133 alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive
9134 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
9135 the target supports DWARF 2 frame unwind information.
9138 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
9139 Contains the value true if the target should add a zero word onto the
9140 end of a Dwarf-2 frame info section when used for exception handling.
9141 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
9145 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
9146 Given a register, this hook should return a parallel of registers to
9147 represent where to find the register pieces. Define this hook if the
9148 register and its mode are represented in Dwarf in non-contiguous
9149 locations, or if the register should be represented in more than one
9150 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
9151 If not defined, the default is to return @code{NULL_RTX}.
9154 @deftypefn {Target Hook} machine_mode TARGET_DWARF_FRAME_REG_MODE (int @var{regno})
9155 Given a register, this hook should return the mode which the
9156 corresponding Dwarf frame register should have. This is normally
9157 used to return a smaller mode than the raw mode to prevent call
9158 clobbered parts of a register altering the frame register size
9161 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
9162 If some registers are represented in Dwarf-2 unwind information in
9163 multiple pieces, define this hook to fill in information about the
9164 sizes of those pieces in the table used by the unwinder at runtime.
9165 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
9166 filling in a single size corresponding to each hard register;
9167 @var{address} is the address of the table.
9170 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
9171 This hook is used to output a reference from a frame unwinding table to
9172 the type_info object identified by @var{sym}. It should return @code{true}
9173 if the reference was output. Returning @code{false} will cause the
9174 reference to be output using the normal Dwarf2 routines.
9177 @deftypevr {Target Hook} bool TARGET_ARM_EABI_UNWINDER
9178 This flag should be set to @code{true} on targets that use an ARM EABI
9179 based unwinding library, and @code{false} on other targets. This effects
9180 the format of unwinding tables, and how the unwinder in entered after
9181 running a cleanup. The default is @code{false}.
9184 @node Alignment Output
9185 @subsection Assembler Commands for Alignment
9187 @c prevent bad page break with this line
9188 This describes commands for alignment.
9190 @defmac JUMP_ALIGN (@var{label})
9191 The alignment (log base 2) to put in front of @var{label}, which is
9192 a common destination of jumps and has no fallthru incoming edge.
9194 This macro need not be defined if you don't want any special alignment
9195 to be done at such a time. Most machine descriptions do not currently
9198 Unless it's necessary to inspect the @var{label} parameter, it is better
9199 to set the variable @var{align_jumps} in the target's
9200 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9201 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
9204 @deftypefn {Target Hook} int TARGET_ASM_JUMP_ALIGN_MAX_SKIP (rtx_insn *@var{label})
9205 The maximum number of bytes to skip before @var{label} when applying
9206 @code{JUMP_ALIGN}. This works only if
9207 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9210 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
9211 The alignment (log base 2) to put in front of @var{label}, which follows
9214 This macro need not be defined if you don't want any special alignment
9215 to be done at such a time. Most machine descriptions do not currently
9219 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP (rtx_insn *@var{label})
9220 The maximum number of bytes to skip before @var{label} when applying
9221 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
9222 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
9225 @defmac LOOP_ALIGN (@var{label})
9226 The alignment (log base 2) to put in front of @var{label} that heads
9227 a frequently executed basic block (usually the header of a loop).
9229 This macro need not be defined if you don't want any special alignment
9230 to be done at such a time. Most machine descriptions do not currently
9233 Unless it's necessary to inspect the @var{label} parameter, it is better
9234 to set the variable @code{align_loops} in the target's
9235 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9236 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
9239 @deftypefn {Target Hook} int TARGET_ASM_LOOP_ALIGN_MAX_SKIP (rtx_insn *@var{label})
9240 The maximum number of bytes to skip when applying @code{LOOP_ALIGN} to
9241 @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is
9245 @defmac LABEL_ALIGN (@var{label})
9246 The alignment (log base 2) to put in front of @var{label}.
9247 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
9248 the maximum of the specified values is used.
9250 Unless it's necessary to inspect the @var{label} parameter, it is better
9251 to set the variable @code{align_labels} in the target's
9252 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
9253 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
9256 @deftypefn {Target Hook} int TARGET_ASM_LABEL_ALIGN_MAX_SKIP (rtx_insn *@var{label})
9257 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}
9258 to @var{label}. This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN}
9262 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
9263 A C statement to output to the stdio stream @var{stream} an assembler
9264 instruction to advance the location counter by @var{nbytes} bytes.
9265 Those bytes should be zero when loaded. @var{nbytes} will be a C
9266 expression of type @code{unsigned HOST_WIDE_INT}.
9269 @defmac ASM_NO_SKIP_IN_TEXT
9270 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
9271 text section because it fails to put zeros in the bytes that are skipped.
9272 This is true on many Unix systems, where the pseudo--op to skip bytes
9273 produces no-op instructions rather than zeros when used in the text
9277 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
9278 A C statement to output to the stdio stream @var{stream} an assembler
9279 command to advance the location counter to a multiple of 2 to the
9280 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
9283 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
9284 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
9285 for padding, if necessary.
9288 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
9289 A C statement to output to the stdio stream @var{stream} an assembler
9290 command to advance the location counter to a multiple of 2 to the
9291 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
9292 satisfy the alignment request. @var{power} and @var{max_skip} will be
9293 a C expression of type @code{int}.
9297 @node Debugging Info
9298 @section Controlling Debugging Information Format
9300 @c prevent bad page break with this line
9301 This describes how to specify debugging information.
9304 * All Debuggers:: Macros that affect all debugging formats uniformly.
9305 * DBX Options:: Macros enabling specific options in DBX format.
9306 * DBX Hooks:: Hook macros for varying DBX format.
9307 * File Names and DBX:: Macros controlling output of file names in DBX format.
9308 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
9309 * VMS Debug:: Macros for VMS debug format.
9313 @subsection Macros Affecting All Debugging Formats
9315 @c prevent bad page break with this line
9316 These macros affect all debugging formats.
9318 @defmac DBX_REGISTER_NUMBER (@var{regno})
9319 A C expression that returns the DBX register number for the compiler
9320 register number @var{regno}. In the default macro provided, the value
9321 of this expression will be @var{regno} itself. But sometimes there are
9322 some registers that the compiler knows about and DBX does not, or vice
9323 versa. In such cases, some register may need to have one number in the
9324 compiler and another for DBX@.
9326 If two registers have consecutive numbers inside GCC, and they can be
9327 used as a pair to hold a multiword value, then they @emph{must} have
9328 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
9329 Otherwise, debuggers will be unable to access such a pair, because they
9330 expect register pairs to be consecutive in their own numbering scheme.
9332 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
9333 does not preserve register pairs, then what you must do instead is
9334 redefine the actual register numbering scheme.
9337 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
9338 A C expression that returns the integer offset value for an automatic
9339 variable having address @var{x} (an RTL expression). The default
9340 computation assumes that @var{x} is based on the frame-pointer and
9341 gives the offset from the frame-pointer. This is required for targets
9342 that produce debugging output for DBX or COFF-style debugging output
9343 for SDB and allow the frame-pointer to be eliminated when the
9344 @option{-g} options is used.
9347 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
9348 A C expression that returns the integer offset value for an argument
9349 having address @var{x} (an RTL expression). The nominal offset is
9353 @defmac PREFERRED_DEBUGGING_TYPE
9354 A C expression that returns the type of debugging output GCC should
9355 produce when the user specifies just @option{-g}. Define
9356 this if you have arranged for GCC to support more than one format of
9357 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
9358 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
9359 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
9361 When the user specifies @option{-ggdb}, GCC normally also uses the
9362 value of this macro to select the debugging output format, but with two
9363 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
9364 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
9365 defined, GCC uses @code{DBX_DEBUG}.
9367 The value of this macro only affects the default debugging output; the
9368 user can always get a specific type of output by using @option{-gstabs},
9369 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
9373 @subsection Specific Options for DBX Output
9375 @c prevent bad page break with this line
9376 These are specific options for DBX output.
9378 @defmac DBX_DEBUGGING_INFO
9379 Define this macro if GCC should produce debugging output for DBX
9380 in response to the @option{-g} option.
9383 @defmac XCOFF_DEBUGGING_INFO
9384 Define this macro if GCC should produce XCOFF format debugging output
9385 in response to the @option{-g} option. This is a variant of DBX format.
9388 @defmac DEFAULT_GDB_EXTENSIONS
9389 Define this macro to control whether GCC should by default generate
9390 GDB's extended version of DBX debugging information (assuming DBX-format
9391 debugging information is enabled at all). If you don't define the
9392 macro, the default is 1: always generate the extended information
9393 if there is any occasion to.
9396 @defmac DEBUG_SYMS_TEXT
9397 Define this macro if all @code{.stabs} commands should be output while
9398 in the text section.
9401 @defmac ASM_STABS_OP
9402 A C string constant, including spacing, naming the assembler pseudo op to
9403 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
9404 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
9405 applies only to DBX debugging information format.
9408 @defmac ASM_STABD_OP
9409 A C string constant, including spacing, naming the assembler pseudo op to
9410 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
9411 value is the current location. If you don't define this macro,
9412 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
9416 @defmac ASM_STABN_OP
9417 A C string constant, including spacing, naming the assembler pseudo op to
9418 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
9419 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
9420 macro applies only to DBX debugging information format.
9423 @defmac DBX_NO_XREFS
9424 Define this macro if DBX on your system does not support the construct
9425 @samp{xs@var{tagname}}. On some systems, this construct is used to
9426 describe a forward reference to a structure named @var{tagname}.
9427 On other systems, this construct is not supported at all.
9430 @defmac DBX_CONTIN_LENGTH
9431 A symbol name in DBX-format debugging information is normally
9432 continued (split into two separate @code{.stabs} directives) when it
9433 exceeds a certain length (by default, 80 characters). On some
9434 operating systems, DBX requires this splitting; on others, splitting
9435 must not be done. You can inhibit splitting by defining this macro
9436 with the value zero. You can override the default splitting-length by
9437 defining this macro as an expression for the length you desire.
9440 @defmac DBX_CONTIN_CHAR
9441 Normally continuation is indicated by adding a @samp{\} character to
9442 the end of a @code{.stabs} string when a continuation follows. To use
9443 a different character instead, define this macro as a character
9444 constant for the character you want to use. Do not define this macro
9445 if backslash is correct for your system.
9448 @defmac DBX_STATIC_STAB_DATA_SECTION
9449 Define this macro if it is necessary to go to the data section before
9450 outputting the @samp{.stabs} pseudo-op for a non-global static
9454 @defmac DBX_TYPE_DECL_STABS_CODE
9455 The value to use in the ``code'' field of the @code{.stabs} directive
9456 for a typedef. The default is @code{N_LSYM}.
9459 @defmac DBX_STATIC_CONST_VAR_CODE
9460 The value to use in the ``code'' field of the @code{.stabs} directive
9461 for a static variable located in the text section. DBX format does not
9462 provide any ``right'' way to do this. The default is @code{N_FUN}.
9465 @defmac DBX_REGPARM_STABS_CODE
9466 The value to use in the ``code'' field of the @code{.stabs} directive
9467 for a parameter passed in registers. DBX format does not provide any
9468 ``right'' way to do this. The default is @code{N_RSYM}.
9471 @defmac DBX_REGPARM_STABS_LETTER
9472 The letter to use in DBX symbol data to identify a symbol as a parameter
9473 passed in registers. DBX format does not customarily provide any way to
9474 do this. The default is @code{'P'}.
9477 @defmac DBX_FUNCTION_FIRST
9478 Define this macro if the DBX information for a function and its
9479 arguments should precede the assembler code for the function. Normally,
9480 in DBX format, the debugging information entirely follows the assembler
9484 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
9485 Define this macro, with value 1, if the value of a symbol describing
9486 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
9487 relative to the start of the enclosing function. Normally, GCC uses
9488 an absolute address.
9491 @defmac DBX_LINES_FUNCTION_RELATIVE
9492 Define this macro, with value 1, if the value of a symbol indicating
9493 the current line number (@code{N_SLINE}) should be relative to the
9494 start of the enclosing function. Normally, GCC uses an absolute address.
9497 @defmac DBX_USE_BINCL
9498 Define this macro if GCC should generate @code{N_BINCL} and
9499 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9500 macro also directs GCC to output a type number as a pair of a file
9501 number and a type number within the file. Normally, GCC does not
9502 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9503 number for a type number.
9507 @subsection Open-Ended Hooks for DBX Format
9509 @c prevent bad page break with this line
9510 These are hooks for DBX format.
9512 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9513 A C statement to output DBX debugging information before code for line
9514 number @var{line} of the current source file to the stdio stream
9515 @var{stream}. @var{counter} is the number of time the macro was
9516 invoked, including the current invocation; it is intended to generate
9517 unique labels in the assembly output.
9519 This macro should not be defined if the default output is correct, or
9520 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9523 @defmac NO_DBX_FUNCTION_END
9524 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9525 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9526 On those machines, define this macro to turn this feature off without
9527 disturbing the rest of the gdb extensions.
9530 @defmac NO_DBX_BNSYM_ENSYM
9531 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9532 extension construct. On those machines, define this macro to turn this
9533 feature off without disturbing the rest of the gdb extensions.
9536 @node File Names and DBX
9537 @subsection File Names in DBX Format
9539 @c prevent bad page break with this line
9540 This describes file names in DBX format.
9542 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9543 A C statement to output DBX debugging information to the stdio stream
9544 @var{stream}, which indicates that file @var{name} is the main source
9545 file---the file specified as the input file for compilation.
9546 This macro is called only once, at the beginning of compilation.
9548 This macro need not be defined if the standard form of output
9549 for DBX debugging information is appropriate.
9551 It may be necessary to refer to a label equal to the beginning of the
9552 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9553 to do so. If you do this, you must also set the variable
9554 @var{used_ltext_label_name} to @code{true}.
9557 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9558 Define this macro, with value 1, if GCC should not emit an indication
9559 of the current directory for compilation and current source language at
9560 the beginning of the file.
9563 @defmac NO_DBX_GCC_MARKER
9564 Define this macro, with value 1, if GCC should not emit an indication
9565 that this object file was compiled by GCC@. The default is to emit
9566 an @code{N_OPT} stab at the beginning of every source file, with
9567 @samp{gcc2_compiled.} for the string and value 0.
9570 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9571 A C statement to output DBX debugging information at the end of
9572 compilation of the main source file @var{name}. Output should be
9573 written to the stdio stream @var{stream}.
9575 If you don't define this macro, nothing special is output at the end
9576 of compilation, which is correct for most machines.
9579 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9580 Define this macro @emph{instead of} defining
9581 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9582 the end of compilation is an @code{N_SO} stab with an empty string,
9583 whose value is the highest absolute text address in the file.
9588 @subsection Macros for SDB and DWARF Output
9590 @c prevent bad page break with this line
9591 Here are macros for SDB and DWARF output.
9593 @defmac SDB_DEBUGGING_INFO
9594 Define this macro to 1 if GCC should produce COFF-style debugging output
9595 for SDB in response to the @option{-g} option.
9598 @defmac DWARF2_DEBUGGING_INFO
9599 Define this macro if GCC should produce dwarf version 2 format
9600 debugging output in response to the @option{-g} option.
9602 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (const_tree @var{function})
9603 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9604 be emitted for each function. Instead of an integer return the enum
9605 value for the @code{DW_CC_} tag.
9608 To support optional call frame debugging information, you must also
9609 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9610 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9611 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9612 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9615 @defmac DWARF2_FRAME_INFO
9616 Define this macro to a nonzero value if GCC should always output
9617 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
9618 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
9619 exceptions are enabled, GCC will output this information not matter
9620 how you define @code{DWARF2_FRAME_INFO}.
9623 @deftypefn {Target Hook} {enum unwind_info_type} TARGET_DEBUG_UNWIND_INFO (void)
9624 This hook defines the mechanism that will be used for describing frame
9625 unwind information to the debugger. Normally the hook will return
9626 @code{UI_DWARF2} if DWARF 2 debug information is enabled, and
9627 return @code{UI_NONE} otherwise.
9629 A target may return @code{UI_DWARF2} even when DWARF 2 debug information
9630 is disabled in order to always output DWARF 2 frame information.
9632 A target may return @code{UI_TARGET} if it has ABI specified unwind tables.
9633 This will suppress generation of the normal debug frame unwind information.
9636 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9637 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9638 line debug info sections. This will result in much more compact line number
9639 tables, and hence is desirable if it works.
9642 @deftypevr {Target Hook} bool TARGET_WANT_DEBUG_PUB_SECTIONS
9643 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.
9646 @deftypevr {Target Hook} bool TARGET_FORCE_AT_COMP_DIR
9647 True if the @code{DW_AT_comp_dir} attribute should be emitted for each compilation unit. This attribute is required for the darwin linker to emit debug information.
9650 @deftypevr {Target Hook} bool TARGET_DELAY_SCHED2
9651 True if sched2 is not to be run at its normal place.
9652 This usually means it will be run as part of machine-specific reorg.
9655 @deftypevr {Target Hook} bool TARGET_DELAY_VARTRACK
9656 True if vartrack is not to be run at its normal place.
9657 This usually means it will be run as part of machine-specific reorg.
9660 @deftypevr {Target Hook} bool TARGET_NO_REGISTER_ALLOCATION
9661 True if register allocation and the passes
9662 following it should not be run. Usually true only for virtual assembler
9666 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9667 A C statement to issue assembly directives that create a difference
9668 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9671 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9672 A C statement to issue assembly directives that create a difference
9673 between the two given labels in system defined units, e.g. instruction
9674 slots on IA64 VMS, using an integer of the given size.
9677 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9678 A C statement to issue assembly directives that create a
9679 section-relative reference to the given @var{label}, using an integer of the
9680 given @var{size}. The label is known to be defined in the given @var{section}.
9683 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9684 A C statement to issue assembly directives that create a self-relative
9685 reference to the given @var{label}, using an integer of the given @var{size}.
9688 @defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label})
9689 A C statement to issue assembly directives that create a reference to the
9690 given @var{label} relative to the dbase, using an integer of the given @var{size}.
9693 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9694 A C statement to issue assembly directives that create a reference to
9695 the DWARF table identifier @var{label} from the current section. This
9696 is used on some systems to avoid garbage collecting a DWARF table which
9697 is referenced by a function.
9700 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{file}, int @var{size}, rtx @var{x})
9701 If defined, this target hook is a function which outputs a DTP-relative
9702 reference to the given TLS symbol of the specified size.
9705 @defmac PUT_SDB_@dots{}
9706 Define these macros to override the assembler syntax for the special
9707 SDB assembler directives. See @file{sdbout.c} for a list of these
9708 macros and their arguments. If the standard syntax is used, you need
9709 not define them yourself.
9713 Some assemblers do not support a semicolon as a delimiter, even between
9714 SDB assembler directives. In that case, define this macro to be the
9715 delimiter to use (usually @samp{\n}). It is not necessary to define
9716 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9720 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9721 Define this macro to allow references to unknown structure,
9722 union, or enumeration tags to be emitted. Standard COFF does not
9723 allow handling of unknown references, MIPS ECOFF has support for
9727 @defmac SDB_ALLOW_FORWARD_REFERENCES
9728 Define this macro to allow references to structure, union, or
9729 enumeration tags that have not yet been seen to be handled. Some
9730 assemblers choke if forward tags are used, while some require it.
9733 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9734 A C statement to output SDB debugging information before code for line
9735 number @var{line} of the current source file to the stdio stream
9736 @var{stream}. The default is to emit an @code{.ln} directive.
9741 @subsection Macros for VMS Debug Format
9743 @c prevent bad page break with this line
9744 Here are macros for VMS debug format.
9746 @defmac VMS_DEBUGGING_INFO
9747 Define this macro if GCC should produce debugging output for VMS
9748 in response to the @option{-g} option. The default behavior for VMS
9749 is to generate minimal debug info for a traceback in the absence of
9750 @option{-g} unless explicitly overridden with @option{-g0}. This
9751 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
9752 @code{TARGET_OPTION_OVERRIDE}.
9755 @node Floating Point
9756 @section Cross Compilation and Floating Point
9757 @cindex cross compilation and floating point
9758 @cindex floating point and cross compilation
9760 While all modern machines use twos-complement representation for integers,
9761 there are a variety of representations for floating point numbers. This
9762 means that in a cross-compiler the representation of floating point numbers
9763 in the compiled program may be different from that used in the machine
9764 doing the compilation.
9766 Because different representation systems may offer different amounts of
9767 range and precision, all floating point constants must be represented in
9768 the target machine's format. Therefore, the cross compiler cannot
9769 safely use the host machine's floating point arithmetic; it must emulate
9770 the target's arithmetic. To ensure consistency, GCC always uses
9771 emulation to work with floating point values, even when the host and
9772 target floating point formats are identical.
9774 The following macros are provided by @file{real.h} for the compiler to
9775 use. All parts of the compiler which generate or optimize
9776 floating-point calculations must use these macros. They may evaluate
9777 their operands more than once, so operands must not have side effects.
9779 @defmac REAL_VALUE_TYPE
9780 The C data type to be used to hold a floating point value in the target
9781 machine's format. Typically this is a @code{struct} containing an
9782 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9786 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9787 Truncates @var{x} to a signed integer, rounding toward zero.
9790 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9791 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9792 @var{x} is negative, returns zero.
9795 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode})
9796 Converts @var{string} into a floating point number in the target machine's
9797 representation for mode @var{mode}. This routine can handle both
9798 decimal and hexadecimal floating point constants, using the syntax
9799 defined by the C language for both.
9802 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9803 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9806 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9807 Determines whether @var{x} represents infinity (positive or negative).
9810 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9811 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9814 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9815 Returns the negative of the floating point value @var{x}.
9818 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9819 Returns the absolute value of @var{x}.
9822 @node Mode Switching
9823 @section Mode Switching Instructions
9824 @cindex mode switching
9825 The following macros control mode switching optimizations:
9827 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9828 Define this macro if the port needs extra instructions inserted for mode
9829 switching in an optimizing compilation.
9831 For an example, the SH4 can perform both single and double precision
9832 floating point operations, but to perform a single precision operation,
9833 the FPSCR PR bit has to be cleared, while for a double precision
9834 operation, this bit has to be set. Changing the PR bit requires a general
9835 purpose register as a scratch register, hence these FPSCR sets have to
9836 be inserted before reload, i.e.@: you can't put this into instruction emitting
9837 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9839 You can have multiple entities that are mode-switched, and select at run time
9840 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9841 return nonzero for any @var{entity} that needs mode-switching.
9842 If you define this macro, you also have to define
9843 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
9844 @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
9845 @code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
9849 @defmac NUM_MODES_FOR_MODE_SWITCHING
9850 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9851 initializer for an array of integers. Each initializer element
9852 N refers to an entity that needs mode switching, and specifies the number
9853 of different modes that might need to be set for this entity.
9854 The position of the initializer in the initializer---starting counting at
9855 zero---determines the integer that is used to refer to the mode-switched
9857 In macros that take mode arguments / yield a mode result, modes are
9858 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9859 switch is needed / supplied.
9862 @deftypefn {Target Hook} void TARGET_MODE_EMIT (int @var{entity}, int @var{mode}, int @var{prev_mode}, HARD_REG_SET @var{regs_live})
9863 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.
9866 @deftypefn {Target Hook} int TARGET_MODE_NEEDED (int @var{entity}, rtx_insn *@var{insn})
9867 @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}.
9870 @deftypefn {Target Hook} int TARGET_MODE_AFTER (int @var{entity}, int @var{mode}, rtx_insn *@var{insn})
9871 @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).
9874 @deftypefn {Target Hook} int TARGET_MODE_ENTRY (int @var{entity})
9875 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.
9878 @deftypefn {Target Hook} int TARGET_MODE_EXIT (int @var{entity})
9879 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.
9882 @deftypefn {Target Hook} int TARGET_MODE_PRIORITY (int @var{entity}, int @var{n})
9883 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}.
9886 @node Target Attributes
9887 @section Defining target-specific uses of @code{__attribute__}
9888 @cindex target attributes
9889 @cindex machine attributes
9890 @cindex attributes, target-specific
9892 Target-specific attributes may be defined for functions, data and types.
9893 These are described using the following target hooks; they also need to
9894 be documented in @file{extend.texi}.
9896 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9897 If defined, this target hook points to an array of @samp{struct
9898 attribute_spec} (defined in @file{tree-core.h}) specifying the machine
9899 specific attributes for this target and some of the restrictions on the
9900 entities to which these attributes are applied and the arguments they
9904 @deftypefn {Target Hook} bool TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P (const_tree @var{name})
9905 If defined, this target hook is a function which returns true if the
9906 machine-specific attribute named @var{name} expects an identifier
9907 given as its first argument to be passed on as a plain identifier, not
9908 subjected to name lookup. If this is not defined, the default is
9909 false for all machine-specific attributes.
9912 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (const_tree @var{type1}, const_tree @var{type2})
9913 If defined, this target hook is a function which returns zero if the attributes on
9914 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9915 and two if they are nearly compatible (which causes a warning to be
9916 generated). If this is not defined, machine-specific attributes are
9917 supposed always to be compatible.
9920 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9921 If defined, this target hook is a function which assigns default attributes to
9922 the newly defined @var{type}.
9925 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9926 Define this target hook if the merging of type attributes needs special
9927 handling. If defined, the result is a list of the combined
9928 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9929 that @code{comptypes} has already been called and returned 1. This
9930 function may call @code{merge_attributes} to handle machine-independent
9934 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9935 Define this target hook if the merging of decl attributes needs special
9936 handling. If defined, the result is a list of the combined
9937 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9938 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9939 when this is needed are when one attribute overrides another, or when an
9940 attribute is nullified by a subsequent definition. This function may
9941 call @code{merge_attributes} to handle machine-independent merging.
9943 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9944 If the only target-specific handling you require is @samp{dllimport}
9945 for Microsoft Windows targets, you should define the macro
9946 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9947 will then define a function called
9948 @code{merge_dllimport_decl_attributes} which can then be defined as
9949 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9950 add @code{handle_dll_attribute} in the attribute table for your port
9951 to perform initial processing of the @samp{dllimport} and
9952 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9953 @file{i386/i386.c}, for example.
9956 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (const_tree @var{decl})
9957 @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}.
9960 @defmac TARGET_DECLSPEC
9961 Define this macro to a nonzero value if you want to treat
9962 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9963 default, this behavior is enabled only for targets that define
9964 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9965 of @code{__declspec} is via a built-in macro, but you should not rely
9966 on this implementation detail.
9969 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9970 Define this target hook if you want to be able to add attributes to a decl
9971 when it is being created. This is normally useful for back ends which
9972 wish to implement a pragma by using the attributes which correspond to
9973 the pragma's effect. The @var{node} argument is the decl which is being
9974 created. The @var{attr_ptr} argument is a pointer to the attribute list
9975 for this decl. The list itself should not be modified, since it may be
9976 shared with other decls, but attributes may be chained on the head of
9977 the list and @code{*@var{attr_ptr}} modified to point to the new
9978 attributes, or a copy of the list may be made if further changes are
9982 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (const_tree @var{fndecl})
9984 This target hook returns @code{true} if it is OK to inline @var{fndecl}
9985 into the current function, despite its having target-specific
9986 attributes, @code{false} otherwise. By default, if a function has a
9987 target specific attribute attached to it, it will not be inlined.
9990 @deftypefn {Target Hook} bool TARGET_OPTION_VALID_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9991 This hook is called to parse @code{attribute(target("..."))}, which
9992 allows setting target-specific options on individual functions.
9993 These function-specific options may differ
9994 from the options specified on the command line. The hook should return
9995 @code{true} if the options are valid.
9997 The hook should set the @code{DECL_FUNCTION_SPECIFIC_TARGET} field in
9998 the function declaration to hold a pointer to a target-specific
9999 @code{struct cl_target_option} structure.
10002 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr}, struct gcc_options *@var{opts})
10003 This hook is called to save any additional target-specific information
10004 in the @code{struct cl_target_option} structure for function-specific
10005 options from the @code{struct gcc_options} structure.
10006 @xref{Option file format}.
10009 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct gcc_options *@var{opts}, struct cl_target_option *@var{ptr})
10010 This hook is called to restore any additional target-specific
10011 information in the @code{struct cl_target_option} structure for
10012 function-specific options to the @code{struct gcc_options} structure.
10015 @deftypefn {Target Hook} void TARGET_OPTION_POST_STREAM_IN (struct cl_target_option *@var{ptr})
10016 This hook is called to update target-specific information in the
10017 @code{struct cl_target_option} structure after it is streamed in from
10021 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (FILE *@var{file}, int @var{indent}, struct cl_target_option *@var{ptr})
10022 This hook is called to print any additional target-specific
10023 information in the @code{struct cl_target_option} structure for
10024 function-specific options.
10027 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (tree @var{args}, tree @var{pop_target})
10028 This target hook parses the options for @code{#pragma GCC target}, which
10029 sets the target-specific options for functions that occur later in the
10030 input stream. The options accepted should be the same as those handled by the
10031 @code{TARGET_OPTION_VALID_ATTRIBUTE_P} hook.
10034 @deftypefn {Target Hook} void TARGET_OPTION_OVERRIDE (void)
10035 Sometimes certain combinations of command options do not make sense on
10036 a particular target machine. You can override the hook
10037 @code{TARGET_OPTION_OVERRIDE} to take account of this. This hooks is called
10038 once just after all the command options have been parsed.
10040 Don't use this hook to turn on various extra optimizations for
10041 @option{-O}. That is what @code{TARGET_OPTION_OPTIMIZATION} is for.
10043 If you need to do something whenever the optimization level is
10044 changed via the optimize attribute or pragma, see
10045 @code{TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE}
10048 @deftypefn {Target Hook} bool TARGET_OPTION_FUNCTION_VERSIONS (tree @var{decl1}, tree @var{decl2})
10049 This target hook returns @code{true} if @var{DECL1} and @var{DECL2} are
10050 versions of the same function. @var{DECL1} and @var{DECL2} are function
10051 versions if and only if they have the same function signature and
10052 different target specific attributes, that is, they are compiled for
10053 different target machines.
10056 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
10057 This target hook returns @code{false} if the @var{caller} function
10058 cannot inline @var{callee}, based on target specific information. By
10059 default, inlining is not allowed if the callee function has function
10060 specific target options and the caller does not use the same options.
10063 @deftypefn {Target Hook} void TARGET_RELAYOUT_FUNCTION (tree @var{fndecl})
10064 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.
10068 @section Emulating TLS
10069 @cindex Emulated TLS
10071 For targets whose psABI does not provide Thread Local Storage via
10072 specific relocations and instruction sequences, an emulation layer is
10073 used. A set of target hooks allows this emulation layer to be
10074 configured for the requirements of a particular target. For instance
10075 the psABI may in fact specify TLS support in terms of an emulation
10078 The emulation layer works by creating a control object for every TLS
10079 object. To access the TLS object, a lookup function is provided
10080 which, when given the address of the control object, will return the
10081 address of the current thread's instance of the TLS object.
10083 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
10084 Contains the name of the helper function that uses a TLS control
10085 object to locate a TLS instance. The default causes libgcc's
10086 emulated TLS helper function to be used.
10089 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
10090 Contains the name of the helper function that should be used at
10091 program startup to register TLS objects that are implicitly
10092 initialized to zero. If this is @code{NULL}, all TLS objects will
10093 have explicit initializers. The default causes libgcc's emulated TLS
10094 registration function to be used.
10097 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
10098 Contains the name of the section in which TLS control variables should
10099 be placed. The default of @code{NULL} allows these to be placed in
10103 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
10104 Contains the name of the section in which TLS initializers should be
10105 placed. The default of @code{NULL} allows these to be placed in any
10109 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
10110 Contains the prefix to be prepended to TLS control variable names.
10111 The default of @code{NULL} uses a target-specific prefix.
10114 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
10115 Contains the prefix to be prepended to TLS initializer objects. The
10116 default of @code{NULL} uses a target-specific prefix.
10119 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
10120 Specifies a function that generates the FIELD_DECLs for a TLS control
10121 object type. @var{type} is the RECORD_TYPE the fields are for and
10122 @var{name} should be filled with the structure tag, if the default of
10123 @code{__emutls_object} is unsuitable. The default creates a type suitable
10124 for libgcc's emulated TLS function.
10127 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
10128 Specifies a function that generates the CONSTRUCTOR to initialize a
10129 TLS control object. @var{var} is the TLS control object, @var{decl}
10130 is the TLS object and @var{tmpl_addr} is the address of the
10131 initializer. The default initializes libgcc's emulated TLS control object.
10134 @deftypevr {Target Hook} bool TARGET_EMUTLS_VAR_ALIGN_FIXED
10135 Specifies whether the alignment of TLS control variable objects is
10136 fixed and should not be increased as some backends may do to optimize
10137 single objects. The default is false.
10140 @deftypevr {Target Hook} bool TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
10141 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
10142 may be used to describe emulated TLS control objects.
10145 @node MIPS Coprocessors
10146 @section Defining coprocessor specifics for MIPS targets.
10147 @cindex MIPS coprocessor-definition macros
10149 The MIPS specification allows MIPS implementations to have as many as 4
10150 coprocessors, each with as many as 32 private registers. GCC supports
10151 accessing these registers and transferring values between the registers
10152 and memory using asm-ized variables. For example:
10155 register unsigned int cp0count asm ("c0r1");
10161 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
10162 names may be added as described below, or the default names may be
10163 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
10165 Coprocessor registers are assumed to be epilogue-used; sets to them will
10166 be preserved even if it does not appear that the register is used again
10167 later in the function.
10169 Another note: according to the MIPS spec, coprocessor 1 (if present) is
10170 the FPU@. One accesses COP1 registers through standard mips
10171 floating-point support; they are not included in this mechanism.
10174 @section Parameters for Precompiled Header Validity Checking
10175 @cindex parameters, precompiled headers
10177 @deftypefn {Target Hook} {void *} TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
10178 This hook returns a pointer to the data needed by
10179 @code{TARGET_PCH_VALID_P} and sets
10180 @samp{*@var{sz}} to the size of the data in bytes.
10183 @deftypefn {Target Hook} {const char *} TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
10184 This hook checks whether the options used to create a PCH file are
10185 compatible with the current settings. It returns @code{NULL}
10186 if so and a suitable error message if not. Error messages will
10187 be presented to the user and must be localized using @samp{_(@var{msg})}.
10189 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
10190 when the PCH file was created and @var{sz} is the size of that data in bytes.
10191 It's safe to assume that the data was created by the same version of the
10192 compiler, so no format checking is needed.
10194 The default definition of @code{default_pch_valid_p} should be
10195 suitable for most targets.
10198 @deftypefn {Target Hook} {const char *} TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
10199 If this hook is nonnull, the default implementation of
10200 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
10201 of @code{target_flags}. @var{pch_flags} specifies the value that
10202 @code{target_flags} had when the PCH file was created. The return
10203 value is the same as for @code{TARGET_PCH_VALID_P}.
10206 @deftypefn {Target Hook} void TARGET_PREPARE_PCH_SAVE (void)
10207 Called before writing out a PCH file. If the target has some
10208 garbage-collected data that needs to be in a particular state on PCH loads,
10209 it can use this hook to enforce that state. Very few targets need
10210 to do anything here.
10214 @section C++ ABI parameters
10215 @cindex parameters, c++ abi
10217 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
10218 Define this hook to override the integer type used for guard variables.
10219 These are used to implement one-time construction of static objects. The
10220 default is long_long_integer_type_node.
10223 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
10224 This hook determines how guard variables are used. It should return
10225 @code{false} (the default) if the first byte should be used. A return value of
10226 @code{true} indicates that only the least significant bit should be used.
10229 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
10230 This hook returns the size of the cookie to use when allocating an array
10231 whose elements have the indicated @var{type}. Assumes that it is already
10232 known that a cookie is needed. The default is
10233 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
10234 IA64/Generic C++ ABI@.
10237 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
10238 This hook should return @code{true} if the element size should be stored in
10239 array cookies. The default is to return @code{false}.
10242 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
10243 If defined by a backend this hook allows the decision made to export
10244 class @var{type} to be overruled. Upon entry @var{import_export}
10245 will contain 1 if the class is going to be exported, @minus{}1 if it is going
10246 to be imported and 0 otherwise. This function should return the
10247 modified value and perform any other actions necessary to support the
10248 backend's targeted operating system.
10251 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
10252 This hook should return @code{true} if constructors and destructors return
10253 the address of the object created/destroyed. The default is to return
10257 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
10258 This hook returns true if the key method for a class (i.e., the method
10259 which, if defined in the current translation unit, causes the virtual
10260 table to be emitted) may be an inline function. Under the standard
10261 Itanium C++ ABI the key method may be an inline function so long as
10262 the function is not declared inline in the class definition. Under
10263 some variants of the ABI, an inline function can never be the key
10264 method. The default is to return @code{true}.
10267 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
10268 @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}.
10271 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
10272 This hook returns true (the default) if virtual tables and other
10273 similar implicit class data objects are always COMDAT if they have
10274 external linkage. If this hook returns false, then class data for
10275 classes whose virtual table will be emitted in only one translation
10276 unit will not be COMDAT.
10279 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
10280 This hook returns true (the default) if the RTTI information for
10281 the basic types which is defined in the C++ runtime should always
10282 be COMDAT, false if it should not be COMDAT.
10285 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
10286 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
10287 should be used to register static destructors when @option{-fuse-cxa-atexit}
10288 is in effect. The default is to return false to use @code{__cxa_atexit}.
10291 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
10292 This hook returns true if the target @code{atexit} function can be used
10293 in the same manner as @code{__cxa_atexit} to register C++ static
10294 destructors. This requires that @code{atexit}-registered functions in
10295 shared libraries are run in the correct order when the libraries are
10296 unloaded. The default is to return false.
10299 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
10300 @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).
10303 @deftypefn {Target Hook} tree TARGET_CXX_DECL_MANGLING_CONTEXT (const_tree @var{decl})
10304 Return target-specific mangling context of @var{decl} or @code{NULL_TREE}.
10307 @node Named Address Spaces
10308 @section Adding support for named address spaces
10309 @cindex named address spaces
10311 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
10312 standards committee, @cite{Programming Languages - C - Extensions to
10313 support embedded processors}, specifies a syntax for embedded
10314 processors to specify alternate address spaces. You can configure a
10315 GCC port to support section 5.1 of the draft report to add support for
10316 address spaces other than the default address space. These address
10317 spaces are new keywords that are similar to the @code{volatile} and
10318 @code{const} type attributes.
10320 Pointers to named address spaces can have a different size than
10321 pointers to the generic address space.
10323 For example, the SPU port uses the @code{__ea} address space to refer
10324 to memory in the host processor, rather than memory local to the SPU
10325 processor. Access to memory in the @code{__ea} address space involves
10326 issuing DMA operations to move data between the host processor and the
10327 local processor memory address space. Pointers in the @code{__ea}
10328 address space are either 32 bits or 64 bits based on the
10329 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
10332 Internally, address spaces are represented as a small integer in the
10333 range 0 to 15 with address space 0 being reserved for the generic
10336 To register a named address space qualifier keyword with the C front end,
10337 the target may call the @code{c_register_addr_space} routine. For example,
10338 the SPU port uses the following to declare @code{__ea} as the keyword for
10339 named address space #1:
10341 #define ADDR_SPACE_EA 1
10342 c_register_addr_space ("__ea", ADDR_SPACE_EA);
10345 @deftypefn {Target Hook} machine_mode TARGET_ADDR_SPACE_POINTER_MODE (addr_space_t @var{address_space})
10346 Define this to return the machine mode to use for pointers to
10347 @var{address_space} if the target supports named address spaces.
10348 The default version of this hook returns @code{ptr_mode}.
10351 @deftypefn {Target Hook} machine_mode TARGET_ADDR_SPACE_ADDRESS_MODE (addr_space_t @var{address_space})
10352 Define this to return the machine mode to use for addresses in
10353 @var{address_space} if the target supports named address spaces.
10354 The default version of this hook returns @code{Pmode}.
10357 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_VALID_POINTER_MODE (machine_mode @var{mode}, addr_space_t @var{as})
10358 Define this to return nonzero if the port can handle pointers
10359 with machine mode @var{mode} to address space @var{as}. This target
10360 hook is the same as the @code{TARGET_VALID_POINTER_MODE} target hook,
10361 except that it includes explicit named address space support. The default
10362 version of this hook returns true for the modes returned by either the
10363 @code{TARGET_ADDR_SPACE_POINTER_MODE} or @code{TARGET_ADDR_SPACE_ADDRESS_MODE}
10364 target hooks for the given address space.
10367 @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})
10368 Define this to return true if @var{exp} is a valid address for mode
10369 @var{mode} in the named address space @var{as}. The @var{strict}
10370 parameter says whether strict addressing is in effect after reload has
10371 finished. This target hook is the same as the
10372 @code{TARGET_LEGITIMATE_ADDRESS_P} target hook, except that it includes
10373 explicit named address space support.
10376 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, machine_mode @var{mode}, addr_space_t @var{as})
10377 Define this to modify an invalid address @var{x} to be a valid address
10378 with mode @var{mode} in the named address space @var{as}. This target
10379 hook is the same as the @code{TARGET_LEGITIMIZE_ADDRESS} target hook,
10380 except that it includes explicit named address space support.
10383 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_SUBSET_P (addr_space_t @var{subset}, addr_space_t @var{superset})
10384 Define this to return whether the @var{subset} named address space is
10385 contained within the @var{superset} named address space. Pointers to
10386 a named address space that is a subset of another named address space
10387 will be converted automatically without a cast if used together in
10388 arithmetic operations. Pointers to a superset address space can be
10389 converted to pointers to a subset address space via explicit casts.
10392 @deftypefn {Target Hook} bool TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID (addr_space_t @var{as})
10393 Define this to modify the default handling of address 0 for the
10394 address space. Return true if 0 should be considered a valid address.
10397 @deftypefn {Target Hook} rtx TARGET_ADDR_SPACE_CONVERT (rtx @var{op}, tree @var{from_type}, tree @var{to_type})
10398 Define this to convert the pointer expression represented by the RTL
10399 @var{op} with type @var{from_type} that points to a named address
10400 space to a new pointer expression with type @var{to_type} that points
10401 to a different named address space. When this hook it called, it is
10402 guaranteed that one of the two address spaces is a subset of the other,
10403 as determined by the @code{TARGET_ADDR_SPACE_SUBSET_P} target hook.
10406 @deftypefn {Target Hook} int TARGET_ADDR_SPACE_DEBUG (addr_space_t @var{as})
10407 Define this to define how the address space is encoded in dwarf.
10408 The result is the value to be used with @code{DW_AT_address_class}.
10412 @section Miscellaneous Parameters
10413 @cindex parameters, miscellaneous
10415 @c prevent bad page break with this line
10416 Here are several miscellaneous parameters.
10418 @defmac HAS_LONG_COND_BRANCH
10419 Define this boolean macro to indicate whether or not your architecture
10420 has conditional branches that can span all of memory. It is used in
10421 conjunction with an optimization that partitions hot and cold basic
10422 blocks into separate sections of the executable. If this macro is
10423 set to false, gcc will convert any conditional branches that attempt
10424 to cross between sections into unconditional branches or indirect jumps.
10427 @defmac HAS_LONG_UNCOND_BRANCH
10428 Define this boolean macro to indicate whether or not your architecture
10429 has unconditional branches that can span all of memory. It is used in
10430 conjunction with an optimization that partitions hot and cold basic
10431 blocks into separate sections of the executable. If this macro is
10432 set to false, gcc will convert any unconditional branches that attempt
10433 to cross between sections into indirect jumps.
10436 @defmac CASE_VECTOR_MODE
10437 An alias for a machine mode name. This is the machine mode that
10438 elements of a jump-table should have.
10441 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
10442 Optional: return the preferred mode for an @code{addr_diff_vec}
10443 when the minimum and maximum offset are known. If you define this,
10444 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
10445 To make this work, you also have to define @code{INSN_ALIGN} and
10446 make the alignment for @code{addr_diff_vec} explicit.
10447 The @var{body} argument is provided so that the offset_unsigned and scale
10448 flags can be updated.
10451 @defmac CASE_VECTOR_PC_RELATIVE
10452 Define this macro to be a C expression to indicate when jump-tables
10453 should contain relative addresses. You need not define this macro if
10454 jump-tables never contain relative addresses, or jump-tables should
10455 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
10459 @deftypefn {Target Hook} {unsigned int} TARGET_CASE_VALUES_THRESHOLD (void)
10460 This function return the smallest number of different values for which it
10461 is best to use a jump-table instead of a tree of conditional branches.
10462 The default is four for machines with a @code{casesi} instruction and
10463 five otherwise. This is best for most machines.
10466 @defmac WORD_REGISTER_OPERATIONS
10467 Define this macro to 1 if operations between registers with integral mode
10468 smaller than a word are always performed on the entire register.
10469 Most RISC machines have this property and most CISC machines do not.
10472 @defmac LOAD_EXTEND_OP (@var{mem_mode})
10473 Define this macro to be a C expression indicating when insns that read
10474 memory in @var{mem_mode}, an integral mode narrower than a word, set the
10475 bits outside of @var{mem_mode} to be either the sign-extension or the
10476 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
10477 of @var{mem_mode} for which the
10478 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
10479 @code{UNKNOWN} for other modes.
10481 This macro is not called with @var{mem_mode} non-integral or with a width
10482 greater than or equal to @code{BITS_PER_WORD}, so you may return any
10483 value in this case. Do not define this macro if it would always return
10484 @code{UNKNOWN}. On machines where this macro is defined, you will normally
10485 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
10487 You may return a non-@code{UNKNOWN} value even if for some hard registers
10488 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
10489 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
10490 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
10491 integral mode larger than this but not larger than @code{word_mode}.
10493 You must return @code{UNKNOWN} if for some hard registers that allow this
10494 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
10495 @code{word_mode}, but that they can change to another integral mode that
10496 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
10499 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
10500 Define this macro to 1 if loading short immediate values into registers sign
10504 @deftypefn {Target Hook} {unsigned int} TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (machine_mode @var{mode})
10505 When @option{-ffast-math} is in effect, GCC tries to optimize
10506 divisions by the same divisor, by turning them into multiplications by
10507 the reciprocal. This target hook specifies the minimum number of divisions
10508 that should be there for GCC to perform the optimization for a variable
10509 of mode @var{mode}. The default implementation returns 3 if the machine
10510 has an instruction for the division, and 2 if it does not.
10514 The maximum number of bytes that a single instruction can move quickly
10515 between memory and registers or between two memory locations.
10518 @defmac MAX_MOVE_MAX
10519 The maximum number of bytes that a single instruction can move quickly
10520 between memory and registers or between two memory locations. If this
10521 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
10522 constant value that is the largest value that @code{MOVE_MAX} can have
10526 @defmac SHIFT_COUNT_TRUNCATED
10527 A C expression that is nonzero if on this machine the number of bits
10528 actually used for the count of a shift operation is equal to the number
10529 of bits needed to represent the size of the object being shifted. When
10530 this macro is nonzero, the compiler will assume that it is safe to omit
10531 a sign-extend, zero-extend, and certain bitwise `and' instructions that
10532 truncates the count of a shift operation. On machines that have
10533 instructions that act on bit-fields at variable positions, which may
10534 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
10535 also enables deletion of truncations of the values that serve as
10536 arguments to bit-field instructions.
10538 If both types of instructions truncate the count (for shifts) and
10539 position (for bit-field operations), or if no variable-position bit-field
10540 instructions exist, you should define this macro.
10542 However, on some machines, such as the 80386 and the 680x0, truncation
10543 only applies to shift operations and not the (real or pretended)
10544 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
10545 such machines. Instead, add patterns to the @file{md} file that include
10546 the implied truncation of the shift instructions.
10548 You need not define this macro if it would always have the value of zero.
10551 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
10552 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_SHIFT_TRUNCATION_MASK (machine_mode @var{mode})
10553 This function describes how the standard shift patterns for @var{mode}
10554 deal with shifts by negative amounts or by more than the width of the mode.
10555 @xref{shift patterns}.
10557 On many machines, the shift patterns will apply a mask @var{m} to the
10558 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
10559 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
10560 this is true for mode @var{mode}, the function should return @var{m},
10561 otherwise it should return 0. A return value of 0 indicates that no
10562 particular behavior is guaranteed.
10564 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
10565 @emph{not} apply to general shift rtxes; it applies only to instructions
10566 that are generated by the named shift patterns.
10568 The default implementation of this function returns
10569 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
10570 and 0 otherwise. This definition is always safe, but if
10571 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
10572 nevertheless truncate the shift count, you may get better code
10576 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10577 A C expression which is nonzero if on this machine it is safe to
10578 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10579 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10580 operating on it as if it had only @var{outprec} bits.
10582 On many machines, this expression can be 1.
10584 @c rearranged this, removed the phrase "it is reported that". this was
10585 @c to fix an overfull hbox. --mew 10feb93
10586 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10587 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10588 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10589 such cases may improve things.
10592 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (machine_mode @var{mode}, machine_mode @var{rep_mode})
10593 The representation of an integral mode can be such that the values
10594 are always extended to a wider integral mode. Return
10595 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10596 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10597 otherwise. (Currently, none of the targets use zero-extended
10598 representation this way so unlike @code{LOAD_EXTEND_OP},
10599 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10600 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10601 @var{mode} to @var{rep_mode} so that @var{rep_mode} is not the next
10602 widest integral mode and currently we take advantage of this fact.)
10604 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10605 value even if the extension is not performed on certain hard registers
10606 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10607 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10609 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10610 describe two related properties. If you define
10611 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10612 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10615 In order to enforce the representation of @code{mode},
10616 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10620 @defmac STORE_FLAG_VALUE
10621 A C expression describing the value returned by a comparison operator
10622 with an integral mode and stored by a store-flag instruction
10623 (@samp{cstore@var{mode}4}) when the condition is true. This description must
10624 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
10625 comparison operators whose results have a @code{MODE_INT} mode.
10627 A value of 1 or @minus{}1 means that the instruction implementing the
10628 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10629 and 0 when the comparison is false. Otherwise, the value indicates
10630 which bits of the result are guaranteed to be 1 when the comparison is
10631 true. This value is interpreted in the mode of the comparison
10632 operation, which is given by the mode of the first operand in the
10633 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
10634 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10637 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10638 generate code that depends only on the specified bits. It can also
10639 replace comparison operators with equivalent operations if they cause
10640 the required bits to be set, even if the remaining bits are undefined.
10641 For example, on a machine whose comparison operators return an
10642 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10643 @samp{0x80000000}, saying that just the sign bit is relevant, the
10647 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10651 can be converted to
10654 (ashift:SI @var{x} (const_int @var{n}))
10658 where @var{n} is the appropriate shift count to move the bit being
10659 tested into the sign bit.
10661 There is no way to describe a machine that always sets the low-order bit
10662 for a true value, but does not guarantee the value of any other bits,
10663 but we do not know of any machine that has such an instruction. If you
10664 are trying to port GCC to such a machine, include an instruction to
10665 perform a logical-and of the result with 1 in the pattern for the
10666 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10668 Often, a machine will have multiple instructions that obtain a value
10669 from a comparison (or the condition codes). Here are rules to guide the
10670 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10675 Use the shortest sequence that yields a valid definition for
10676 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10677 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10678 comparison operators to do so because there may be opportunities to
10679 combine the normalization with other operations.
10682 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10683 slightly preferred on machines with expensive jumps and 1 preferred on
10687 As a second choice, choose a value of @samp{0x80000001} if instructions
10688 exist that set both the sign and low-order bits but do not define the
10692 Otherwise, use a value of @samp{0x80000000}.
10695 Many machines can produce both the value chosen for
10696 @code{STORE_FLAG_VALUE} and its negation in the same number of
10697 instructions. On those machines, you should also define a pattern for
10698 those cases, e.g., one matching
10701 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10704 Some machines can also perform @code{and} or @code{plus} operations on
10705 condition code values with less instructions than the corresponding
10706 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
10707 machines, define the appropriate patterns. Use the names @code{incscc}
10708 and @code{decscc}, respectively, for the patterns which perform
10709 @code{plus} or @code{minus} operations on condition code values. See
10710 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
10711 find such instruction sequences on other machines.
10713 If this macro is not defined, the default value, 1, is used. You need
10714 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10715 instructions, or if the value generated by these instructions is 1.
10718 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10719 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10720 returned when comparison operators with floating-point results are true.
10721 Define this macro on machines that have comparison operations that return
10722 floating-point values. If there are no such operations, do not define
10726 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10727 A C expression that gives a rtx representing the nonzero true element
10728 for vector comparisons. The returned rtx should be valid for the inner
10729 mode of @var{mode} which is guaranteed to be a vector mode. Define
10730 this macro on machines that have vector comparison operations that
10731 return a vector result. If there are no such operations, do not define
10732 this macro. Typically, this macro is defined as @code{const1_rtx} or
10733 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10734 the compiler optimizing such vector comparison operations for the
10738 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10739 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10740 A C expression that indicates whether the architecture defines a value
10741 for @code{clz} or @code{ctz} with a zero operand.
10742 A result of @code{0} indicates the value is undefined.
10743 If the value is defined for only the RTL expression, the macro should
10744 evaluate to @code{1}; if the value applies also to the corresponding optab
10745 entry (which is normally the case if it expands directly into
10746 the corresponding RTL), then the macro should evaluate to @code{2}.
10747 In the cases where the value is defined, @var{value} should be set to
10750 If this macro is not defined, the value of @code{clz} or
10751 @code{ctz} at zero is assumed to be undefined.
10753 This macro must be defined if the target's expansion for @code{ffs}
10754 relies on a particular value to get correct results. Otherwise it
10755 is not necessary, though it may be used to optimize some corner cases, and
10756 to provide a default expansion for the @code{ffs} optab.
10758 Note that regardless of this macro the ``definedness'' of @code{clz}
10759 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10760 visible to the user. Thus one may be free to adjust the value at will
10761 to match the target expansion of these operations without fear of
10766 An alias for the machine mode for pointers. On most machines, define
10767 this to be the integer mode corresponding to the width of a hardware
10768 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10769 On some machines you must define this to be one of the partial integer
10770 modes, such as @code{PSImode}.
10772 The width of @code{Pmode} must be at least as large as the value of
10773 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10774 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10778 @defmac FUNCTION_MODE
10779 An alias for the machine mode used for memory references to functions
10780 being called, in @code{call} RTL expressions. On most CISC machines,
10781 where an instruction can begin at any byte address, this should be
10782 @code{QImode}. On most RISC machines, where all instructions have fixed
10783 size and alignment, this should be a mode with the same size and alignment
10784 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10787 @defmac STDC_0_IN_SYSTEM_HEADERS
10788 In normal operation, the preprocessor expands @code{__STDC__} to the
10789 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10790 hosts, like Solaris, the system compiler uses a different convention,
10791 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10792 strict conformance to the C Standard.
10794 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10795 convention when processing system header files, but when processing user
10796 files @code{__STDC__} will always expand to 1.
10799 @deftypefn {C Target Hook} {const char *} TARGET_C_PREINCLUDE (void)
10800 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.
10802 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.
10805 @deftypefn {C Target Hook} bool TARGET_CXX_IMPLICIT_EXTERN_C (const char*@var{})
10806 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.
10809 @defmac NO_IMPLICIT_EXTERN_C
10810 Define this macro if the system header files support C++ as well as C@.
10811 This macro inhibits the usual method of using system header files in
10812 C++, which is to pretend that the file's contents are enclosed in
10813 @samp{extern "C" @{@dots{}@}}.
10818 @defmac REGISTER_TARGET_PRAGMAS ()
10819 Define this macro if you want to implement any target-specific pragmas.
10820 If defined, it is a C expression which makes a series of calls to
10821 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10822 for each pragma. The macro may also do any
10823 setup required for the pragmas.
10825 The primary reason to define this macro is to provide compatibility with
10826 other compilers for the same target. In general, we discourage
10827 definition of target-specific pragmas for GCC@.
10829 If the pragma can be implemented by attributes then you should consider
10830 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10832 Preprocessor macros that appear on pragma lines are not expanded. All
10833 @samp{#pragma} directives that do not match any registered pragma are
10834 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10837 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10838 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10840 Each call to @code{c_register_pragma} or
10841 @code{c_register_pragma_with_expansion} establishes one pragma. The
10842 @var{callback} routine will be called when the preprocessor encounters a
10846 #pragma [@var{space}] @var{name} @dots{}
10849 @var{space} is the case-sensitive namespace of the pragma, or
10850 @code{NULL} to put the pragma in the global namespace. The callback
10851 routine receives @var{pfile} as its first argument, which can be passed
10852 on to cpplib's functions if necessary. You can lex tokens after the
10853 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10854 callback will be silently ignored. The end of the line is indicated by
10855 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10856 arguments of pragmas registered with
10857 @code{c_register_pragma_with_expansion} but not on the arguments of
10858 pragmas registered with @code{c_register_pragma}.
10860 Note that the use of @code{pragma_lex} is specific to the C and C++
10861 compilers. It will not work in the Java or Fortran compilers, or any
10862 other language compilers for that matter. Thus if @code{pragma_lex} is going
10863 to be called from target-specific code, it must only be done so when
10864 building the C and C++ compilers. This can be done by defining the
10865 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10866 target entry in the @file{config.gcc} file. These variables should name
10867 the target-specific, language-specific object file which contains the
10868 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10869 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10870 how to build this object file.
10873 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10874 Define this macro if macros should be expanded in the
10875 arguments of @samp{#pragma pack}.
10878 @defmac TARGET_DEFAULT_PACK_STRUCT
10879 If your target requires a structure packing default other than 0 (meaning
10880 the machine default), define this macro to the necessary value (in bytes).
10881 This must be a value that would also be valid to use with
10882 @samp{#pragma pack()} (that is, a small power of two).
10885 @defmac DOLLARS_IN_IDENTIFIERS
10886 Define this macro to control use of the character @samp{$} in
10887 identifier names for the C family of languages. 0 means @samp{$} is
10888 not allowed by default; 1 means it is allowed. 1 is the default;
10889 there is no need to define this macro in that case.
10892 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10893 Define this macro as a C expression that is nonzero if it is safe for the
10894 delay slot scheduler to place instructions in the delay slot of @var{insn},
10895 even if they appear to use a resource set or clobbered in @var{insn}.
10896 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10897 every @code{call_insn} has this behavior. On machines where some @code{insn}
10898 or @code{jump_insn} is really a function call and hence has this behavior,
10899 you should define this macro.
10901 You need not define this macro if it would always return zero.
10904 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10905 Define this macro as a C expression that is nonzero if it is safe for the
10906 delay slot scheduler to place instructions in the delay slot of @var{insn},
10907 even if they appear to set or clobber a resource referenced in @var{insn}.
10908 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10909 some @code{insn} or @code{jump_insn} is really a function call and its operands
10910 are registers whose use is actually in the subroutine it calls, you should
10911 define this macro. Doing so allows the delay slot scheduler to move
10912 instructions which copy arguments into the argument registers into the delay
10913 slot of @var{insn}.
10915 You need not define this macro if it would always return zero.
10918 @defmac MULTIPLE_SYMBOL_SPACES
10919 Define this macro as a C expression that is nonzero if, in some cases,
10920 global symbols from one translation unit may not be bound to undefined
10921 symbols in another translation unit without user intervention. For
10922 instance, under Microsoft Windows symbols must be explicitly imported
10923 from shared libraries (DLLs).
10925 You need not define this macro if it would always evaluate to zero.
10928 @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})
10929 This target hook may add @dfn{clobbers} to @var{clobbers} and
10930 @var{clobbered_regs} for any hard regs the port wishes to automatically
10931 clobber for an asm. The @var{outputs} and @var{inputs} may be inspected
10932 to avoid clobbering a register that is already used by the asm.
10934 It may modify the @var{outputs}, @var{inputs}, and @var{constraints}
10935 as necessary for other pre-processing. In this case the return value is
10936 a sequence of insns to emit after the asm.
10939 @defmac MATH_LIBRARY
10940 Define this macro as a C string constant for the linker argument to link
10941 in the system math library, minus the initial @samp{"-l"}, or
10942 @samp{""} if the target does not have a
10943 separate math library.
10945 You need only define this macro if the default of @samp{"m"} is wrong.
10948 @defmac LIBRARY_PATH_ENV
10949 Define this macro as a C string constant for the environment variable that
10950 specifies where the linker should look for libraries.
10952 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10956 @defmac TARGET_POSIX_IO
10957 Define this macro if the target supports the following POSIX@ file
10958 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10959 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10960 to use file locking when exiting a program, which avoids race conditions
10961 if the program has forked. It will also create directories at run-time
10962 for cross-profiling.
10965 @defmac MAX_CONDITIONAL_EXECUTE
10967 A C expression for the maximum number of instructions to execute via
10968 conditional execution instructions instead of a branch. A value of
10969 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10970 1 if it does use cc0.
10973 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10974 Used if the target needs to perform machine-dependent modifications on the
10975 conditionals used for turning basic blocks into conditionally executed code.
10976 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10977 contains information about the currently processed blocks. @var{true_expr}
10978 and @var{false_expr} are the tests that are used for converting the
10979 then-block and the else-block, respectively. Set either @var{true_expr} or
10980 @var{false_expr} to a null pointer if the tests cannot be converted.
10983 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10984 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10985 if-statements into conditions combined by @code{and} and @code{or} operations.
10986 @var{bb} contains the basic block that contains the test that is currently
10987 being processed and about to be turned into a condition.
10990 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10991 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10992 be converted to conditional execution format. @var{ce_info} points to
10993 a data structure, @code{struct ce_if_block}, which contains information
10994 about the currently processed blocks.
10997 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10998 A C expression to perform any final machine dependent modifications in
10999 converting code to conditional execution. The involved basic blocks
11000 can be found in the @code{struct ce_if_block} structure that is pointed
11001 to by @var{ce_info}.
11004 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
11005 A C expression to cancel any machine dependent modifications in
11006 converting code to conditional execution. The involved basic blocks
11007 can be found in the @code{struct ce_if_block} structure that is pointed
11008 to by @var{ce_info}.
11011 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
11012 A C expression to initialize any machine specific data for if-conversion
11013 of the if-block in the @code{struct ce_if_block} structure that is pointed
11014 to by @var{ce_info}.
11017 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG (void)
11018 If non-null, this hook performs a target-specific pass over the
11019 instruction stream. The compiler will run it at all optimization levels,
11020 just before the point at which it normally does delayed-branch scheduling.
11022 The exact purpose of the hook varies from target to target. Some use
11023 it to do transformations that are necessary for correctness, such as
11024 laying out in-function constant pools or avoiding hardware hazards.
11025 Others use it as an opportunity to do some machine-dependent optimizations.
11027 You need not implement the hook if it has nothing to do. The default
11028 definition is null.
11031 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS (void)
11032 Define this hook if you have any machine-specific built-in functions
11033 that need to be defined. It should be a function that performs the
11036 Machine specific built-in functions can be useful to expand special machine
11037 instructions that would otherwise not normally be generated because
11038 they have no equivalent in the source language (for example, SIMD vector
11039 instructions or prefetch instructions).
11041 To create a built-in function, call the function
11042 @code{lang_hooks.builtin_function}
11043 which is defined by the language front end. You can use any type nodes set
11044 up by @code{build_common_tree_nodes};
11045 only language front ends that use those two functions will call
11046 @samp{TARGET_INIT_BUILTINS}.
11049 @deftypefn {Target Hook} tree TARGET_BUILTIN_DECL (unsigned @var{code}, bool @var{initialize_p})
11050 Define this hook if you have any machine-specific built-in functions
11051 that need to be defined. It should be a function that returns the
11052 builtin function declaration for the builtin function code @var{code}.
11053 If there is no such builtin and it cannot be initialized at this time
11054 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
11055 If @var{code} is out of range the function should return
11056 @code{error_mark_node}.
11059 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, machine_mode @var{mode}, int @var{ignore})
11061 Expand a call to a machine specific built-in function that was set up by
11062 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
11063 function call; the result should go to @var{target} if that is
11064 convenient, and have mode @var{mode} if that is convenient.
11065 @var{subtarget} may be used as the target for computing one of
11066 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
11067 ignored. This function should return the result of the call to the
11071 @deftypefn {Target Hook} tree TARGET_BUILTIN_CHKP_FUNCTION (unsigned @var{fcode})
11072 This hook allows target to redefine built-in functions used by
11073 Pointer Bounds Checker for code instrumentation. Hook should return
11074 fndecl of function implementing generic builtin whose code is
11075 passed in @var{fcode}. Currently following built-in functions are
11076 obtained using this hook:
11077 @deftypefn {Built-in Function} __bounds_type __chkp_bndmk (const void *@var{lb}, size_t @var{size})
11078 Function code - BUILT_IN_CHKP_BNDMK. This built-in function is used
11079 by Pointer Bounds Checker to create bound values. @var{lb} holds low
11080 bound of the resulting bounds. @var{size} holds size of created bounds.
11083 @deftypefn {Built-in Function} void __chkp_bndstx (const void *@var{ptr}, __bounds_type @var{b}, const void **@var{loc})
11084 Function code - @code{BUILT_IN_CHKP_BNDSTX}. This built-in function is used
11085 by Pointer Bounds Checker to store bounds @var{b} for pointer @var{ptr}
11086 when @var{ptr} is stored by address @var{loc}.
11089 @deftypefn {Built-in Function} __bounds_type __chkp_bndldx (const void **@var{loc}, const void *@var{ptr})
11090 Function code - @code{BUILT_IN_CHKP_BNDLDX}. This built-in function is used
11091 by Pointer Bounds Checker to get bounds of pointer @var{ptr} loaded by
11095 @deftypefn {Built-in Function} void __chkp_bndcl (const void *@var{ptr}, __bounds_type @var{b})
11096 Function code - @code{BUILT_IN_CHKP_BNDCL}. This built-in function is used
11097 by Pointer Bounds Checker to perform check for pointer @var{ptr} against
11098 lower bound of bounds @var{b}.
11101 @deftypefn {Built-in Function} void __chkp_bndcu (const void *@var{ptr}, __bounds_type @var{b})
11102 Function code - @code{BUILT_IN_CHKP_BNDCU}. This built-in function is used
11103 by Pointer Bounds Checker to perform check for pointer @var{ptr} against
11104 upper bound of bounds @var{b}.
11107 @deftypefn {Built-in Function} __bounds_type __chkp_bndret (void *@var{ptr})
11108 Function code - @code{BUILT_IN_CHKP_BNDRET}. This built-in function is used
11109 by Pointer Bounds Checker to obtain bounds returned by a call statement.
11110 @var{ptr} passed to built-in is @code{SSA_NAME} returned by the call.
11113 @deftypefn {Built-in Function} __bounds_type __chkp_intersect (__bounds_type @var{b1}, __bounds_type @var{b2})
11114 Function code - @code{BUILT_IN_CHKP_INTERSECT}. This built-in function
11115 returns intersection of bounds @var{b1} and @var{b2}.
11118 @deftypefn {Built-in Function} __bounds_type __chkp_narrow (const void *@var{ptr}, __bounds_type @var{b}, size_t @var{s})
11119 Function code - @code{BUILT_IN_CHKP_NARROW}. This built-in function
11120 returns intersection of bounds @var{b} and
11121 [@var{ptr}, @var{ptr} + @var{s} - @code{1}].
11124 @deftypefn {Built-in Function} size_t __chkp_sizeof (const void *@var{ptr})
11125 Function code - @code{BUILT_IN_CHKP_SIZEOF}. This built-in function
11126 returns size of object referenced by @var{ptr}. @var{ptr} is always
11127 @code{ADDR_EXPR} of @code{VAR_DECL}. This built-in is used by
11128 Pointer Bounds Checker when bounds of object cannot be computed statically
11129 (e.g. object has incomplete type).
11132 @deftypefn {Built-in Function} const void *__chkp_extract_lower (__bounds_type @var{b})
11133 Function code - @code{BUILT_IN_CHKP_EXTRACT_LOWER}. This built-in function
11134 returns lower bound of bounds @var{b}.
11137 @deftypefn {Built-in Function} const void *__chkp_extract_upper (__bounds_type @var{b})
11138 Function code - @code{BUILT_IN_CHKP_EXTRACT_UPPER}. This built-in function
11139 returns upper bound of bounds @var{b}.
11142 @deftypefn {Target Hook} tree TARGET_CHKP_BOUND_TYPE (void)
11143 Return type to be used for bounds
11145 @deftypefn {Target Hook} {enum machine_mode} TARGET_CHKP_BOUND_MODE (void)
11146 Return mode to be used for bounds.
11148 @deftypefn {Target Hook} tree TARGET_CHKP_MAKE_BOUNDS_CONSTANT (HOST_WIDE_INT @var{lb}, HOST_WIDE_INT @var{ub})
11149 Return constant used to statically initialize constant bounds
11150 with specified lower bound @var{lb} and upper bounds @var{ub}.
11152 @deftypefn {Target Hook} int TARGET_CHKP_INITIALIZE_BOUNDS (tree @var{var}, tree @var{lb}, tree @var{ub}, tree *@var{stmts})
11153 Generate a list of statements @var{stmts} to initialize pointer
11154 bounds variable @var{var} with bounds @var{lb} and @var{ub}. Return
11155 the number of generated statements.
11158 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (unsigned int @var{loc}, tree @var{fndecl}, void *@var{arglist})
11159 Select a replacement for a machine specific built-in function that
11160 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
11161 @emph{before} regular type checking, and so allows the target to
11162 implement a crude form of function overloading. @var{fndecl} is the
11163 declaration of the built-in function. @var{arglist} is the list of
11164 arguments passed to the built-in function. The result is a
11165 complete expression that implements the operation, usually
11166 another @code{CALL_EXPR}.
11167 @var{arglist} really has type @samp{VEC(tree,gc)*}
11170 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, int @var{n_args}, tree *@var{argp}, bool @var{ignore})
11171 Fold a call to a machine specific built-in function that was set up by
11172 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
11173 built-in function. @var{n_args} is the number of arguments passed to
11174 the function; the arguments themselves are pointed to by @var{argp}.
11175 The result is another tree, valid for both GIMPLE and GENERIC,
11176 containing a simplified expression for the call's result. If
11177 @var{ignore} is true the value will be ignored.
11180 @deftypefn {Target Hook} bool TARGET_GIMPLE_FOLD_BUILTIN (gimple_stmt_iterator *@var{gsi})
11181 Fold a call to a machine specific built-in function that was set up
11182 by @samp{TARGET_INIT_BUILTINS}. @var{gsi} points to the gimple
11183 statement holding the function call. Returns true if any change
11184 was made to the GIMPLE stream.
11187 @deftypefn {Target Hook} int TARGET_COMPARE_VERSION_PRIORITY (tree @var{decl1}, tree @var{decl2})
11188 This hook is used to compare the target attributes in two functions to
11189 determine which function's features get higher priority. This is used
11190 during function multi-versioning to figure out the order in which two
11191 versions must be dispatched. A function version with a higher priority
11192 is checked for dispatching earlier. @var{decl1} and @var{decl2} are
11193 the two function decls that will be compared.
11196 @deftypefn {Target Hook} tree TARGET_GET_FUNCTION_VERSIONS_DISPATCHER (void *@var{decl})
11197 This hook is used to get the dispatcher function for a set of function
11198 versions. The dispatcher function is called to invoke the right function
11199 version at run-time. @var{decl} is one version from a set of semantically
11200 identical versions.
11203 @deftypefn {Target Hook} tree TARGET_GENERATE_VERSION_DISPATCHER_BODY (void *@var{arg})
11204 This hook is used to generate the dispatcher logic to invoke the right
11205 function version at run-time for a given set of function versions.
11206 @var{arg} points to the callgraph node of the dispatcher function whose
11207 body must be generated.
11210 @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})
11211 Return true if it is possible to use low-overhead loops (@code{doloop_end}
11212 and @code{doloop_begin}) for a particular loop. @var{iterations} gives the
11213 exact number of iterations, or 0 if not known. @var{iterations_max} gives
11214 the maximum number of iterations, or 0 if not known. @var{loop_depth} is
11215 the nesting depth of the loop, with 1 for innermost loops, 2 for loops that
11216 contain innermost loops, and so on. @var{entered_at_top} is true if the
11217 loop is only entered from the top.
11219 This hook is only used if @code{doloop_end} is available. The default
11220 implementation returns true. You can use @code{can_use_doloop_if_innermost}
11221 if the loop must be the innermost, and if there are no other restrictions.
11224 @deftypefn {Target Hook} {const char *} TARGET_INVALID_WITHIN_DOLOOP (const rtx_insn *@var{insn})
11226 Take an instruction in @var{insn} and return NULL if it is valid within a
11227 low-overhead loop, otherwise return a string explaining why doloop
11228 could not be applied.
11230 Many targets use special registers for low-overhead looping. For any
11231 instruction that clobbers these this function should return a string indicating
11232 the reason why the doloop could not be applied.
11233 By default, the RTL loop optimizer does not use a present doloop pattern for
11234 loops containing function calls or branch on table instructions.
11237 @deftypefn {Target Hook} bool TARGET_LEGITIMATE_COMBINED_INSN (rtx_insn *@var{insn})
11238 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.
11241 @deftypefn {Target Hook} bool TARGET_CAN_FOLLOW_JUMP (const rtx_insn *@var{follower}, const rtx_insn *@var{followee})
11242 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.
11245 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (const_rtx @var{x}, int @var{outer_code})
11246 This target hook returns @code{true} if @var{x} is considered to be commutative.
11247 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
11248 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
11249 of the enclosing rtl, if known, otherwise it is UNKNOWN.
11252 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
11254 When the initial value of a hard register has been copied in a pseudo
11255 register, it is often not necessary to actually allocate another register
11256 to this pseudo register, because the original hard register or a stack slot
11257 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
11258 is called at the start of register allocation once for each hard register
11259 that had its initial value copied by using
11260 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
11261 Possible values are @code{NULL_RTX}, if you don't want
11262 to do any special allocation, a @code{REG} rtx---that would typically be
11263 the hard register itself, if it is known not to be clobbered---or a
11265 If you are returning a @code{MEM}, this is only a hint for the allocator;
11266 it might decide to use another register anyways.
11267 You may use @code{current_function_is_leaf} or
11268 @code{REG_N_SETS} in the hook to determine if the hard
11269 register in question will not be clobbered.
11270 The default value of this hook is @code{NULL}, which disables any special
11274 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
11275 This target hook returns nonzero if @var{x}, an @code{unspec} or
11276 @code{unspec_volatile} operation, might cause a trap. Targets can use
11277 this hook to enhance precision of analysis for @code{unspec} and
11278 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
11279 to analyze inner elements of @var{x} in which case @var{flags} should be
11283 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
11284 The compiler invokes this hook whenever it changes its current function
11285 context (@code{cfun}). You can define this function if
11286 the back end needs to perform any initialization or reset actions on a
11287 per-function basis. For example, it may be used to implement function
11288 attributes that affect register usage or code generation patterns.
11289 The argument @var{decl} is the declaration for the new function context,
11290 and may be null to indicate that the compiler has left a function context
11291 and is returning to processing at the top level.
11292 The default hook function does nothing.
11294 GCC sets @code{cfun} to a dummy function context during initialization of
11295 some parts of the back end. The hook function is not invoked in this
11296 situation; you need not worry about the hook being invoked recursively,
11297 or when the back end is in a partially-initialized state.
11298 @code{cfun} might be @code{NULL} to indicate processing at top level,
11299 outside of any function scope.
11302 @defmac TARGET_OBJECT_SUFFIX
11303 Define this macro to be a C string representing the suffix for object
11304 files on your target machine. If you do not define this macro, GCC will
11305 use @samp{.o} as the suffix for object files.
11308 @defmac TARGET_EXECUTABLE_SUFFIX
11309 Define this macro to be a C string representing the suffix to be
11310 automatically added to executable files on your target machine. If you
11311 do not define this macro, GCC will use the null string as the suffix for
11315 @defmac COLLECT_EXPORT_LIST
11316 If defined, @code{collect2} will scan the individual object files
11317 specified on its command line and create an export list for the linker.
11318 Define this macro for systems like AIX, where the linker discards
11319 object files that are not referenced from @code{main} and uses export
11323 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
11324 Define this macro to a C expression representing a variant of the
11325 method call @var{mdecl}, if Java Native Interface (JNI) methods
11326 must be invoked differently from other methods on your target.
11327 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
11328 the @code{stdcall} calling convention and this macro is then
11329 defined as this expression:
11332 build_type_attribute_variant (@var{mdecl},
11334 (get_identifier ("stdcall"),
11339 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
11340 This target hook returns @code{true} past the point in which new jump
11341 instructions could be created. On machines that require a register for
11342 every jump such as the SHmedia ISA of SH5, this point would typically be
11343 reload, so this target hook should be defined to a function such as:
11347 cannot_modify_jumps_past_reload_p ()
11349 return (reload_completed || reload_in_progress);
11354 @deftypefn {Target Hook} reg_class_t TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
11355 This target hook returns a register class for which branch target register
11356 optimizations should be applied. All registers in this class should be
11357 usable interchangeably. After reload, registers in this class will be
11358 re-allocated and loads will be hoisted out of loops and be subjected
11359 to inter-block scheduling.
11362 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
11363 Branch target register optimization will by default exclude callee-saved
11365 that are not already live during the current function; if this target hook
11366 returns true, they will be included. The target code must than make sure
11367 that all target registers in the class returned by
11368 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
11369 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
11370 epilogues have already been generated. Note, even if you only return
11371 true when @var{after_prologue_epilogue_gen} is false, you still are likely
11372 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
11373 to reserve space for caller-saved target registers.
11376 @deftypefn {Target Hook} bool TARGET_HAVE_CONDITIONAL_EXECUTION (void)
11377 This target hook returns true if the target supports conditional execution.
11378 This target hook is required only when the target has several different
11379 modes and they have different conditional execution capability, such as ARM.
11382 @deftypefn {Target Hook} rtx TARGET_GEN_CCMP_FIRST (rtx *@var{prep_seq}, rtx *@var{gen_seq}, int @var{code}, tree @var{op0}, tree @var{op1})
11383 This function prepares to emit a comparison insn for the first compare in a
11384 sequence of conditional comparisions. It returns an appropriate comparison
11385 with @code{CC} for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11386 The insns to prepare the compare are saved in @var{prep_seq} and the compare
11387 insns are saved in @var{gen_seq}. They will be emitted when all the
11388 compares in the the conditional comparision are generated without error.
11389 @var{code} is the @code{rtx_code} of the compare for @var{op0} and @var{op1}.
11392 @deftypefn {Target Hook} rtx TARGET_GEN_CCMP_NEXT (rtx *@var{prep_seq}, rtx *@var{gen_seq}, rtx @var{prev}, int @var{cmp_code}, tree @var{op0}, tree @var{op1}, int @var{bit_code})
11393 This function prepares to emit a conditional comparison within a sequence
11394 of conditional comparisons. It returns an appropriate comparison with
11395 @code{CC} for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11396 The insns to prepare the compare are saved in @var{prep_seq} and the compare
11397 insns are saved in @var{gen_seq}. They will be emitted when all the
11398 compares in the conditional comparision are generated without error. The
11399 @var{prev} expression is the result of a prior call to @code{gen_ccmp_first}
11400 or @code{gen_ccmp_next}. It may return @code{NULL} if the combination of
11401 @var{prev} and this comparison is not supported, otherwise the result must
11402 be appropriate for passing to @code{gen_ccmp_next} or @code{cbranch_optab}.
11403 @var{code} is the @code{rtx_code} of the compare for @var{op0} and @var{op1}.
11404 @var{bit_code} is @code{AND} or @code{IOR}, which is the op on the compares.
11407 @deftypefn {Target Hook} unsigned TARGET_LOOP_UNROLL_ADJUST (unsigned @var{nunroll}, struct loop *@var{loop})
11408 This target hook returns a new value for the number of times @var{loop}
11409 should be unrolled. The parameter @var{nunroll} is the number of times
11410 the loop is to be unrolled. The parameter @var{loop} is a pointer to
11411 the loop, which is going to be checked for unrolling. This target hook
11412 is required only when the target has special constraints like maximum
11413 number of memory accesses.
11416 @defmac POWI_MAX_MULTS
11417 If defined, this macro is interpreted as a signed integer C expression
11418 that specifies the maximum number of floating point multiplications
11419 that should be emitted when expanding exponentiation by an integer
11420 constant inline. When this value is defined, exponentiation requiring
11421 more than this number of multiplications is implemented by calling the
11422 system library's @code{pow}, @code{powf} or @code{powl} routines.
11423 The default value places no upper bound on the multiplication count.
11426 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11427 This target hook should register any extra include files for the
11428 target. The parameter @var{stdinc} indicates if normal include files
11429 are present. The parameter @var{sysroot} is the system root directory.
11430 The parameter @var{iprefix} is the prefix for the gcc directory.
11433 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
11434 This target hook should register any extra include files for the
11435 target before any standard headers. The parameter @var{stdinc}
11436 indicates if normal include files are present. The parameter
11437 @var{sysroot} is the system root directory. The parameter
11438 @var{iprefix} is the prefix for the gcc directory.
11441 @deftypefn Macro void TARGET_OPTF (char *@var{path})
11442 This target hook should register special include paths for the target.
11443 The parameter @var{path} is the include to register. On Darwin
11444 systems, this is used for Framework includes, which have semantics
11445 that are different from @option{-I}.
11448 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
11449 This target macro returns @code{true} if it is safe to use a local alias
11450 for a virtual function @var{fndecl} when constructing thunks,
11451 @code{false} otherwise. By default, the macro returns @code{true} for all
11452 functions, if a target supports aliases (i.e.@: defines
11453 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
11456 @defmac TARGET_FORMAT_TYPES
11457 If defined, this macro is the name of a global variable containing
11458 target-specific format checking information for the @option{-Wformat}
11459 option. The default is to have no target-specific format checks.
11462 @defmac TARGET_N_FORMAT_TYPES
11463 If defined, this macro is the number of entries in
11464 @code{TARGET_FORMAT_TYPES}.
11467 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
11468 If defined, this macro is the name of a global variable containing
11469 target-specific format overrides for the @option{-Wformat} option. The
11470 default is to have no target-specific format overrides. If defined,
11471 @code{TARGET_FORMAT_TYPES} must be defined, too.
11474 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
11475 If defined, this macro specifies the number of entries in
11476 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
11479 @defmac TARGET_OVERRIDES_FORMAT_INIT
11480 If defined, this macro specifies the optional initialization
11481 routine for target specific customizations of the system printf
11482 and scanf formatter settings.
11485 @deftypefn {Target Hook} {const char *} TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (const_tree @var{typelist}, const_tree @var{funcdecl}, const_tree @var{val})
11486 If defined, this macro returns the diagnostic message when it is
11487 illegal to pass argument @var{val} to function @var{funcdecl}
11488 with prototype @var{typelist}.
11491 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (const_tree @var{fromtype}, const_tree @var{totype})
11492 If defined, this macro returns the diagnostic message when it is
11493 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
11494 if validity should be determined by the front end.
11497 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, const_tree @var{type})
11498 If defined, this macro returns the diagnostic message when it is
11499 invalid to apply operation @var{op} (where unary plus is denoted by
11500 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
11501 if validity should be determined by the front end.
11504 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, const_tree @var{type1}, const_tree @var{type2})
11505 If defined, this macro returns the diagnostic message when it is
11506 invalid to apply operation @var{op} to operands of types @var{type1}
11507 and @var{type2}, or @code{NULL} if validity should be determined by
11511 @deftypefn {Target Hook} tree TARGET_PROMOTED_TYPE (const_tree @var{type})
11512 If defined, this target hook returns the type to which values of
11513 @var{type} should be promoted when they appear in expressions,
11514 analogous to the integer promotions, or @code{NULL_TREE} to use the
11515 front end's normal promotion rules. This hook is useful when there are
11516 target-specific types with special promotion rules.
11517 This is currently used only by the C and C++ front ends.
11520 @deftypefn {Target Hook} tree TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
11521 If defined, this hook returns the result of converting @var{expr} to
11522 @var{type}. It should return the converted expression,
11523 or @code{NULL_TREE} to apply the front end's normal conversion rules.
11524 This hook is useful when there are target-specific types with special
11526 This is currently used only by the C and C++ front ends.
11529 @defmac TARGET_USE_JCR_SECTION
11530 This macro determines whether to use the JCR section to register Java
11531 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
11532 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
11536 This macro determines the size of the objective C jump buffer for the
11537 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
11540 @defmac LIBGCC2_UNWIND_ATTRIBUTE
11541 Define this macro if any target-specific attributes need to be attached
11542 to the functions in @file{libgcc} that provide low-level support for
11543 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
11544 and the associated definitions of those functions.
11547 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
11548 Define this macro to update the current function stack boundary if
11552 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
11553 This hook should return an rtx for Dynamic Realign Argument Pointer (DRAP) if a
11554 different argument pointer register is needed to access the function's
11555 argument list due to stack realignment. Return @code{NULL} if no DRAP
11559 @deftypefn {Target Hook} bool TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
11560 When optimization is disabled, this hook indicates whether or not
11561 arguments should be allocated to stack slots. Normally, GCC allocates
11562 stacks slots for arguments when not optimizing in order to make
11563 debugging easier. However, when a function is declared with
11564 @code{__attribute__((naked))}, there is no stack frame, and the compiler
11565 cannot safely move arguments from the registers in which they are passed
11566 to the stack. Therefore, this hook should return true in general, but
11567 false for naked functions. The default implementation always returns true.
11570 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
11571 On some architectures it can take multiple instructions to synthesize
11572 a constant. If there is another constant already in a register that
11573 is close enough in value then it is preferable that the new constant
11574 is computed from this register using immediate addition or
11575 subtraction. We accomplish this through CSE. Besides the value of
11576 the constant we also add a lower and an upper constant anchor to the
11577 available expressions. These are then queried when encountering new
11578 constants. The anchors are computed by rounding the constant up and
11579 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
11580 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
11581 accepted by immediate-add plus one. We currently assume that the
11582 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
11583 MIPS, where add-immediate takes a 16-bit signed value,
11584 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
11585 is zero, which disables this optimization.
11588 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_ASAN_SHADOW_OFFSET (void)
11589 Return the offset bitwise ored into shifted address to get corresponding
11590 Address Sanitizer shadow memory address. NULL if Address Sanitizer is not
11591 supported by the target.
11594 @deftypefn {Target Hook} {unsigned HOST_WIDE_INT} TARGET_MEMMODEL_CHECK (unsigned HOST_WIDE_INT @var{val})
11595 Validate target specific memory model mask bits. When NULL no target specific
11596 memory model bits are allowed.
11599 @deftypevr {Target Hook} {unsigned char} TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
11600 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}.
11603 @deftypefn {Target Hook} bool TARGET_HAS_IFUNC_P (void)
11604 It returns true if the target supports GNU indirect functions.
11605 The support includes the assembler, linker and dynamic linker.
11606 The default value of this hook is based on target's libc.
11609 @deftypefn {Target Hook} {unsigned int} TARGET_ATOMIC_ALIGN_FOR_MODE (machine_mode @var{mode})
11610 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.
11613 @deftypefn {Target Hook} void TARGET_ATOMIC_ASSIGN_EXPAND_FENV (tree *@var{hold}, tree *@var{clear}, tree *@var{update})
11614 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}}.
11617 @deftypefn {Target Hook} void TARGET_RECORD_OFFLOAD_SYMBOL (tree)
11618 Used when offloaded functions are seen in the compilation unit and no named
11619 sections are available. It is called once for each symbol that must be
11620 recorded in the offload function and variable table.
11623 @deftypefn {Target Hook} {char *} TARGET_OFFLOAD_OPTIONS (void)
11624 Used when writing out the list of options into an LTO file. It should
11625 translate any relevant target-specific options (such as the ABI in use)
11626 into one of the @option{-foffload} options that exist as a common interface
11627 to express such options. It should return a string containing these options,
11628 separated by spaces, which the caller will free.
11632 @defmac TARGET_SUPPORTS_WIDE_INT
11634 On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
11635 objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
11636 to indicate that large integers are stored in
11637 @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows
11638 very large integer constants to be represented. @code{CONST_DOUBLE}
11639 is limited to twice the size of the host's @code{HOST_WIDE_INT}
11642 Converting a port mostly requires looking for the places where
11643 @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
11644 code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i
11645 const_double"} at the port level gets you to 95% of the changes that
11646 need to be made. There are a few places that require a deeper look.
11650 There is no equivalent to @code{hval} and @code{lval} for
11651 @code{CONST_WIDE_INT}s. This would be difficult to express in the md
11652 language since there are a variable number of elements.
11654 Most ports only check that @code{hval} is either 0 or -1 to see if the
11655 value is small. As mentioned above, this will no longer be necessary
11656 since small constants are always @code{CONST_INT}. Of course there
11657 are still a few exceptions, the alpha's constraint used by the zap
11658 instruction certainly requires careful examination by C code.
11659 However, all the current code does is pass the hval and lval to C
11660 code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
11661 not really a large change.
11664 Because there is no standard template that ports use to materialize
11665 constants, there is likely to be some futzing that is unique to each
11669 The rtx costs may have to be adjusted to properly account for larger
11670 constants that are represented as @code{CONST_WIDE_INT}.
11673 All and all it does not take long to convert ports that the
11674 maintainer is familiar with.