1 @c Copyright (C) 1988-2014 Free Software Foundation, Inc.
2 @c This is part of the GCC manual.
3 @c For copying conditions, see the file gcc.texi.
6 @chapter Target Description Macros and Functions
7 @cindex machine description macros
8 @cindex target description macros
9 @cindex macros, target description
10 @cindex @file{tm.h} macros
12 In addition to the file @file{@var{machine}.md}, a machine description
13 includes a C header file conventionally given the name
14 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
15 The header file defines numerous macros that convey the information
16 about the target machine that does not fit into the scheme of the
17 @file{.md} file. The file @file{tm.h} should be a link to
18 @file{@var{machine}.h}. The header file @file{config.h} includes
19 @file{tm.h} and most compiler source files include @file{config.h}. The
20 source file defines a variable @code{targetm}, which is a structure
21 containing pointers to functions and data relating to the target
22 machine. @file{@var{machine}.c} should also contain their definitions,
23 if they are not defined elsewhere in GCC, and other functions called
24 through the macros defined in the @file{.h} file.
27 * Target Structure:: The @code{targetm} variable.
28 * Driver:: Controlling how the driver runs the compilation passes.
29 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
30 * Per-Function Data:: Defining data structures for per-function information.
31 * Storage Layout:: Defining sizes and alignments of data.
32 * Type Layout:: Defining sizes and properties of basic user data types.
33 * Registers:: Naming and describing the hardware registers.
34 * Register Classes:: Defining the classes of hardware registers.
35 * 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 LINK_COMMAND_SPEC
379 A C string constant giving the complete command line need to execute the
380 linker. When you do this, you will need to update your port each time a
381 change is made to the link command line within @file{gcc.c}. Therefore,
382 define this macro only if you need to completely redefine the command
383 line for invoking the linker and there is no other way to accomplish
384 the effect you need. Overriding this macro may be avoidable by overriding
385 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
388 @hook TARGET_ALWAYS_STRIP_DOTDOT
390 @defmac MULTILIB_DEFAULTS
391 Define this macro as a C expression for the initializer of an array of
392 string to tell the driver program which options are defaults for this
393 target and thus do not need to be handled specially when using
394 @code{MULTILIB_OPTIONS}.
396 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
397 the target makefile fragment or if none of the options listed in
398 @code{MULTILIB_OPTIONS} are set by default.
399 @xref{Target Fragment}.
402 @defmac RELATIVE_PREFIX_NOT_LINKDIR
403 Define this macro to tell @command{gcc} that it should only translate
404 a @option{-B} prefix into a @option{-L} linker option if the prefix
405 indicates an absolute file name.
408 @defmac MD_EXEC_PREFIX
409 If defined, this macro is an additional prefix to try after
410 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
411 when the compiler is built as a cross
412 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
413 to the list of directories used to find the assembler in @file{configure.in}.
416 @defmac STANDARD_STARTFILE_PREFIX
417 Define this macro as a C string constant if you wish to override the
418 standard choice of @code{libdir} as the default prefix to
419 try when searching for startup files such as @file{crt0.o}.
420 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
421 is built as a cross compiler.
424 @defmac STANDARD_STARTFILE_PREFIX_1
425 Define this macro as a C string constant if you wish to override the
426 standard choice of @code{/lib} as a prefix to try after the default prefix
427 when searching for startup files such as @file{crt0.o}.
428 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
429 is built as a cross compiler.
432 @defmac STANDARD_STARTFILE_PREFIX_2
433 Define this macro as a C string constant if you wish to override the
434 standard choice of @code{/lib} as yet another prefix to try after the
435 default prefix when searching for startup files such as @file{crt0.o}.
436 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
437 is built as a cross compiler.
440 @defmac MD_STARTFILE_PREFIX
441 If defined, this macro supplies an additional prefix to try after the
442 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
443 compiler is built as a cross compiler.
446 @defmac MD_STARTFILE_PREFIX_1
447 If defined, this macro supplies yet another prefix to try after the
448 standard prefixes. It is not searched when the compiler is built as a
452 @defmac INIT_ENVIRONMENT
453 Define this macro as a C string constant if you wish to set environment
454 variables for programs called by the driver, such as the assembler and
455 loader. The driver passes the value of this macro to @code{putenv} to
456 initialize the necessary environment variables.
459 @defmac LOCAL_INCLUDE_DIR
460 Define this macro as a C string constant if you wish to override the
461 standard choice of @file{/usr/local/include} as the default prefix to
462 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
463 comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
464 @file{config.gcc}, normally @file{/usr/include}) in the search order.
466 Cross compilers do not search either @file{/usr/local/include} or its
470 @defmac NATIVE_SYSTEM_HEADER_COMPONENT
471 The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
472 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
473 If you do not define this macro, no component is used.
476 @defmac INCLUDE_DEFAULTS
477 Define this macro if you wish to override the entire default search path
478 for include files. For a native compiler, the default search path
479 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
480 @code{GPLUSPLUS_INCLUDE_DIR}, and
481 @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
482 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
483 and specify private search areas for GCC@. The directory
484 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
486 The definition should be an initializer for an array of structures.
487 Each array element should have four elements: the directory name (a
488 string constant), the component name (also a string constant), a flag
489 for C++-only directories,
490 and a flag showing that the includes in the directory don't need to be
491 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
492 the array with a null element.
494 The component name denotes what GNU package the include file is part of,
495 if any, in all uppercase letters. For example, it might be @samp{GCC}
496 or @samp{BINUTILS}. If the package is part of a vendor-supplied
497 operating system, code the component name as @samp{0}.
499 For example, here is the definition used for VAX/VMS:
502 #define INCLUDE_DEFAULTS \
504 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
505 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
506 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
513 Here is the order of prefixes tried for exec files:
517 Any prefixes specified by the user with @option{-B}.
520 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
521 is not set and the compiler has not been installed in the configure-time
522 @var{prefix}, the location in which the compiler has actually been installed.
525 The directories specified by the environment variable @code{COMPILER_PATH}.
528 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
529 in the configured-time @var{prefix}.
532 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
535 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
538 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
542 Here is the order of prefixes tried for startfiles:
546 Any prefixes specified by the user with @option{-B}.
549 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
550 value based on the installed toolchain location.
553 The directories specified by the environment variable @code{LIBRARY_PATH}
554 (or port-specific name; native only, cross compilers do not use this).
557 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
558 in the configured @var{prefix} or this is a native compiler.
561 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
564 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
568 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
569 native compiler, or we have a target system root.
572 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
573 native compiler, or we have a target system root.
576 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
577 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
578 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
581 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
582 compiler, or we have a target system root. The default for this macro is
586 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
587 compiler, or we have a target system root. The default for this macro is
591 @node Run-time Target
592 @section Run-time Target Specification
593 @cindex run-time target specification
594 @cindex predefined macros
595 @cindex target specifications
597 @c prevent bad page break with this line
598 Here are run-time target specifications.
600 @defmac TARGET_CPU_CPP_BUILTINS ()
601 This function-like macro expands to a block of code that defines
602 built-in preprocessor macros and assertions for the target CPU, using
603 the functions @code{builtin_define}, @code{builtin_define_std} and
604 @code{builtin_assert}. When the front end
605 calls this macro it provides a trailing semicolon, and since it has
606 finished command line option processing your code can use those
609 @code{builtin_assert} takes a string in the form you pass to the
610 command-line option @option{-A}, such as @code{cpu=mips}, and creates
611 the assertion. @code{builtin_define} takes a string in the form
612 accepted by option @option{-D} and unconditionally defines the macro.
614 @code{builtin_define_std} takes a string representing the name of an
615 object-like macro. If it doesn't lie in the user's namespace,
616 @code{builtin_define_std} defines it unconditionally. Otherwise, it
617 defines a version with two leading underscores, and another version
618 with two leading and trailing underscores, and defines the original
619 only if an ISO standard was not requested on the command line. For
620 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
621 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
622 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
623 defines only @code{_ABI64}.
625 You can also test for the C dialect being compiled. The variable
626 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
627 or @code{clk_objective_c}. Note that if we are preprocessing
628 assembler, this variable will be @code{clk_c} but the function-like
629 macro @code{preprocessing_asm_p()} will return true, so you might want
630 to check for that first. If you need to check for strict ANSI, the
631 variable @code{flag_iso} can be used. The function-like macro
632 @code{preprocessing_trad_p()} can be used to check for traditional
636 @defmac TARGET_OS_CPP_BUILTINS ()
637 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
638 and is used for the target operating system instead.
641 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
642 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
643 and is used for the target object format. @file{elfos.h} uses this
644 macro to define @code{__ELF__}, so you probably do not need to define
648 @deftypevar {extern int} target_flags
649 This variable is declared in @file{options.h}, which is included before
650 any target-specific headers.
653 @hook TARGET_DEFAULT_TARGET_FLAGS
654 This variable specifies the initial value of @code{target_flags}.
655 Its default setting is 0.
658 @cindex optional hardware or system features
659 @cindex features, optional, in system conventions
661 @hook TARGET_HANDLE_OPTION
662 This hook is called whenever the user specifies one of the
663 target-specific options described by the @file{.opt} definition files
664 (@pxref{Options}). It has the opportunity to do some option-specific
665 processing and should return true if the option is valid. The default
666 definition does nothing but return true.
668 @var{decoded} specifies the option and its arguments. @var{opts} and
669 @var{opts_set} are the @code{gcc_options} structures to be used for
670 storing option state, and @var{loc} is the location at which the
671 option was passed (@code{UNKNOWN_LOCATION} except for options passed
675 @hook TARGET_HANDLE_C_OPTION
676 This target hook is called whenever the user specifies one of the
677 target-specific C language family options described by the @file{.opt}
678 definition files(@pxref{Options}). It has the opportunity to do some
679 option-specific processing and should return true if the option is
680 valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The
681 default definition does nothing but return false.
683 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
684 options. However, if processing an option requires routines that are
685 only available in the C (and related language) front ends, then you
686 should use @code{TARGET_HANDLE_C_OPTION} instead.
689 @hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT
691 @hook TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE
693 @hook TARGET_OBJC_DECLARE_CLASS_DEFINITION
695 @hook TARGET_STRING_OBJECT_REF_TYPE_P
697 @hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG
699 @hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE
701 @defmac C_COMMON_OVERRIDE_OPTIONS
702 This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
703 but is only used in the C
704 language frontends (C, Objective-C, C++, Objective-C++) and so can be
705 used to alter option flag variables which only exist in those
709 @hook TARGET_OPTION_OPTIMIZATION_TABLE
710 Some machines may desire to change what optimizations are performed for
711 various optimization levels. This variable, if defined, describes
712 options to enable at particular sets of optimization levels. These
713 options are processed once
714 just after the optimization level is determined and before the remainder
715 of the command options have been parsed, so may be overridden by other
716 options passed explicitly.
718 This processing is run once at program startup and when the optimization
719 options are changed via @code{#pragma GCC optimize} or by using the
720 @code{optimize} attribute.
723 @hook TARGET_OPTION_INIT_STRUCT
725 @hook TARGET_OPTION_DEFAULT_PARAMS
727 @defmac SWITCHABLE_TARGET
728 Some targets need to switch between substantially different subtargets
729 during compilation. For example, the MIPS target has one subtarget for
730 the traditional MIPS architecture and another for MIPS16. Source code
731 can switch between these two subarchitectures using the @code{mips16}
732 and @code{nomips16} attributes.
734 Such subtargets can differ in things like the set of available
735 registers, the set of available instructions, the costs of various
736 operations, and so on. GCC caches a lot of this type of information
737 in global variables, and recomputing them for each subtarget takes a
738 significant amount of time. The compiler therefore provides a facility
739 for maintaining several versions of the global variables and quickly
740 switching between them; see @file{target-globals.h} for details.
742 Define this macro to 1 if your target needs this facility. The default
746 @hook TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P
748 @node Per-Function Data
749 @section Defining data structures for per-function information.
750 @cindex per-function data
751 @cindex data structures
753 If the target needs to store information on a per-function basis, GCC
754 provides a macro and a couple of variables to allow this. Note, just
755 using statics to store the information is a bad idea, since GCC supports
756 nested functions, so you can be halfway through encoding one function
757 when another one comes along.
759 GCC defines a data structure called @code{struct function} which
760 contains all of the data specific to an individual function. This
761 structure contains a field called @code{machine} whose type is
762 @code{struct machine_function *}, which can be used by targets to point
763 to their own specific data.
765 If a target needs per-function specific data it should define the type
766 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
767 This macro should be used to initialize the function pointer
768 @code{init_machine_status}. This pointer is explained below.
770 One typical use of per-function, target specific data is to create an
771 RTX to hold the register containing the function's return address. This
772 RTX can then be used to implement the @code{__builtin_return_address}
773 function, for level 0.
775 Note---earlier implementations of GCC used a single data area to hold
776 all of the per-function information. Thus when processing of a nested
777 function began the old per-function data had to be pushed onto a
778 stack, and when the processing was finished, it had to be popped off the
779 stack. GCC used to provide function pointers called
780 @code{save_machine_status} and @code{restore_machine_status} to handle
781 the saving and restoring of the target specific information. Since the
782 single data area approach is no longer used, these pointers are no
785 @defmac INIT_EXPANDERS
786 Macro called to initialize any target specific information. This macro
787 is called once per function, before generation of any RTL has begun.
788 The intention of this macro is to allow the initialization of the
789 function pointer @code{init_machine_status}.
792 @deftypevar {void (*)(struct function *)} init_machine_status
793 If this function pointer is non-@code{NULL} it will be called once per
794 function, before function compilation starts, in order to allow the
795 target to perform any target specific initialization of the
796 @code{struct function} structure. It is intended that this would be
797 used to initialize the @code{machine} of that structure.
799 @code{struct machine_function} structures are expected to be freed by GC@.
800 Generally, any memory that they reference must be allocated by using
801 GC allocation, including the structure itself.
805 @section Storage Layout
806 @cindex storage layout
808 Note that the definitions of the macros in this table which are sizes or
809 alignments measured in bits do not need to be constant. They can be C
810 expressions that refer to static variables, such as the @code{target_flags}.
811 @xref{Run-time Target}.
813 @defmac BITS_BIG_ENDIAN
814 Define this macro to have the value 1 if the most significant bit in a
815 byte has the lowest number; otherwise define it to have the value zero.
816 This means that bit-field instructions count from the most significant
817 bit. If the machine has no bit-field instructions, then this must still
818 be defined, but it doesn't matter which value it is defined to. This
819 macro need not be a constant.
821 This macro does not affect the way structure fields are packed into
822 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
825 @defmac BYTES_BIG_ENDIAN
826 Define this macro to have the value 1 if the most significant byte in a
827 word has the lowest number. This macro need not be a constant.
830 @defmac WORDS_BIG_ENDIAN
831 Define this macro to have the value 1 if, in a multiword object, the
832 most significant word has the lowest number. This applies to both
833 memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
834 order of words in memory is not the same as the order in registers. This
835 macro need not be a constant.
838 @defmac REG_WORDS_BIG_ENDIAN
839 On some machines, the order of words in a multiword object differs between
840 registers in memory. In such a situation, define this macro to describe
841 the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls
842 the order of words in memory.
845 @defmac FLOAT_WORDS_BIG_ENDIAN
846 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
847 @code{TFmode} floating point numbers are stored in memory with the word
848 containing the sign bit at the lowest address; otherwise define it to
849 have the value 0. This macro need not be a constant.
851 You need not define this macro if the ordering is the same as for
855 @defmac BITS_PER_WORD
856 Number of bits in a word. If you do not define this macro, the default
857 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
860 @defmac MAX_BITS_PER_WORD
861 Maximum number of bits in a word. If this is undefined, the default is
862 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
863 largest value that @code{BITS_PER_WORD} can have at run-time.
866 @defmac UNITS_PER_WORD
867 Number of storage units in a word; normally the size of a general-purpose
868 register, a power of two from 1 or 8.
871 @defmac MIN_UNITS_PER_WORD
872 Minimum number of units in a word. If this is undefined, the default is
873 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
874 smallest value that @code{UNITS_PER_WORD} can have at run-time.
878 Width of a pointer, in bits. You must specify a value no wider than the
879 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
880 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
881 a value the default is @code{BITS_PER_WORD}.
884 @defmac POINTERS_EXTEND_UNSIGNED
885 A C expression that determines how pointers should be extended from
886 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
887 greater than zero if pointers should be zero-extended, zero if they
888 should be sign-extended, and negative if some other sort of conversion
889 is needed. In the last case, the extension is done by the target's
890 @code{ptr_extend} instruction.
892 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
893 and @code{word_mode} are all the same width.
896 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
897 A macro to update @var{m} and @var{unsignedp} when an object whose type
898 is @var{type} and which has the specified mode and signedness is to be
899 stored in a register. This macro is only called when @var{type} is a
902 On most RISC machines, which only have operations that operate on a full
903 register, define this macro to set @var{m} to @code{word_mode} if
904 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
905 cases, only integer modes should be widened because wider-precision
906 floating-point operations are usually more expensive than their narrower
909 For most machines, the macro definition does not change @var{unsignedp}.
910 However, some machines, have instructions that preferentially handle
911 either signed or unsigned quantities of certain modes. For example, on
912 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
913 sign-extend the result to 64 bits. On such machines, set
914 @var{unsignedp} according to which kind of extension is more efficient.
916 Do not define this macro if it would never modify @var{m}.
919 @hook TARGET_PROMOTE_FUNCTION_MODE
921 @defmac PARM_BOUNDARY
922 Normal alignment required for function parameters on the stack, in
923 bits. All stack parameters receive at least this much alignment
924 regardless of data type. On most machines, this is the same as the
928 @defmac STACK_BOUNDARY
929 Define this macro to the minimum alignment enforced by hardware for the
930 stack pointer on this machine. The definition is a C expression for the
931 desired alignment (measured in bits). This value is used as a default
932 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
933 this should be the same as @code{PARM_BOUNDARY}.
936 @defmac PREFERRED_STACK_BOUNDARY
937 Define this macro if you wish to preserve a certain alignment for the
938 stack pointer, greater than what the hardware enforces. The definition
939 is a C expression for the desired alignment (measured in bits). This
940 macro must evaluate to a value equal to or larger than
941 @code{STACK_BOUNDARY}.
944 @defmac INCOMING_STACK_BOUNDARY
945 Define this macro if the incoming stack boundary may be different
946 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
947 to a value equal to or larger than @code{STACK_BOUNDARY}.
950 @defmac FUNCTION_BOUNDARY
951 Alignment required for a function entry point, in bits.
954 @defmac BIGGEST_ALIGNMENT
955 Biggest alignment that any data type can require on this machine, in
956 bits. Note that this is not the biggest alignment that is supported,
957 just the biggest alignment that, when violated, may cause a fault.
960 @defmac MALLOC_ABI_ALIGNMENT
961 Alignment, in bits, a C conformant malloc implementation has to
962 provide. If not defined, the default value is @code{BITS_PER_WORD}.
965 @defmac ATTRIBUTE_ALIGNED_VALUE
966 Alignment used by the @code{__attribute__ ((aligned))} construct. If
967 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
970 @defmac MINIMUM_ATOMIC_ALIGNMENT
971 If defined, the smallest alignment, in bits, that can be given to an
972 object that can be referenced in one operation, without disturbing any
973 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
974 on machines that don't have byte or half-word store operations.
977 @defmac BIGGEST_FIELD_ALIGNMENT
978 Biggest alignment that any structure or union field can require on this
979 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
980 structure and union fields only, unless the field alignment has been set
981 by the @code{__attribute__ ((aligned (@var{n})))} construct.
984 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
985 An expression for the alignment of a structure field @var{field} if the
986 alignment computed in the usual way (including applying of
987 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
988 alignment) is @var{computed}. It overrides alignment only if the
989 field alignment has not been set by the
990 @code{__attribute__ ((aligned (@var{n})))} construct.
993 @defmac MAX_STACK_ALIGNMENT
994 Biggest stack alignment guaranteed by the backend. Use this macro
995 to specify the maximum alignment of a variable on stack.
997 If not defined, the default value is @code{STACK_BOUNDARY}.
999 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1000 @c But the fix for PR 32893 indicates that we can only guarantee
1001 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1002 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1005 @defmac MAX_OFILE_ALIGNMENT
1006 Biggest alignment supported by the object file format of this machine.
1007 Use this macro to limit the alignment which can be specified using the
1008 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1009 the default value is @code{BIGGEST_ALIGNMENT}.
1011 On systems that use ELF, the default (in @file{config/elfos.h}) is
1012 the largest supported 32-bit ELF section alignment representable on
1013 a 32-bit host e.g. @samp{(((uint64_t) 1 << 28) * 8)}.
1014 On 32-bit ELF the largest supported section alignment in bits is
1015 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1018 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1019 If defined, a C expression to compute the alignment for a variable in
1020 the static store. @var{type} is the data type, and @var{basic-align} is
1021 the alignment that the object would ordinarily have. The value of this
1022 macro is used instead of that alignment to align the object.
1024 If this macro is not defined, then @var{basic-align} is used.
1027 One use of this macro is to increase alignment of medium-size data to
1028 make it all fit in fewer cache lines. Another is to cause character
1029 arrays to be word-aligned so that @code{strcpy} calls that copy
1030 constants to character arrays can be done inline.
1033 @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1034 Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1035 some alignment increase, instead of optimization only purposes. E.g.@
1036 AMD x86-64 psABI says that variables with array type larger than 15 bytes
1037 must be aligned to 16 byte boundaries.
1039 If this macro is not defined, then @var{basic-align} is used.
1042 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1043 If defined, a C expression to compute the alignment given to a constant
1044 that is being placed in memory. @var{constant} is the constant and
1045 @var{basic-align} is the alignment that the object would ordinarily
1046 have. The value of this macro is used instead of that alignment to
1049 If this macro is not defined, then @var{basic-align} is used.
1051 The typical use of this macro is to increase alignment for string
1052 constants to be word aligned so that @code{strcpy} calls that copy
1053 constants can be done inline.
1056 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1057 If defined, a C expression to compute the alignment for a variable in
1058 the local store. @var{type} is the data type, and @var{basic-align} is
1059 the alignment that the object would ordinarily have. The value of this
1060 macro is used instead of that alignment to align the object.
1062 If this macro is not defined, then @var{basic-align} is used.
1064 One use of this macro is to increase alignment of medium-size data to
1065 make it all fit in fewer cache lines.
1067 If the value of this macro has a type, it should be an unsigned type.
1070 @hook TARGET_VECTOR_ALIGNMENT
1072 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1073 If defined, a C expression to compute the alignment for stack slot.
1074 @var{type} is the data type, @var{mode} is the widest mode available,
1075 and @var{basic-align} is the alignment that the slot would ordinarily
1076 have. The value of this macro is used instead of that alignment to
1079 If this macro is not defined, then @var{basic-align} is used when
1080 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1083 This macro is to set alignment of stack slot to the maximum alignment
1084 of all possible modes which the slot may have.
1086 If the value of this macro has a type, it should be an unsigned type.
1089 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1090 If defined, a C expression to compute the alignment for a local
1091 variable @var{decl}.
1093 If this macro is not defined, then
1094 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1097 One use of this macro is to increase alignment of medium-size data to
1098 make it all fit in fewer cache lines.
1100 If the value of this macro has a type, it should be an unsigned type.
1103 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1104 If defined, a C expression to compute the minimum required alignment
1105 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1106 @var{mode}, assuming normal alignment @var{align}.
1108 If this macro is not defined, then @var{align} will be used.
1111 @defmac EMPTY_FIELD_BOUNDARY
1112 Alignment in bits to be given to a structure bit-field that follows an
1113 empty field such as @code{int : 0;}.
1115 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1118 @defmac STRUCTURE_SIZE_BOUNDARY
1119 Number of bits which any structure or union's size must be a multiple of.
1120 Each structure or union's size is rounded up to a multiple of this.
1122 If you do not define this macro, the default is the same as
1123 @code{BITS_PER_UNIT}.
1126 @defmac STRICT_ALIGNMENT
1127 Define this macro to be the value 1 if instructions will fail to work
1128 if given data not on the nominal alignment. If instructions will merely
1129 go slower in that case, define this macro as 0.
1132 @defmac PCC_BITFIELD_TYPE_MATTERS
1133 Define this if you wish to imitate the way many other C compilers handle
1134 alignment of bit-fields and the structures that contain them.
1136 The behavior is that the type written for a named bit-field (@code{int},
1137 @code{short}, or other integer type) imposes an alignment for the entire
1138 structure, as if the structure really did contain an ordinary field of
1139 that type. In addition, the bit-field is placed within the structure so
1140 that it would fit within such a field, not crossing a boundary for it.
1142 Thus, on most machines, a named bit-field whose type is written as
1143 @code{int} would not cross a four-byte boundary, and would force
1144 four-byte alignment for the whole structure. (The alignment used may
1145 not be four bytes; it is controlled by the other alignment parameters.)
1147 An unnamed bit-field will not affect the alignment of the containing
1150 If the macro is defined, its definition should be a C expression;
1151 a nonzero value for the expression enables this behavior.
1153 Note that if this macro is not defined, or its value is zero, some
1154 bit-fields may cross more than one alignment boundary. The compiler can
1155 support such references if there are @samp{insv}, @samp{extv}, and
1156 @samp{extzv} insns that can directly reference memory.
1158 The other known way of making bit-fields work is to define
1159 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1160 Then every structure can be accessed with fullwords.
1162 Unless the machine has bit-field instructions or you define
1163 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1164 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1166 If your aim is to make GCC use the same conventions for laying out
1167 bit-fields as are used by another compiler, here is how to investigate
1168 what the other compiler does. Compile and run this program:
1187 printf ("Size of foo1 is %d\n",
1188 sizeof (struct foo1));
1189 printf ("Size of foo2 is %d\n",
1190 sizeof (struct foo2));
1195 If this prints 2 and 5, then the compiler's behavior is what you would
1196 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1199 @defmac BITFIELD_NBYTES_LIMITED
1200 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1201 to aligning a bit-field within the structure.
1204 @hook TARGET_ALIGN_ANON_BITFIELD
1206 @hook TARGET_NARROW_VOLATILE_BITFIELD
1208 @hook TARGET_MEMBER_TYPE_FORCES_BLK
1210 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1211 Define this macro as an expression for the alignment of a type (given
1212 by @var{type} as a tree node) if the alignment computed in the usual
1213 way is @var{computed} and the alignment explicitly specified was
1216 The default is to use @var{specified} if it is larger; otherwise, use
1217 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1220 @defmac MAX_FIXED_MODE_SIZE
1221 An integer expression for the size in bits of the largest integer
1222 machine mode that should actually be used. All integer machine modes of
1223 this size or smaller can be used for structures and unions with the
1224 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1225 (DImode)} is assumed.
1228 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1229 If defined, an expression of type @code{enum machine_mode} that
1230 specifies the mode of the save area operand of a
1231 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1232 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1233 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1234 having its mode specified.
1236 You need not define this macro if it always returns @code{Pmode}. You
1237 would most commonly define this macro if the
1238 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1242 @defmac STACK_SIZE_MODE
1243 If defined, an expression of type @code{enum machine_mode} that
1244 specifies the mode of the size increment operand of an
1245 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1247 You need not define this macro if it always returns @code{word_mode}.
1248 You would most commonly define this macro if the @code{allocate_stack}
1249 pattern needs to support both a 32- and a 64-bit mode.
1252 @hook TARGET_LIBGCC_CMP_RETURN_MODE
1254 @hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1256 @hook TARGET_UNWIND_WORD_MODE
1258 @hook TARGET_MS_BITFIELD_LAYOUT_P
1260 @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1262 @hook TARGET_FIXED_POINT_SUPPORTED_P
1264 @hook TARGET_EXPAND_TO_RTL_HOOK
1266 @hook TARGET_INSTANTIATE_DECLS
1268 @hook TARGET_MANGLE_TYPE
1271 @section Layout of Source Language Data Types
1273 These macros define the sizes and other characteristics of the standard
1274 basic data types used in programs being compiled. Unlike the macros in
1275 the previous section, these apply to specific features of C and related
1276 languages, rather than to fundamental aspects of storage layout.
1278 @defmac INT_TYPE_SIZE
1279 A C expression for the size in bits of the type @code{int} on the
1280 target machine. If you don't define this, the default is one word.
1283 @defmac SHORT_TYPE_SIZE
1284 A C expression for the size in bits of the type @code{short} on the
1285 target machine. If you don't define this, the default is half a word.
1286 (If this would be less than one storage unit, it is rounded up to one
1290 @defmac LONG_TYPE_SIZE
1291 A C expression for the size in bits of the type @code{long} on the
1292 target machine. If you don't define this, the default is one word.
1295 @defmac ADA_LONG_TYPE_SIZE
1296 On some machines, the size used for the Ada equivalent of the type
1297 @code{long} by a native Ada compiler differs from that used by C@. In
1298 that situation, define this macro to be a C expression to be used for
1299 the size of that type. If you don't define this, the default is the
1300 value of @code{LONG_TYPE_SIZE}.
1303 @defmac LONG_LONG_TYPE_SIZE
1304 A C expression for the size in bits of the type @code{long long} on the
1305 target machine. If you don't define this, the default is two
1306 words. If you want to support GNU Ada on your machine, the value of this
1307 macro must be at least 64.
1310 @defmac CHAR_TYPE_SIZE
1311 A C expression for the size in bits of the type @code{char} on the
1312 target machine. If you don't define this, the default is
1313 @code{BITS_PER_UNIT}.
1316 @defmac BOOL_TYPE_SIZE
1317 A C expression for the size in bits of the C++ type @code{bool} and
1318 C99 type @code{_Bool} on the target machine. If you don't define
1319 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1322 @defmac FLOAT_TYPE_SIZE
1323 A C expression for the size in bits of the type @code{float} on the
1324 target machine. If you don't define this, the default is one word.
1327 @defmac DOUBLE_TYPE_SIZE
1328 A C expression for the size in bits of the type @code{double} on the
1329 target machine. If you don't define this, the default is two
1333 @defmac LONG_DOUBLE_TYPE_SIZE
1334 A C expression for the size in bits of the type @code{long double} on
1335 the target machine. If you don't define this, the default is two
1339 @defmac SHORT_FRACT_TYPE_SIZE
1340 A C expression for the size in bits of the type @code{short _Fract} on
1341 the target machine. If you don't define this, the default is
1342 @code{BITS_PER_UNIT}.
1345 @defmac FRACT_TYPE_SIZE
1346 A C expression for the size in bits of the type @code{_Fract} on
1347 the target machine. If you don't define this, the default is
1348 @code{BITS_PER_UNIT * 2}.
1351 @defmac LONG_FRACT_TYPE_SIZE
1352 A C expression for the size in bits of the type @code{long _Fract} on
1353 the target machine. If you don't define this, the default is
1354 @code{BITS_PER_UNIT * 4}.
1357 @defmac LONG_LONG_FRACT_TYPE_SIZE
1358 A C expression for the size in bits of the type @code{long long _Fract} on
1359 the target machine. If you don't define this, the default is
1360 @code{BITS_PER_UNIT * 8}.
1363 @defmac SHORT_ACCUM_TYPE_SIZE
1364 A C expression for the size in bits of the type @code{short _Accum} on
1365 the target machine. If you don't define this, the default is
1366 @code{BITS_PER_UNIT * 2}.
1369 @defmac ACCUM_TYPE_SIZE
1370 A C expression for the size in bits of the type @code{_Accum} on
1371 the target machine. If you don't define this, the default is
1372 @code{BITS_PER_UNIT * 4}.
1375 @defmac LONG_ACCUM_TYPE_SIZE
1376 A C expression for the size in bits of the type @code{long _Accum} on
1377 the target machine. If you don't define this, the default is
1378 @code{BITS_PER_UNIT * 8}.
1381 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1382 A C expression for the size in bits of the type @code{long long _Accum} on
1383 the target machine. If you don't define this, the default is
1384 @code{BITS_PER_UNIT * 16}.
1387 @defmac LIBGCC2_GNU_PREFIX
1388 This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1389 hook and should be defined if that hook is overriden to be true. It
1390 causes function names in libgcc to be changed to use a @code{__gnu_}
1391 prefix for their name rather than the default @code{__}. A port which
1392 uses this macro should also arrange to use @file{t-gnu-prefix} in
1393 the libgcc @file{config.host}.
1396 @defmac TARGET_FLT_EVAL_METHOD
1397 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1398 assuming, if applicable, that the floating-point control word is in its
1399 default state. If you do not define this macro the value of
1400 @code{FLT_EVAL_METHOD} will be zero.
1403 @defmac WIDEST_HARDWARE_FP_SIZE
1404 A C expression for the size in bits of the widest floating-point format
1405 supported by the hardware. If you define this macro, you must specify a
1406 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1407 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1411 @defmac DEFAULT_SIGNED_CHAR
1412 An expression whose value is 1 or 0, according to whether the type
1413 @code{char} should be signed or unsigned by default. The user can
1414 always override this default with the options @option{-fsigned-char}
1415 and @option{-funsigned-char}.
1418 @hook TARGET_DEFAULT_SHORT_ENUMS
1421 A C expression for a string describing the name of the data type to use
1422 for size values. The typedef name @code{size_t} is defined using the
1423 contents of the string.
1425 The string can contain more than one keyword. If so, separate them with
1426 spaces, and write first any length keyword, then @code{unsigned} if
1427 appropriate, and finally @code{int}. The string must exactly match one
1428 of the data type names defined in the function
1429 @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.c}.
1430 You may not omit @code{int} or change the order---that would cause the
1431 compiler to crash on startup.
1433 If you don't define this macro, the default is @code{"long unsigned
1438 GCC defines internal types (@code{sizetype}, @code{ssizetype},
1439 @code{bitsizetype} and @code{sbitsizetype}) for expressions
1440 dealing with size. This macro is a C expression for a string describing
1441 the name of the data type from which the precision of @code{sizetype}
1444 The string has the same restrictions as @code{SIZE_TYPE} string.
1446 If you don't define this macro, the default is @code{SIZE_TYPE}.
1449 @defmac PTRDIFF_TYPE
1450 A C expression for a string describing the name of the data type to use
1451 for the result of subtracting two pointers. The typedef name
1452 @code{ptrdiff_t} is defined using the contents of the string. See
1453 @code{SIZE_TYPE} above for more information.
1455 If you don't define this macro, the default is @code{"long int"}.
1459 A C expression for a string describing the name of the data type to use
1460 for wide characters. The typedef name @code{wchar_t} is defined using
1461 the contents of the string. See @code{SIZE_TYPE} above for more
1464 If you don't define this macro, the default is @code{"int"}.
1467 @defmac WCHAR_TYPE_SIZE
1468 A C expression for the size in bits of the data type for wide
1469 characters. This is used in @code{cpp}, which cannot make use of
1474 A C expression for a string describing the name of the data type to
1475 use for wide characters passed to @code{printf} and returned from
1476 @code{getwc}. The typedef name @code{wint_t} is defined using the
1477 contents of the string. See @code{SIZE_TYPE} above for more
1480 If you don't define this macro, the default is @code{"unsigned int"}.
1484 A C expression for a string describing the name of the data type that
1485 can represent any value of any standard or extended signed integer type.
1486 The typedef name @code{intmax_t} is defined using the contents of the
1487 string. See @code{SIZE_TYPE} above for more information.
1489 If you don't define this macro, the default is the first of
1490 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1491 much precision as @code{long long int}.
1494 @defmac UINTMAX_TYPE
1495 A C expression for a string describing the name of the data type that
1496 can represent any value of any standard or extended unsigned integer
1497 type. The typedef name @code{uintmax_t} is defined using the contents
1498 of the string. See @code{SIZE_TYPE} above for more information.
1500 If you don't define this macro, the default is the first of
1501 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1502 unsigned int"} that has as much precision as @code{long long unsigned
1506 @defmac SIG_ATOMIC_TYPE
1512 @defmacx UINT16_TYPE
1513 @defmacx UINT32_TYPE
1514 @defmacx UINT64_TYPE
1515 @defmacx INT_LEAST8_TYPE
1516 @defmacx INT_LEAST16_TYPE
1517 @defmacx INT_LEAST32_TYPE
1518 @defmacx INT_LEAST64_TYPE
1519 @defmacx UINT_LEAST8_TYPE
1520 @defmacx UINT_LEAST16_TYPE
1521 @defmacx UINT_LEAST32_TYPE
1522 @defmacx UINT_LEAST64_TYPE
1523 @defmacx INT_FAST8_TYPE
1524 @defmacx INT_FAST16_TYPE
1525 @defmacx INT_FAST32_TYPE
1526 @defmacx INT_FAST64_TYPE
1527 @defmacx UINT_FAST8_TYPE
1528 @defmacx UINT_FAST16_TYPE
1529 @defmacx UINT_FAST32_TYPE
1530 @defmacx UINT_FAST64_TYPE
1531 @defmacx INTPTR_TYPE
1532 @defmacx UINTPTR_TYPE
1533 C expressions for the standard types @code{sig_atomic_t},
1534 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1535 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1536 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1537 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1538 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1539 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1540 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1541 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1542 @code{SIZE_TYPE} above for more information.
1544 If any of these macros evaluates to a null pointer, the corresponding
1545 type is not supported; if GCC is configured to provide
1546 @code{<stdint.h>} in such a case, the header provided may not conform
1547 to C99, depending on the type in question. The defaults for all of
1548 these macros are null pointers.
1551 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1552 The C++ compiler represents a pointer-to-member-function with a struct
1559 ptrdiff_t vtable_index;
1566 The C++ compiler must use one bit to indicate whether the function that
1567 will be called through a pointer-to-member-function is virtual.
1568 Normally, we assume that the low-order bit of a function pointer must
1569 always be zero. Then, by ensuring that the vtable_index is odd, we can
1570 distinguish which variant of the union is in use. But, on some
1571 platforms function pointers can be odd, and so this doesn't work. In
1572 that case, we use the low-order bit of the @code{delta} field, and shift
1573 the remainder of the @code{delta} field to the left.
1575 GCC will automatically make the right selection about where to store
1576 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1577 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1578 set such that functions always start at even addresses, but the lowest
1579 bit of pointers to functions indicate whether the function at that
1580 address is in ARM or Thumb mode. If this is the case of your
1581 architecture, you should define this macro to
1582 @code{ptrmemfunc_vbit_in_delta}.
1584 In general, you should not have to define this macro. On architectures
1585 in which function addresses are always even, according to
1586 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1587 @code{ptrmemfunc_vbit_in_pfn}.
1590 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1591 Normally, the C++ compiler uses function pointers in vtables. This
1592 macro allows the target to change to use ``function descriptors''
1593 instead. Function descriptors are found on targets for whom a
1594 function pointer is actually a small data structure. Normally the
1595 data structure consists of the actual code address plus a data
1596 pointer to which the function's data is relative.
1598 If vtables are used, the value of this macro should be the number
1599 of words that the function descriptor occupies.
1602 @defmac TARGET_VTABLE_ENTRY_ALIGN
1603 By default, the vtable entries are void pointers, the so the alignment
1604 is the same as pointer alignment. The value of this macro specifies
1605 the alignment of the vtable entry in bits. It should be defined only
1606 when special alignment is necessary. */
1609 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1610 There are a few non-descriptor entries in the vtable at offsets below
1611 zero. If these entries must be padded (say, to preserve the alignment
1612 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1613 of words in each data entry.
1617 @section Register Usage
1618 @cindex register usage
1620 This section explains how to describe what registers the target machine
1621 has, and how (in general) they can be used.
1623 The description of which registers a specific instruction can use is
1624 done with register classes; see @ref{Register Classes}. For information
1625 on using registers to access a stack frame, see @ref{Frame Registers}.
1626 For passing values in registers, see @ref{Register Arguments}.
1627 For returning values in registers, see @ref{Scalar Return}.
1630 * Register Basics:: Number and kinds of registers.
1631 * Allocation Order:: Order in which registers are allocated.
1632 * Values in Registers:: What kinds of values each reg can hold.
1633 * Leaf Functions:: Renumbering registers for leaf functions.
1634 * Stack Registers:: Handling a register stack such as 80387.
1637 @node Register Basics
1638 @subsection Basic Characteristics of Registers
1640 @c prevent bad page break with this line
1641 Registers have various characteristics.
1643 @defmac FIRST_PSEUDO_REGISTER
1644 Number of hardware registers known to the compiler. They receive
1645 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1646 pseudo register's number really is assigned the number
1647 @code{FIRST_PSEUDO_REGISTER}.
1650 @defmac FIXED_REGISTERS
1651 @cindex fixed register
1652 An initializer that says which registers are used for fixed purposes
1653 all throughout the compiled code and are therefore not available for
1654 general allocation. These would include the stack pointer, the frame
1655 pointer (except on machines where that can be used as a general
1656 register when no frame pointer is needed), the program counter on
1657 machines where that is considered one of the addressable registers,
1658 and any other numbered register with a standard use.
1660 This information is expressed as a sequence of numbers, separated by
1661 commas and surrounded by braces. The @var{n}th number is 1 if
1662 register @var{n} is fixed, 0 otherwise.
1664 The table initialized from this macro, and the table initialized by
1665 the following one, may be overridden at run time either automatically,
1666 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1667 the user with the command options @option{-ffixed-@var{reg}},
1668 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1671 @defmac CALL_USED_REGISTERS
1672 @cindex call-used register
1673 @cindex call-clobbered register
1674 @cindex call-saved register
1675 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1676 clobbered (in general) by function calls as well as for fixed
1677 registers. This macro therefore identifies the registers that are not
1678 available for general allocation of values that must live across
1681 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1682 automatically saves it on function entry and restores it on function
1683 exit, if the register is used within the function.
1686 @defmac CALL_REALLY_USED_REGISTERS
1687 @cindex call-used register
1688 @cindex call-clobbered register
1689 @cindex call-saved register
1690 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1691 that the entire set of @code{FIXED_REGISTERS} be included.
1692 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1693 This macro is optional. If not specified, it defaults to the value
1694 of @code{CALL_USED_REGISTERS}.
1697 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1698 @cindex call-used register
1699 @cindex call-clobbered register
1700 @cindex call-saved register
1701 A C expression that is nonzero if it is not permissible to store a
1702 value of mode @var{mode} in hard register number @var{regno} across a
1703 call without some part of it being clobbered. For most machines this
1704 macro need not be defined. It is only required for machines that do not
1705 preserve the entire contents of a register across a call.
1709 @findex call_used_regs
1712 @findex reg_class_contents
1713 @hook TARGET_CONDITIONAL_REGISTER_USAGE
1715 @defmac INCOMING_REGNO (@var{out})
1716 Define this macro if the target machine has register windows. This C
1717 expression returns the register number as seen by the called function
1718 corresponding to the register number @var{out} as seen by the calling
1719 function. Return @var{out} if register number @var{out} is not an
1723 @defmac OUTGOING_REGNO (@var{in})
1724 Define this macro if the target machine has register windows. This C
1725 expression returns the register number as seen by the calling function
1726 corresponding to the register number @var{in} as seen by the called
1727 function. Return @var{in} if register number @var{in} is not an inbound
1731 @defmac LOCAL_REGNO (@var{regno})
1732 Define this macro if the target machine has register windows. This C
1733 expression returns true if the register is call-saved but is in the
1734 register window. Unlike most call-saved registers, such registers
1735 need not be explicitly restored on function exit or during non-local
1740 If the program counter has a register number, define this as that
1741 register number. Otherwise, do not define it.
1744 @node Allocation Order
1745 @subsection Order of Allocation of Registers
1746 @cindex order of register allocation
1747 @cindex register allocation order
1749 @c prevent bad page break with this line
1750 Registers are allocated in order.
1752 @defmac REG_ALLOC_ORDER
1753 If defined, an initializer for a vector of integers, containing the
1754 numbers of hard registers in the order in which GCC should prefer
1755 to use them (from most preferred to least).
1757 If this macro is not defined, registers are used lowest numbered first
1758 (all else being equal).
1760 One use of this macro is on machines where the highest numbered
1761 registers must always be saved and the save-multiple-registers
1762 instruction supports only sequences of consecutive registers. On such
1763 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1764 the highest numbered allocable register first.
1767 @defmac ADJUST_REG_ALLOC_ORDER
1768 A C statement (sans semicolon) to choose the order in which to allocate
1769 hard registers for pseudo-registers local to a basic block.
1771 Store the desired register order in the array @code{reg_alloc_order}.
1772 Element 0 should be the register to allocate first; element 1, the next
1773 register; and so on.
1775 The macro body should not assume anything about the contents of
1776 @code{reg_alloc_order} before execution of the macro.
1778 On most machines, it is not necessary to define this macro.
1781 @defmac HONOR_REG_ALLOC_ORDER
1782 Normally, IRA tries to estimate the costs for saving a register in the
1783 prologue and restoring it in the epilogue. This discourages it from
1784 using call-saved registers. If a machine wants to ensure that IRA
1785 allocates registers in the order given by REG_ALLOC_ORDER even if some
1786 call-saved registers appear earlier than call-used ones, then define this
1787 macro as a C expression to nonzero. Default is 0.
1790 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
1791 In some case register allocation order is not enough for the
1792 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
1793 If this macro is defined, it should return a floating point value
1794 based on @var{regno}. The cost of using @var{regno} for a pseudo will
1795 be increased by approximately the pseudo's usage frequency times the
1796 value returned by this macro. Not defining this macro is equivalent
1797 to having it always return @code{0.0}.
1799 On most machines, it is not necessary to define this macro.
1802 @node Values in Registers
1803 @subsection How Values Fit in Registers
1805 This section discusses the macros that describe which kinds of values
1806 (specifically, which machine modes) each register can hold, and how many
1807 consecutive registers are needed for a given mode.
1809 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
1810 A C expression for the number of consecutive hard registers, starting
1811 at register number @var{regno}, required to hold a value of mode
1812 @var{mode}. This macro must never return zero, even if a register
1813 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
1814 and/or CANNOT_CHANGE_MODE_CLASS instead.
1816 On a machine where all registers are exactly one word, a suitable
1817 definition of this macro is
1820 #define HARD_REGNO_NREGS(REGNO, MODE) \
1821 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
1826 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
1827 A C expression that is nonzero if a value of mode @var{mode}, stored
1828 in memory, ends with padding that causes it to take up more space than
1829 in registers starting at register number @var{regno} (as determined by
1830 multiplying GCC's notion of the size of the register when containing
1831 this mode by the number of registers returned by
1832 @code{HARD_REGNO_NREGS}). By default this is zero.
1834 For example, if a floating-point value is stored in three 32-bit
1835 registers but takes up 128 bits in memory, then this would be
1838 This macros only needs to be defined if there are cases where
1839 @code{subreg_get_info}
1840 would otherwise wrongly determine that a @code{subreg} can be
1841 represented by an offset to the register number, when in fact such a
1842 @code{subreg} would contain some of the padding not stored in
1843 registers and so not be representable.
1846 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
1847 For values of @var{regno} and @var{mode} for which
1848 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
1849 returning the greater number of registers required to hold the value
1850 including any padding. In the example above, the value would be four.
1853 @defmac REGMODE_NATURAL_SIZE (@var{mode})
1854 Define this macro if the natural size of registers that hold values
1855 of mode @var{mode} is not the word size. It is a C expression that
1856 should give the natural size in bytes for the specified mode. It is
1857 used by the register allocator to try to optimize its results. This
1858 happens for example on SPARC 64-bit where the natural size of
1859 floating-point registers is still 32-bit.
1862 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
1863 A C expression that is nonzero if it is permissible to store a value
1864 of mode @var{mode} in hard register number @var{regno} (or in several
1865 registers starting with that one). For a machine where all registers
1866 are equivalent, a suitable definition is
1869 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
1872 You need not include code to check for the numbers of fixed registers,
1873 because the allocation mechanism considers them to be always occupied.
1875 @cindex register pairs
1876 On some machines, double-precision values must be kept in even/odd
1877 register pairs. You can implement that by defining this macro to reject
1878 odd register numbers for such modes.
1880 The minimum requirement for a mode to be OK in a register is that the
1881 @samp{mov@var{mode}} instruction pattern support moves between the
1882 register and other hard register in the same class and that moving a
1883 value into the register and back out not alter it.
1885 Since the same instruction used to move @code{word_mode} will work for
1886 all narrower integer modes, it is not necessary on any machine for
1887 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
1888 you define patterns @samp{movhi}, etc., to take advantage of this. This
1889 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
1890 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
1893 Many machines have special registers for floating point arithmetic.
1894 Often people assume that floating point machine modes are allowed only
1895 in floating point registers. This is not true. Any registers that
1896 can hold integers can safely @emph{hold} a floating point machine
1897 mode, whether or not floating arithmetic can be done on it in those
1898 registers. Integer move instructions can be used to move the values.
1900 On some machines, though, the converse is true: fixed-point machine
1901 modes may not go in floating registers. This is true if the floating
1902 registers normalize any value stored in them, because storing a
1903 non-floating value there would garble it. In this case,
1904 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
1905 floating registers. But if the floating registers do not automatically
1906 normalize, if you can store any bit pattern in one and retrieve it
1907 unchanged without a trap, then any machine mode may go in a floating
1908 register, so you can define this macro to say so.
1910 The primary significance of special floating registers is rather that
1911 they are the registers acceptable in floating point arithmetic
1912 instructions. However, this is of no concern to
1913 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
1914 constraints for those instructions.
1916 On some machines, the floating registers are especially slow to access,
1917 so that it is better to store a value in a stack frame than in such a
1918 register if floating point arithmetic is not being done. As long as the
1919 floating registers are not in class @code{GENERAL_REGS}, they will not
1920 be used unless some pattern's constraint asks for one.
1923 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
1924 A C expression that is nonzero if it is OK to rename a hard register
1925 @var{from} to another hard register @var{to}.
1927 One common use of this macro is to prevent renaming of a register to
1928 another register that is not saved by a prologue in an interrupt
1931 The default is always nonzero.
1934 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
1935 A C expression that is nonzero if a value of mode
1936 @var{mode1} is accessible in mode @var{mode2} without copying.
1938 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
1939 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
1940 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
1941 should be nonzero. If they differ for any @var{r}, you should define
1942 this macro to return zero unless some other mechanism ensures the
1943 accessibility of the value in a narrower mode.
1945 You should define this macro to return nonzero in as many cases as
1946 possible since doing so will allow GCC to perform better register
1950 @hook TARGET_HARD_REGNO_SCRATCH_OK
1952 @defmac AVOID_CCMODE_COPIES
1953 Define this macro if the compiler should avoid copies to/from @code{CCmode}
1954 registers. You should only define this macro if support for copying to/from
1955 @code{CCmode} is incomplete.
1958 @node Leaf Functions
1959 @subsection Handling Leaf Functions
1961 @cindex leaf functions
1962 @cindex functions, leaf
1963 On some machines, a leaf function (i.e., one which makes no calls) can run
1964 more efficiently if it does not make its own register window. Often this
1965 means it is required to receive its arguments in the registers where they
1966 are passed by the caller, instead of the registers where they would
1969 The special treatment for leaf functions generally applies only when
1970 other conditions are met; for example, often they may use only those
1971 registers for its own variables and temporaries. We use the term ``leaf
1972 function'' to mean a function that is suitable for this special
1973 handling, so that functions with no calls are not necessarily ``leaf
1976 GCC assigns register numbers before it knows whether the function is
1977 suitable for leaf function treatment. So it needs to renumber the
1978 registers in order to output a leaf function. The following macros
1981 @defmac LEAF_REGISTERS
1982 Name of a char vector, indexed by hard register number, which
1983 contains 1 for a register that is allowable in a candidate for leaf
1986 If leaf function treatment involves renumbering the registers, then the
1987 registers marked here should be the ones before renumbering---those that
1988 GCC would ordinarily allocate. The registers which will actually be
1989 used in the assembler code, after renumbering, should not be marked with 1
1992 Define this macro only if the target machine offers a way to optimize
1993 the treatment of leaf functions.
1996 @defmac LEAF_REG_REMAP (@var{regno})
1997 A C expression whose value is the register number to which @var{regno}
1998 should be renumbered, when a function is treated as a leaf function.
2000 If @var{regno} is a register number which should not appear in a leaf
2001 function before renumbering, then the expression should yield @minus{}1, which
2002 will cause the compiler to abort.
2004 Define this macro only if the target machine offers a way to optimize the
2005 treatment of leaf functions, and registers need to be renumbered to do
2009 @findex current_function_is_leaf
2010 @findex current_function_uses_only_leaf_regs
2011 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2012 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2013 specially. They can test the C variable @code{current_function_is_leaf}
2014 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2015 set prior to local register allocation and is valid for the remaining
2016 compiler passes. They can also test the C variable
2017 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2018 functions which only use leaf registers.
2019 @code{current_function_uses_only_leaf_regs} is valid after all passes
2020 that modify the instructions have been run and is only useful if
2021 @code{LEAF_REGISTERS} is defined.
2022 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2023 @c of the next paragraph?! --mew 2feb93
2025 @node Stack Registers
2026 @subsection Registers That Form a Stack
2028 There are special features to handle computers where some of the
2029 ``registers'' form a stack. Stack registers are normally written by
2030 pushing onto the stack, and are numbered relative to the top of the
2033 Currently, GCC can only handle one group of stack-like registers, and
2034 they must be consecutively numbered. Furthermore, the existing
2035 support for stack-like registers is specific to the 80387 floating
2036 point coprocessor. If you have a new architecture that uses
2037 stack-like registers, you will need to do substantial work on
2038 @file{reg-stack.c} and write your machine description to cooperate
2039 with it, as well as defining these macros.
2042 Define this if the machine has any stack-like registers.
2045 @defmac STACK_REG_COVER_CLASS
2046 This is a cover class containing the stack registers. Define this if
2047 the machine has any stack-like registers.
2050 @defmac FIRST_STACK_REG
2051 The number of the first stack-like register. This one is the top
2055 @defmac LAST_STACK_REG
2056 The number of the last stack-like register. This one is the bottom of
2060 @node Register Classes
2061 @section Register Classes
2062 @cindex register class definitions
2063 @cindex class definitions, register
2065 On many machines, the numbered registers are not all equivalent.
2066 For example, certain registers may not be allowed for indexed addressing;
2067 certain registers may not be allowed in some instructions. These machine
2068 restrictions are described to the compiler using @dfn{register classes}.
2070 You define a number of register classes, giving each one a name and saying
2071 which of the registers belong to it. Then you can specify register classes
2072 that are allowed as operands to particular instruction patterns.
2076 In general, each register will belong to several classes. In fact, one
2077 class must be named @code{ALL_REGS} and contain all the registers. Another
2078 class must be named @code{NO_REGS} and contain no registers. Often the
2079 union of two classes will be another class; however, this is not required.
2081 @findex GENERAL_REGS
2082 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2083 terribly special about the name, but the operand constraint letters
2084 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2085 the same as @code{ALL_REGS}, just define it as a macro which expands
2088 Order the classes so that if class @var{x} is contained in class @var{y}
2089 then @var{x} has a lower class number than @var{y}.
2091 The way classes other than @code{GENERAL_REGS} are specified in operand
2092 constraints is through machine-dependent operand constraint letters.
2093 You can define such letters to correspond to various classes, then use
2094 them in operand constraints.
2096 You must define the narrowest register classes for allocatable
2097 registers, so that each class either has no subclasses, or that for
2098 some mode, the move cost between registers within the class is
2099 cheaper than moving a register in the class to or from memory
2102 You should define a class for the union of two classes whenever some
2103 instruction allows both classes. For example, if an instruction allows
2104 either a floating point (coprocessor) register or a general register for a
2105 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2106 which includes both of them. Otherwise you will get suboptimal code,
2107 or even internal compiler errors when reload cannot find a register in the
2108 class computed via @code{reg_class_subunion}.
2110 You must also specify certain redundant information about the register
2111 classes: for each class, which classes contain it and which ones are
2112 contained in it; for each pair of classes, the largest class contained
2115 When a value occupying several consecutive registers is expected in a
2116 certain class, all the registers used must belong to that class.
2117 Therefore, register classes cannot be used to enforce a requirement for
2118 a register pair to start with an even-numbered register. The way to
2119 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2121 Register classes used for input-operands of bitwise-and or shift
2122 instructions have a special requirement: each such class must have, for
2123 each fixed-point machine mode, a subclass whose registers can transfer that
2124 mode to or from memory. For example, on some machines, the operations for
2125 single-byte values (@code{QImode}) are limited to certain registers. When
2126 this is so, each register class that is used in a bitwise-and or shift
2127 instruction must have a subclass consisting of registers from which
2128 single-byte values can be loaded or stored. This is so that
2129 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2131 @deftp {Data type} {enum reg_class}
2132 An enumerated type that must be defined with all the register class names
2133 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2134 must be the last register class, followed by one more enumerated value,
2135 @code{LIM_REG_CLASSES}, which is not a register class but rather
2136 tells how many classes there are.
2138 Each register class has a number, which is the value of casting
2139 the class name to type @code{int}. The number serves as an index
2140 in many of the tables described below.
2143 @defmac N_REG_CLASSES
2144 The number of distinct register classes, defined as follows:
2147 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2151 @defmac REG_CLASS_NAMES
2152 An initializer containing the names of the register classes as C string
2153 constants. These names are used in writing some of the debugging dumps.
2156 @defmac REG_CLASS_CONTENTS
2157 An initializer containing the contents of the register classes, as integers
2158 which are bit masks. The @var{n}th integer specifies the contents of class
2159 @var{n}. The way the integer @var{mask} is interpreted is that
2160 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2162 When the machine has more than 32 registers, an integer does not suffice.
2163 Then the integers are replaced by sub-initializers, braced groupings containing
2164 several integers. Each sub-initializer must be suitable as an initializer
2165 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2166 In this situation, the first integer in each sub-initializer corresponds to
2167 registers 0 through 31, the second integer to registers 32 through 63, and
2171 @defmac REGNO_REG_CLASS (@var{regno})
2172 A C expression whose value is a register class containing hard register
2173 @var{regno}. In general there is more than one such class; choose a class
2174 which is @dfn{minimal}, meaning that no smaller class also contains the
2178 @defmac BASE_REG_CLASS
2179 A macro whose definition is the name of the class to which a valid
2180 base register must belong. A base register is one used in an address
2181 which is the register value plus a displacement.
2184 @defmac MODE_BASE_REG_CLASS (@var{mode})
2185 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2186 the selection of a base register in a mode dependent manner. If
2187 @var{mode} is VOIDmode then it should return the same value as
2188 @code{BASE_REG_CLASS}.
2191 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2192 A C expression whose value is the register class to which a valid
2193 base register must belong in order to be used in a base plus index
2194 register address. You should define this macro if base plus index
2195 addresses have different requirements than other base register uses.
2198 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2199 A C expression whose value is the register class to which a valid
2200 base register for a memory reference in mode @var{mode} to address
2201 space @var{address_space} must belong. @var{outer_code} and @var{index_code}
2202 define the context in which the base register occurs. @var{outer_code} is
2203 the code of the immediately enclosing expression (@code{MEM} for the top level
2204 of an address, @code{ADDRESS} for something that occurs in an
2205 @code{address_operand}). @var{index_code} is the code of the corresponding
2206 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2209 @defmac INDEX_REG_CLASS
2210 A macro whose definition is the name of the class to which a valid
2211 index register must belong. An index register is one used in an
2212 address where its value is either multiplied by a scale factor or
2213 added to another register (as well as added to a displacement).
2216 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2217 A C expression which is nonzero if register number @var{num} is
2218 suitable for use as a base register in operand addresses.
2221 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2222 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2223 that expression may examine the mode of the memory reference in
2224 @var{mode}. You should define this macro if the mode of the memory
2225 reference affects whether a register may be used as a base register. If
2226 you define this macro, the compiler will use it instead of
2227 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2228 addresses that appear outside a @code{MEM}, i.e., as an
2229 @code{address_operand}.
2232 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2233 A C expression which is nonzero if register number @var{num} is suitable for
2234 use as a base register in base plus index operand addresses, accessing
2235 memory in mode @var{mode}. It may be either a suitable hard register or a
2236 pseudo register that has been allocated such a hard register. You should
2237 define this macro if base plus index addresses have different requirements
2238 than other base register uses.
2240 Use of this macro is deprecated; please use the more general
2241 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2244 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2245 A C expression which is nonzero if register number @var{num} is
2246 suitable for use as a base register in operand addresses, accessing
2247 memory in mode @var{mode} in address space @var{address_space}.
2248 This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2249 that that expression may examine the context in which the register
2250 appears in the memory reference. @var{outer_code} is the code of the
2251 immediately enclosing expression (@code{MEM} if at the top level of the
2252 address, @code{ADDRESS} for something that occurs in an
2253 @code{address_operand}). @var{index_code} is the code of the
2254 corresponding index expression if @var{outer_code} is @code{PLUS};
2255 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2256 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2259 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2260 A C expression which is nonzero if register number @var{num} is
2261 suitable for use as an index register in operand addresses. It may be
2262 either a suitable hard register or a pseudo register that has been
2263 allocated such a hard register.
2265 The difference between an index register and a base register is that
2266 the index register may be scaled. If an address involves the sum of
2267 two registers, neither one of them scaled, then either one may be
2268 labeled the ``base'' and the other the ``index''; but whichever
2269 labeling is used must fit the machine's constraints of which registers
2270 may serve in each capacity. The compiler will try both labelings,
2271 looking for one that is valid, and will reload one or both registers
2272 only if neither labeling works.
2275 @hook TARGET_PREFERRED_RENAME_CLASS
2277 @hook TARGET_PREFERRED_RELOAD_CLASS
2279 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2280 A C expression that places additional restrictions on the register class
2281 to use when it is necessary to copy value @var{x} into a register in class
2282 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2283 another, smaller class. On many machines, the following definition is
2287 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2290 Sometimes returning a more restrictive class makes better code. For
2291 example, on the 68000, when @var{x} is an integer constant that is in range
2292 for a @samp{moveq} instruction, the value of this macro is always
2293 @code{DATA_REGS} as long as @var{class} includes the data registers.
2294 Requiring a data register guarantees that a @samp{moveq} will be used.
2296 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2297 @var{class} is if @var{x} is a legitimate constant which cannot be
2298 loaded into some register class. By returning @code{NO_REGS} you can
2299 force @var{x} into a memory location. For example, rs6000 can load
2300 immediate values into general-purpose registers, but does not have an
2301 instruction for loading an immediate value into a floating-point
2302 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2303 @var{x} is a floating-point constant. If the constant can't be loaded
2304 into any kind of register, code generation will be better if
2305 @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2306 of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2308 If an insn has pseudos in it after register allocation, reload will go
2309 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2310 to find the best one. Returning @code{NO_REGS}, in this case, makes
2311 reload add a @code{!} in front of the constraint: the x86 back-end uses
2312 this feature to discourage usage of 387 registers when math is done in
2313 the SSE registers (and vice versa).
2316 @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2318 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2319 A C expression that places additional restrictions on the register class
2320 to use when it is necessary to be able to hold a value of mode
2321 @var{mode} in a reload register for which class @var{class} would
2324 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2325 there are certain modes that simply can't go in certain reload classes.
2327 The value is a register class; perhaps @var{class}, or perhaps another,
2330 Don't define this macro unless the target machine has limitations which
2331 require the macro to do something nontrivial.
2334 @hook TARGET_SECONDARY_RELOAD
2336 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2337 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2338 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2339 These macros are obsolete, new ports should use the target hook
2340 @code{TARGET_SECONDARY_RELOAD} instead.
2342 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2343 target hook. Older ports still define these macros to indicate to the
2344 reload phase that it may
2345 need to allocate at least one register for a reload in addition to the
2346 register to contain the data. Specifically, if copying @var{x} to a
2347 register @var{class} in @var{mode} requires an intermediate register,
2348 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2349 largest register class all of whose registers can be used as
2350 intermediate registers or scratch registers.
2352 If copying a register @var{class} in @var{mode} to @var{x} requires an
2353 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2354 was supposed to be defined be defined to return the largest register
2355 class required. If the
2356 requirements for input and output reloads were the same, the macro
2357 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2360 The values returned by these macros are often @code{GENERAL_REGS}.
2361 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2362 can be directly copied to or from a register of @var{class} in
2363 @var{mode} without requiring a scratch register. Do not define this
2364 macro if it would always return @code{NO_REGS}.
2366 If a scratch register is required (either with or without an
2367 intermediate register), you were supposed to define patterns for
2368 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2369 (@pxref{Standard Names}. These patterns, which were normally
2370 implemented with a @code{define_expand}, should be similar to the
2371 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2374 These patterns need constraints for the reload register and scratch
2376 contain a single register class. If the original reload register (whose
2377 class is @var{class}) can meet the constraint given in the pattern, the
2378 value returned by these macros is used for the class of the scratch
2379 register. Otherwise, two additional reload registers are required.
2380 Their classes are obtained from the constraints in the insn pattern.
2382 @var{x} might be a pseudo-register or a @code{subreg} of a
2383 pseudo-register, which could either be in a hard register or in memory.
2384 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2385 in memory and the hard register number if it is in a register.
2387 These macros should not be used in the case where a particular class of
2388 registers can only be copied to memory and not to another class of
2389 registers. In that case, secondary reload registers are not needed and
2390 would not be helpful. Instead, a stack location must be used to perform
2391 the copy and the @code{mov@var{m}} pattern should use memory as an
2392 intermediate storage. This case often occurs between floating-point and
2396 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2397 Certain machines have the property that some registers cannot be copied
2398 to some other registers without using memory. Define this macro on
2399 those machines to be a C expression that is nonzero if objects of mode
2400 @var{m} in registers of @var{class1} can only be copied to registers of
2401 class @var{class2} by storing a register of @var{class1} into memory
2402 and loading that memory location into a register of @var{class2}.
2404 Do not define this macro if its value would always be zero.
2407 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2408 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2409 allocates a stack slot for a memory location needed for register copies.
2410 If this macro is defined, the compiler instead uses the memory location
2411 defined by this macro.
2413 Do not define this macro if you do not define
2414 @code{SECONDARY_MEMORY_NEEDED}.
2417 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2418 When the compiler needs a secondary memory location to copy between two
2419 registers of mode @var{mode}, it normally allocates sufficient memory to
2420 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2421 load operations in a mode that many bits wide and whose class is the
2422 same as that of @var{mode}.
2424 This is right thing to do on most machines because it ensures that all
2425 bits of the register are copied and prevents accesses to the registers
2426 in a narrower mode, which some machines prohibit for floating-point
2429 However, this default behavior is not correct on some machines, such as
2430 the DEC Alpha, that store short integers in floating-point registers
2431 differently than in integer registers. On those machines, the default
2432 widening will not work correctly and you must define this macro to
2433 suppress that widening in some cases. See the file @file{alpha.h} for
2436 Do not define this macro if you do not define
2437 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2438 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2441 @hook TARGET_CLASS_LIKELY_SPILLED_P
2443 @hook TARGET_CLASS_MAX_NREGS
2445 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2446 A C expression for the maximum number of consecutive registers
2447 of class @var{class} needed to hold a value of mode @var{mode}.
2449 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2450 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2451 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2452 @var{mode})} for all @var{regno} values in the class @var{class}.
2454 This macro helps control the handling of multiple-word values
2458 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2459 If defined, a C expression that returns nonzero for a @var{class} for which
2460 a change from mode @var{from} to mode @var{to} is invalid.
2462 For the example, loading 32-bit integer or floating-point objects into
2463 floating-point registers on the Alpha extends them to 64 bits.
2464 Therefore loading a 64-bit object and then storing it as a 32-bit object
2465 does not store the low-order 32 bits, as would be the case for a normal
2466 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2470 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2471 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2472 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2478 @hook TARGET_REGISTER_PRIORITY
2480 @hook TARGET_REGISTER_USAGE_LEVELING_P
2482 @hook TARGET_DIFFERENT_ADDR_DISPLACEMENT_P
2484 @hook TARGET_SPILL_CLASS
2486 @hook TARGET_CSTORE_MODE
2488 @node Stack and Calling
2489 @section Stack Layout and Calling Conventions
2490 @cindex calling conventions
2492 @c prevent bad page break with this line
2493 This describes the stack layout and calling conventions.
2497 * Exception Handling::
2502 * Register Arguments::
2504 * Aggregate Return::
2509 * Stack Smashing Protection::
2510 * Miscellaneous Register Hooks::
2514 @subsection Basic Stack Layout
2515 @cindex stack frame layout
2516 @cindex frame layout
2518 @c prevent bad page break with this line
2519 Here is the basic stack layout.
2521 @defmac STACK_GROWS_DOWNWARD
2522 Define this macro if pushing a word onto the stack moves the stack
2523 pointer to a smaller address.
2525 When we say, ``define this macro if @dots{}'', it means that the
2526 compiler checks this macro only with @code{#ifdef} so the precise
2527 definition used does not matter.
2530 @defmac STACK_PUSH_CODE
2531 This macro defines the operation used when something is pushed
2532 on the stack. In RTL, a push operation will be
2533 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2535 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2536 and @code{POST_INC}. Which of these is correct depends on
2537 the stack direction and on whether the stack pointer points
2538 to the last item on the stack or whether it points to the
2539 space for the next item on the stack.
2541 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2542 defined, which is almost always right, and @code{PRE_INC} otherwise,
2543 which is often wrong.
2546 @defmac FRAME_GROWS_DOWNWARD
2547 Define this macro to nonzero value if the addresses of local variable slots
2548 are at negative offsets from the frame pointer.
2551 @defmac ARGS_GROW_DOWNWARD
2552 Define this macro if successive arguments to a function occupy decreasing
2553 addresses on the stack.
2556 @defmac STARTING_FRAME_OFFSET
2557 Offset from the frame pointer to the first local variable slot to be allocated.
2559 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
2560 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
2561 Otherwise, it is found by adding the length of the first slot to the
2562 value @code{STARTING_FRAME_OFFSET}.
2563 @c i'm not sure if the above is still correct.. had to change it to get
2564 @c rid of an overfull. --mew 2feb93
2567 @defmac STACK_ALIGNMENT_NEEDED
2568 Define to zero to disable final alignment of the stack during reload.
2569 The nonzero default for this macro is suitable for most ports.
2571 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
2572 is a register save block following the local block that doesn't require
2573 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2574 stack alignment and do it in the backend.
2577 @defmac STACK_POINTER_OFFSET
2578 Offset from the stack pointer register to the first location at which
2579 outgoing arguments are placed. If not specified, the default value of
2580 zero is used. This is the proper value for most machines.
2582 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2583 the first location at which outgoing arguments are placed.
2586 @defmac FIRST_PARM_OFFSET (@var{fundecl})
2587 Offset from the argument pointer register to the first argument's
2588 address. On some machines it may depend on the data type of the
2591 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2592 the first argument's address.
2595 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2596 Offset from the stack pointer register to an item dynamically allocated
2597 on the stack, e.g., by @code{alloca}.
2599 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2600 length of the outgoing arguments. The default is correct for most
2601 machines. See @file{function.c} for details.
2604 @defmac INITIAL_FRAME_ADDRESS_RTX
2605 A C expression whose value is RTL representing the address of the initial
2606 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2607 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
2608 default value will be used. Define this macro in order to make frame pointer
2609 elimination work in the presence of @code{__builtin_frame_address (count)} and
2610 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
2613 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2614 A C expression whose value is RTL representing the address in a stack
2615 frame where the pointer to the caller's frame is stored. Assume that
2616 @var{frameaddr} is an RTL expression for the address of the stack frame
2619 If you don't define this macro, the default is to return the value
2620 of @var{frameaddr}---that is, the stack frame address is also the
2621 address of the stack word that points to the previous frame.
2624 @defmac SETUP_FRAME_ADDRESSES
2625 If defined, a C expression that produces the machine-specific code to
2626 setup the stack so that arbitrary frames can be accessed. For example,
2627 on the SPARC, we must flush all of the register windows to the stack
2628 before we can access arbitrary stack frames. You will seldom need to
2632 @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
2634 @defmac FRAME_ADDR_RTX (@var{frameaddr})
2635 A C expression whose value is RTL representing the value of the frame
2636 address for the current frame. @var{frameaddr} is the frame pointer
2637 of the current frame. This is used for __builtin_frame_address.
2638 You need only define this macro if the frame address is not the same
2639 as the frame pointer. Most machines do not need to define it.
2642 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2643 A C expression whose value is RTL representing the value of the return
2644 address for the frame @var{count} steps up from the current frame, after
2645 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
2646 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2647 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
2649 The value of the expression must always be the correct address when
2650 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
2651 determine the return address of other frames.
2654 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2655 Define this if the return address of a particular stack frame is accessed
2656 from the frame pointer of the previous stack frame.
2659 @defmac INCOMING_RETURN_ADDR_RTX
2660 A C expression whose value is RTL representing the location of the
2661 incoming return address at the beginning of any function, before the
2662 prologue. This RTL is either a @code{REG}, indicating that the return
2663 value is saved in @samp{REG}, or a @code{MEM} representing a location in
2666 You only need to define this macro if you want to support call frame
2667 debugging information like that provided by DWARF 2.
2669 If this RTL is a @code{REG}, you should also define
2670 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2673 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
2674 A C expression whose value is an integer giving a DWARF 2 column
2675 number that may be used as an alternative return column. The column
2676 must not correspond to any gcc hard register (that is, it must not
2677 be in the range of @code{DWARF_FRAME_REGNUM}).
2679 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2680 general register, but an alternative column needs to be used for signal
2681 frames. Some targets have also used different frame return columns
2685 @defmac DWARF_ZERO_REG
2686 A C expression whose value is an integer giving a DWARF 2 register
2687 number that is considered to always have the value zero. This should
2688 only be defined if the target has an architected zero register, and
2689 someone decided it was a good idea to use that register number to
2690 terminate the stack backtrace. New ports should avoid this.
2693 @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
2695 @defmac INCOMING_FRAME_SP_OFFSET
2696 A C expression whose value is an integer giving the offset, in bytes,
2697 from the value of the stack pointer register to the top of the stack
2698 frame at the beginning of any function, before the prologue. The top of
2699 the frame is defined to be the value of the stack pointer in the
2700 previous frame, just before the call instruction.
2702 You only need to define this macro if you want to support call frame
2703 debugging information like that provided by DWARF 2.
2706 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2707 A C expression whose value is an integer giving the offset, in bytes,
2708 from the argument pointer to the canonical frame address (cfa). The
2709 final value should coincide with that calculated by
2710 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
2711 during virtual register instantiation.
2713 The default value for this macro is
2714 @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
2715 which is correct for most machines; in general, the arguments are found
2716 immediately before the stack frame. Note that this is not the case on
2717 some targets that save registers into the caller's frame, such as SPARC
2718 and rs6000, and so such targets need to define this macro.
2720 You only need to define this macro if the default is incorrect, and you
2721 want to support call frame debugging information like that provided by
2725 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
2726 If defined, a C expression whose value is an integer giving the offset
2727 in bytes from the frame pointer to the canonical frame address (cfa).
2728 The final value should coincide with that calculated by
2729 @code{INCOMING_FRAME_SP_OFFSET}.
2731 Normally the CFA is calculated as an offset from the argument pointer,
2732 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
2733 variable due to the ABI, this may not be possible. If this macro is
2734 defined, it implies that the virtual register instantiation should be
2735 based on the frame pointer instead of the argument pointer. Only one
2736 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
2740 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
2741 If defined, a C expression whose value is an integer giving the offset
2742 in bytes from the canonical frame address (cfa) to the frame base used
2743 in DWARF 2 debug information. The default is zero. A different value
2744 may reduce the size of debug information on some ports.
2747 @node Exception Handling
2748 @subsection Exception Handling Support
2749 @cindex exception handling
2751 @defmac EH_RETURN_DATA_REGNO (@var{N})
2752 A C expression whose value is the @var{N}th register number used for
2753 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2754 @var{N} registers are usable.
2756 The exception handling library routines communicate with the exception
2757 handlers via a set of agreed upon registers. Ideally these registers
2758 should be call-clobbered; it is possible to use call-saved registers,
2759 but may negatively impact code size. The target must support at least
2760 2 data registers, but should define 4 if there are enough free registers.
2762 You must define this macro if you want to support call frame exception
2763 handling like that provided by DWARF 2.
2766 @defmac EH_RETURN_STACKADJ_RTX
2767 A C expression whose value is RTL representing a location in which
2768 to store a stack adjustment to be applied before function return.
2769 This is used to unwind the stack to an exception handler's call frame.
2770 It will be assigned zero on code paths that return normally.
2772 Typically this is a call-clobbered hard register that is otherwise
2773 untouched by the epilogue, but could also be a stack slot.
2775 Do not define this macro if the stack pointer is saved and restored
2776 by the regular prolog and epilog code in the call frame itself; in
2777 this case, the exception handling library routines will update the
2778 stack location to be restored in place. Otherwise, you must define
2779 this macro if you want to support call frame exception handling like
2780 that provided by DWARF 2.
2783 @defmac EH_RETURN_HANDLER_RTX
2784 A C expression whose value is RTL representing a location in which
2785 to store the address of an exception handler to which we should
2786 return. It will not be assigned on code paths that return normally.
2788 Typically this is the location in the call frame at which the normal
2789 return address is stored. For targets that return by popping an
2790 address off the stack, this might be a memory address just below
2791 the @emph{target} call frame rather than inside the current call
2792 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2793 been assigned, so it may be used to calculate the location of the
2796 Some targets have more complex requirements than storing to an
2797 address calculable during initial code generation. In that case
2798 the @code{eh_return} instruction pattern should be used instead.
2800 If you want to support call frame exception handling, you must
2801 define either this macro or the @code{eh_return} instruction pattern.
2804 @defmac RETURN_ADDR_OFFSET
2805 If defined, an integer-valued C expression for which rtl will be generated
2806 to add it to the exception handler address before it is searched in the
2807 exception handling tables, and to subtract it again from the address before
2808 using it to return to the exception handler.
2811 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
2812 This macro chooses the encoding of pointers embedded in the exception
2813 handling sections. If at all possible, this should be defined such
2814 that the exception handling section will not require dynamic relocations,
2815 and so may be read-only.
2817 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2818 @var{global} is true if the symbol may be affected by dynamic relocations.
2819 The macro should return a combination of the @code{DW_EH_PE_*} defines
2820 as found in @file{dwarf2.h}.
2822 If this macro is not defined, pointers will not be encoded but
2823 represented directly.
2826 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2827 This macro allows the target to emit whatever special magic is required
2828 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2829 Generic code takes care of pc-relative and indirect encodings; this must
2830 be defined if the target uses text-relative or data-relative encodings.
2832 This is a C statement that branches to @var{done} if the format was
2833 handled. @var{encoding} is the format chosen, @var{size} is the number
2834 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2838 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
2839 This macro allows the target to add CPU and operating system specific
2840 code to the call-frame unwinder for use when there is no unwind data
2841 available. The most common reason to implement this macro is to unwind
2842 through signal frames.
2844 This macro is called from @code{uw_frame_state_for} in
2845 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
2846 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2847 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
2848 for the address of the code being executed and @code{context->cfa} for
2849 the stack pointer value. If the frame can be decoded, the register
2850 save addresses should be updated in @var{fs} and the macro should
2851 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
2852 the macro should evaluate to @code{_URC_END_OF_STACK}.
2854 For proper signal handling in Java this macro is accompanied by
2855 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
2858 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
2859 This macro allows the target to add operating system specific code to the
2860 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
2861 usually used for signal or interrupt frames.
2863 This macro is called from @code{uw_update_context} in libgcc's
2864 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
2865 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
2866 for the abi and context in the @code{.unwabi} directive. If the
2867 @code{.unwabi} directive can be handled, the register save addresses should
2868 be updated in @var{fs}.
2871 @defmac TARGET_USES_WEAK_UNWIND_INFO
2872 A C expression that evaluates to true if the target requires unwind
2873 info to be given comdat linkage. Define it to be @code{1} if comdat
2874 linkage is necessary. The default is @code{0}.
2877 @node Stack Checking
2878 @subsection Specifying How Stack Checking is Done
2880 GCC will check that stack references are within the boundaries of the
2881 stack, if the option @option{-fstack-check} is specified, in one of
2886 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2887 will assume that you have arranged for full stack checking to be done
2888 at appropriate places in the configuration files. GCC will not do
2889 other special processing.
2892 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
2893 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
2894 that you have arranged for static stack checking (checking of the
2895 static stack frame of functions) to be done at appropriate places
2896 in the configuration files. GCC will only emit code to do dynamic
2897 stack checking (checking on dynamic stack allocations) using the third
2901 If neither of the above are true, GCC will generate code to periodically
2902 ``probe'' the stack pointer using the values of the macros defined below.
2905 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
2906 GCC will change its allocation strategy for large objects if the option
2907 @option{-fstack-check} is specified: they will always be allocated
2908 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
2910 @defmac STACK_CHECK_BUILTIN
2911 A nonzero value if stack checking is done by the configuration files in a
2912 machine-dependent manner. You should define this macro if stack checking
2913 is required by the ABI of your machine or if you would like to do stack
2914 checking in some more efficient way than the generic approach. The default
2915 value of this macro is zero.
2918 @defmac STACK_CHECK_STATIC_BUILTIN
2919 A nonzero value if static stack checking is done by the configuration files
2920 in a machine-dependent manner. You should define this macro if you would
2921 like to do static stack checking in some more efficient way than the generic
2922 approach. The default value of this macro is zero.
2925 @defmac STACK_CHECK_PROBE_INTERVAL_EXP
2926 An integer specifying the interval at which GCC must generate stack probe
2927 instructions, defined as 2 raised to this integer. You will normally
2928 define this macro so that the interval be no larger than the size of
2929 the ``guard pages'' at the end of a stack area. The default value
2930 of 12 (4096-byte interval) is suitable for most systems.
2933 @defmac STACK_CHECK_MOVING_SP
2934 An integer which is nonzero if GCC should move the stack pointer page by page
2935 when doing probes. This can be necessary on systems where the stack pointer
2936 contains the bottom address of the memory area accessible to the executing
2937 thread at any point in time. In this situation an alternate signal stack
2938 is required in order to be able to recover from a stack overflow. The
2939 default value of this macro is zero.
2942 @defmac STACK_CHECK_PROTECT
2943 The number of bytes of stack needed to recover from a stack overflow, for
2944 languages where such a recovery is supported. The default value of 75 words
2945 with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
2946 8192 bytes with other exception handling mechanisms should be adequate for
2950 The following macros are relevant only if neither STACK_CHECK_BUILTIN
2951 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
2952 in the opposite case.
2954 @defmac STACK_CHECK_MAX_FRAME_SIZE
2955 The maximum size of a stack frame, in bytes. GCC will generate probe
2956 instructions in non-leaf functions to ensure at least this many bytes of
2957 stack are available. If a stack frame is larger than this size, stack
2958 checking will not be reliable and GCC will issue a warning. The
2959 default is chosen so that GCC only generates one instruction on most
2960 systems. You should normally not change the default value of this macro.
2963 @defmac STACK_CHECK_FIXED_FRAME_SIZE
2964 GCC uses this value to generate the above warning message. It
2965 represents the amount of fixed frame used by a function, not including
2966 space for any callee-saved registers, temporaries and user variables.
2967 You need only specify an upper bound for this amount and will normally
2968 use the default of four words.
2971 @defmac STACK_CHECK_MAX_VAR_SIZE
2972 The maximum size, in bytes, of an object that GCC will place in the
2973 fixed area of the stack frame when the user specifies
2974 @option{-fstack-check}.
2975 GCC computed the default from the values of the above macros and you will
2976 normally not need to override that default.
2980 @node Frame Registers
2981 @subsection Registers That Address the Stack Frame
2983 @c prevent bad page break with this line
2984 This discusses registers that address the stack frame.
2986 @defmac STACK_POINTER_REGNUM
2987 The register number of the stack pointer register, which must also be a
2988 fixed register according to @code{FIXED_REGISTERS}. On most machines,
2989 the hardware determines which register this is.
2992 @defmac FRAME_POINTER_REGNUM
2993 The register number of the frame pointer register, which is used to
2994 access automatic variables in the stack frame. On some machines, the
2995 hardware determines which register this is. On other machines, you can
2996 choose any register you wish for this purpose.
2999 @defmac HARD_FRAME_POINTER_REGNUM
3000 On some machines the offset between the frame pointer and starting
3001 offset of the automatic variables is not known until after register
3002 allocation has been done (for example, because the saved registers are
3003 between these two locations). On those machines, define
3004 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3005 be used internally until the offset is known, and define
3006 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3007 used for the frame pointer.
3009 You should define this macro only in the very rare circumstances when it
3010 is not possible to calculate the offset between the frame pointer and
3011 the automatic variables until after register allocation has been
3012 completed. When this macro is defined, you must also indicate in your
3013 definition of @code{ELIMINABLE_REGS} how to eliminate
3014 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3015 or @code{STACK_POINTER_REGNUM}.
3017 Do not define this macro if it would be the same as
3018 @code{FRAME_POINTER_REGNUM}.
3021 @defmac ARG_POINTER_REGNUM
3022 The register number of the arg pointer register, which is used to access
3023 the function's argument list. On some machines, this is the same as the
3024 frame pointer register. On some machines, the hardware determines which
3025 register this is. On other machines, you can choose any register you
3026 wish for this purpose. If this is not the same register as the frame
3027 pointer register, then you must mark it as a fixed register according to
3028 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3029 (@pxref{Elimination}).
3032 @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
3033 Define this to a preprocessor constant that is nonzero if
3034 @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
3035 the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
3036 == FRAME_POINTER_REGNUM)}; you only need to define this macro if that
3037 definition is not suitable for use in preprocessor conditionals.
3040 @defmac HARD_FRAME_POINTER_IS_ARG_POINTER
3041 Define this to a preprocessor constant that is nonzero if
3042 @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
3043 same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
3044 ARG_POINTER_REGNUM)}; you only need to define this macro if that
3045 definition is not suitable for use in preprocessor conditionals.
3048 @defmac RETURN_ADDRESS_POINTER_REGNUM
3049 The register number of the return address pointer register, which is used to
3050 access the current function's return address from the stack. On some
3051 machines, the return address is not at a fixed offset from the frame
3052 pointer or stack pointer or argument pointer. This register can be defined
3053 to point to the return address on the stack, and then be converted by
3054 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3056 Do not define this macro unless there is no other way to get the return
3057 address from the stack.
3060 @defmac STATIC_CHAIN_REGNUM
3061 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3062 Register numbers used for passing a function's static chain pointer. If
3063 register windows are used, the register number as seen by the called
3064 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3065 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3066 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3069 The static chain register need not be a fixed register.
3071 If the static chain is passed in memory, these macros should not be
3072 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3075 @hook TARGET_STATIC_CHAIN
3077 @defmac DWARF_FRAME_REGISTERS
3078 This macro specifies the maximum number of hard registers that can be
3079 saved in a call frame. This is used to size data structures used in
3080 DWARF2 exception handling.
3082 Prior to GCC 3.0, this macro was needed in order to establish a stable
3083 exception handling ABI in the face of adding new hard registers for ISA
3084 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3085 in the number of hard registers. Nevertheless, this macro can still be
3086 used to reduce the runtime memory requirements of the exception handling
3087 routines, which can be substantial if the ISA contains a lot of
3088 registers that are not call-saved.
3090 If this macro is not defined, it defaults to
3091 @code{FIRST_PSEUDO_REGISTER}.
3094 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3096 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3097 for backward compatibility in pre GCC 3.0 compiled code.
3099 If this macro is not defined, it defaults to
3100 @code{DWARF_FRAME_REGISTERS}.
3103 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3105 Define this macro if the target's representation for dwarf registers
3106 is different than the internal representation for unwind column.
3107 Given a dwarf register, this macro should return the internal unwind
3108 column number to use instead.
3110 See the PowerPC's SPE target for an example.
3113 @defmac DWARF_FRAME_REGNUM (@var{regno})
3115 Define this macro if the target's representation for dwarf registers
3116 used in .eh_frame or .debug_frame is different from that used in other
3117 debug info sections. Given a GCC hard register number, this macro
3118 should return the .eh_frame register number. The default is
3119 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3123 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3125 Define this macro to map register numbers held in the call frame info
3126 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3127 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3128 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3129 return @code{@var{regno}}.
3133 @defmac REG_VALUE_IN_UNWIND_CONTEXT
3135 Define this macro if the target stores register values as
3136 @code{_Unwind_Word} type in unwind context. It should be defined if
3137 target register size is larger than the size of @code{void *}. The
3138 default is to store register values as @code{void *} type.
3142 @defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3144 Define this macro to be 1 if the target always uses extended unwind
3145 context with version, args_size and by_value fields. If it is undefined,
3146 it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3147 defined and 0 otherwise.
3152 @subsection Eliminating Frame Pointer and Arg Pointer
3154 @c prevent bad page break with this line
3155 This is about eliminating the frame pointer and arg pointer.
3157 @hook TARGET_FRAME_POINTER_REQUIRED
3159 @findex get_frame_size
3160 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3161 A C statement to store in the variable @var{depth-var} the difference
3162 between the frame pointer and the stack pointer values immediately after
3163 the function prologue. The value would be computed from information
3164 such as the result of @code{get_frame_size ()} and the tables of
3165 registers @code{regs_ever_live} and @code{call_used_regs}.
3167 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3168 need not be defined. Otherwise, it must be defined even if
3169 @code{TARGET_FRAME_POINTER_REQUIRED} always returns true; in that
3170 case, you may set @var{depth-var} to anything.
3173 @defmac ELIMINABLE_REGS
3174 If defined, this macro specifies a table of register pairs used to
3175 eliminate unneeded registers that point into the stack frame. If it is not
3176 defined, the only elimination attempted by the compiler is to replace
3177 references to the frame pointer with references to the stack pointer.
3179 The definition of this macro is a list of structure initializations, each
3180 of which specifies an original and replacement register.
3182 On some machines, the position of the argument pointer is not known until
3183 the compilation is completed. In such a case, a separate hard register
3184 must be used for the argument pointer. This register can be eliminated by
3185 replacing it with either the frame pointer or the argument pointer,
3186 depending on whether or not the frame pointer has been eliminated.
3188 In this case, you might specify:
3190 #define ELIMINABLE_REGS \
3191 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3192 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3193 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3196 Note that the elimination of the argument pointer with the stack pointer is
3197 specified first since that is the preferred elimination.
3200 @hook TARGET_CAN_ELIMINATE
3202 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3203 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3204 specifies the initial difference between the specified pair of
3205 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3209 @node Stack Arguments
3210 @subsection Passing Function Arguments on the Stack
3211 @cindex arguments on stack
3212 @cindex stack arguments
3214 The macros in this section control how arguments are passed
3215 on the stack. See the following section for other macros that
3216 control passing certain arguments in registers.
3218 @hook TARGET_PROMOTE_PROTOTYPES
3221 A C expression. If nonzero, push insns will be used to pass
3223 If the target machine does not have a push instruction, set it to zero.
3224 That directs GCC to use an alternate strategy: to
3225 allocate the entire argument block and then store the arguments into
3226 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3229 @defmac PUSH_ARGS_REVERSED
3230 A C expression. If nonzero, function arguments will be evaluated from
3231 last to first, rather than from first to last. If this macro is not
3232 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3233 and args grow in opposite directions, and 0 otherwise.
3236 @defmac PUSH_ROUNDING (@var{npushed})
3237 A C expression that is the number of bytes actually pushed onto the
3238 stack when an instruction attempts to push @var{npushed} bytes.
3240 On some machines, the definition
3243 #define PUSH_ROUNDING(BYTES) (BYTES)
3247 will suffice. But on other machines, instructions that appear
3248 to push one byte actually push two bytes in an attempt to maintain
3249 alignment. Then the definition should be
3252 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3255 If the value of this macro has a type, it should be an unsigned type.
3258 @findex outgoing_args_size
3259 @findex crtl->outgoing_args_size
3260 @defmac ACCUMULATE_OUTGOING_ARGS
3261 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3262 will be computed and placed into
3263 @code{crtl->outgoing_args_size}. No space will be pushed
3264 onto the stack for each call; instead, the function prologue should
3265 increase the stack frame size by this amount.
3267 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3271 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3272 Define this macro if functions should assume that stack space has been
3273 allocated for arguments even when their values are passed in
3276 The value of this macro is the size, in bytes, of the area reserved for
3277 arguments passed in registers for the function represented by @var{fndecl},
3278 which can be zero if GCC is calling a library function.
3279 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3282 This space can be allocated by the caller, or be a part of the
3283 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3286 @c above is overfull. not sure what to do. --mew 5feb93 did
3287 @c something, not sure if it looks good. --mew 10feb93
3289 @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3290 Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3291 Define this macro if space guaranteed when compiling a function body
3292 is different to space required when making a call, a situation that
3293 can arise with K&R style function definitions.
3296 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3297 Define this to a nonzero value if it is the responsibility of the
3298 caller to allocate the area reserved for arguments passed in registers
3299 when calling a function of @var{fntype}. @var{fntype} may be NULL
3300 if the function called is a library function.
3302 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3303 whether the space for these arguments counts in the value of
3304 @code{crtl->outgoing_args_size}.
3307 @defmac STACK_PARMS_IN_REG_PARM_AREA
3308 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3309 stack parameters don't skip the area specified by it.
3310 @c i changed this, makes more sens and it should have taken care of the
3311 @c overfull.. not as specific, tho. --mew 5feb93
3313 Normally, when a parameter is not passed in registers, it is placed on the
3314 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3315 suppresses this behavior and causes the parameter to be passed on the
3316 stack in its natural location.
3319 @hook TARGET_RETURN_POPS_ARGS
3321 @defmac CALL_POPS_ARGS (@var{cum})
3322 A C expression that should indicate the number of bytes a call sequence
3323 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3324 when compiling a function call.
3326 @var{cum} is the variable in which all arguments to the called function
3327 have been accumulated.
3329 On certain architectures, such as the SH5, a call trampoline is used
3330 that pops certain registers off the stack, depending on the arguments
3331 that have been passed to the function. Since this is a property of the
3332 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3336 @node Register Arguments
3337 @subsection Passing Arguments in Registers
3338 @cindex arguments in registers
3339 @cindex registers arguments
3341 This section describes the macros which let you control how various
3342 types of arguments are passed in registers or how they are arranged in
3345 @hook TARGET_FUNCTION_ARG
3347 @hook TARGET_MUST_PASS_IN_STACK
3349 @hook TARGET_FUNCTION_INCOMING_ARG
3351 @hook TARGET_USE_PSEUDO_PIC_REG
3353 @hook TARGET_INIT_PIC_REG
3355 @hook TARGET_ARG_PARTIAL_BYTES
3357 @hook TARGET_PASS_BY_REFERENCE
3359 @hook TARGET_CALLEE_COPIES
3361 @defmac CUMULATIVE_ARGS
3362 A C type for declaring a variable that is used as the first argument
3363 of @code{TARGET_FUNCTION_ARG} and other related values. For some
3364 target machines, the type @code{int} suffices and can hold the number
3365 of bytes of argument so far.
3367 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3368 arguments that have been passed on the stack. The compiler has other
3369 variables to keep track of that. For target machines on which all
3370 arguments are passed on the stack, there is no need to store anything in
3371 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
3372 should not be empty, so use @code{int}.
3375 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
3376 If defined, this macro is called before generating any code for a
3377 function, but after the @var{cfun} descriptor for the function has been
3378 created. The back end may use this macro to update @var{cfun} to
3379 reflect an ABI other than that which would normally be used by default.
3380 If the compiler is generating code for a compiler-generated function,
3381 @var{fndecl} may be @code{NULL}.
3384 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3385 A C statement (sans semicolon) for initializing the variable
3386 @var{cum} for the state at the beginning of the argument list. The
3387 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
3388 is the tree node for the data type of the function which will receive
3389 the args, or 0 if the args are to a compiler support library function.
3390 For direct calls that are not libcalls, @var{fndecl} contain the
3391 declaration node of the function. @var{fndecl} is also set when
3392 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3393 being compiled. @var{n_named_args} is set to the number of named
3394 arguments, including a structure return address if it is passed as a
3395 parameter, when making a call. When processing incoming arguments,
3396 @var{n_named_args} is set to @minus{}1.
3398 When processing a call to a compiler support library function,
3399 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
3400 contains the name of the function, as a string. @var{libname} is 0 when
3401 an ordinary C function call is being processed. Thus, each time this
3402 macro is called, either @var{libname} or @var{fntype} is nonzero, but
3403 never both of them at once.
3406 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3407 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3408 it gets a @code{MODE} argument instead of @var{fntype}, that would be
3409 @code{NULL}. @var{indirect} would always be zero, too. If this macro
3410 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
3411 0)} is used instead.
3414 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3415 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3416 finding the arguments for the function being compiled. If this macro is
3417 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3419 The value passed for @var{libname} is always 0, since library routines
3420 with special calling conventions are never compiled with GCC@. The
3421 argument @var{libname} exists for symmetry with
3422 @code{INIT_CUMULATIVE_ARGS}.
3423 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3424 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
3427 @hook TARGET_FUNCTION_ARG_ADVANCE
3429 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
3430 If defined, a C expression that is the number of bytes to add to the
3431 offset of the argument passed in memory. This is needed for the SPU,
3432 which passes @code{char} and @code{short} arguments in the preferred
3433 slot that is in the middle of the quad word instead of starting at the
3437 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
3438 If defined, a C expression which determines whether, and in which direction,
3439 to pad out an argument with extra space. The value should be of type
3440 @code{enum direction}: either @code{upward} to pad above the argument,
3441 @code{downward} to pad below, or @code{none} to inhibit padding.
3443 The @emph{amount} of padding is not controlled by this macro, but by the
3444 target hook @code{TARGET_FUNCTION_ARG_ROUND_BOUNDARY}. It is
3445 always just enough to reach the next multiple of that boundary.
3447 This macro has a default definition which is right for most systems.
3448 For little-endian machines, the default is to pad upward. For
3449 big-endian machines, the default is to pad downward for an argument of
3450 constant size shorter than an @code{int}, and upward otherwise.
3453 @defmac PAD_VARARGS_DOWN
3454 If defined, a C expression which determines whether the default
3455 implementation of va_arg will attempt to pad down before reading the
3456 next argument, if that argument is smaller than its aligned space as
3457 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
3458 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3461 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3462 Specify padding for the last element of a block move between registers and
3463 memory. @var{first} is nonzero if this is the only element. Defining this
3464 macro allows better control of register function parameters on big-endian
3465 machines, without using @code{PARALLEL} rtl. In particular,
3466 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3467 registers, as there is no longer a "wrong" part of a register; For example,
3468 a three byte aggregate may be passed in the high part of a register if so
3472 @hook TARGET_FUNCTION_ARG_BOUNDARY
3474 @hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY
3476 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
3477 A C expression that is nonzero if @var{regno} is the number of a hard
3478 register in which function arguments are sometimes passed. This does
3479 @emph{not} include implicit arguments such as the static chain and
3480 the structure-value address. On many machines, no registers can be
3481 used for this purpose since all function arguments are pushed on the
3485 @hook TARGET_SPLIT_COMPLEX_ARG
3487 @hook TARGET_BUILD_BUILTIN_VA_LIST
3489 @hook TARGET_ENUM_VA_LIST_P
3491 @hook TARGET_FN_ABI_VA_LIST
3493 @hook TARGET_CANONICAL_VA_LIST_TYPE
3495 @hook TARGET_GIMPLIFY_VA_ARG_EXPR
3497 @hook TARGET_VALID_POINTER_MODE
3499 @hook TARGET_REF_MAY_ALIAS_ERRNO
3501 @hook TARGET_SCALAR_MODE_SUPPORTED_P
3503 @hook TARGET_VECTOR_MODE_SUPPORTED_P
3505 @hook TARGET_ARRAY_MODE_SUPPORTED_P
3507 @hook TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P
3509 @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
3511 @hook TARGET_FLAGS_REGNUM
3514 @subsection How Scalar Function Values Are Returned
3515 @cindex return values in registers
3516 @cindex values, returned by functions
3517 @cindex scalars, returned as values
3519 This section discusses the macros that control returning scalars as
3520 values---values that can fit in registers.
3522 @hook TARGET_FUNCTION_VALUE
3524 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3525 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
3526 a new target instead.
3529 @defmac LIBCALL_VALUE (@var{mode})
3530 A C expression to create an RTX representing the place where a library
3531 function returns a value of mode @var{mode}.
3533 Note that ``library function'' in this context means a compiler
3534 support routine, used to perform arithmetic, whose name is known
3535 specially by the compiler and was not mentioned in the C code being
3539 @hook TARGET_LIBCALL_VALUE
3541 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3542 A C expression that is nonzero if @var{regno} is the number of a hard
3543 register in which the values of called function may come back.
3545 A register whose use for returning values is limited to serving as the
3546 second of a pair (for a value of type @code{double}, say) need not be
3547 recognized by this macro. So for most machines, this definition
3551 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3554 If the machine has register windows, so that the caller and the called
3555 function use different registers for the return value, this macro
3556 should recognize only the caller's register numbers.
3558 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
3559 for a new target instead.
3562 @hook TARGET_FUNCTION_VALUE_REGNO_P
3564 @defmac APPLY_RESULT_SIZE
3565 Define this macro if @samp{untyped_call} and @samp{untyped_return}
3566 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3567 saving and restoring an arbitrary return value.
3570 @hook TARGET_RETURN_IN_MSB
3572 @node Aggregate Return
3573 @subsection How Large Values Are Returned
3574 @cindex aggregates as return values
3575 @cindex large return values
3576 @cindex returning aggregate values
3577 @cindex structure value address
3579 When a function value's mode is @code{BLKmode} (and in some other
3580 cases), the value is not returned according to
3581 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
3582 caller passes the address of a block of memory in which the value
3583 should be stored. This address is called the @dfn{structure value
3586 This section describes how to control returning structure values in
3589 @hook TARGET_RETURN_IN_MEMORY
3591 @defmac DEFAULT_PCC_STRUCT_RETURN
3592 Define this macro to be 1 if all structure and union return values must be
3593 in memory. Since this results in slower code, this should be defined
3594 only if needed for compatibility with other compilers or with an ABI@.
3595 If you define this macro to be 0, then the conventions used for structure
3596 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
3599 If not defined, this defaults to the value 1.
3602 @hook TARGET_STRUCT_VALUE_RTX
3604 @defmac PCC_STATIC_STRUCT_RETURN
3605 Define this macro if the usual system convention on the target machine
3606 for returning structures and unions is for the called function to return
3607 the address of a static variable containing the value.
3609 Do not define this if the usual system convention is for the caller to
3610 pass an address to the subroutine.
3612 This macro has effect in @option{-fpcc-struct-return} mode, but it does
3613 nothing when you use @option{-freg-struct-return} mode.
3616 @hook TARGET_GET_RAW_RESULT_MODE
3618 @hook TARGET_GET_RAW_ARG_MODE
3621 @subsection Caller-Saves Register Allocation
3623 If you enable it, GCC can save registers around function calls. This
3624 makes it possible to use call-clobbered registers to hold variables that
3625 must live across calls.
3627 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3628 A C expression specifying which mode is required for saving @var{nregs}
3629 of a pseudo-register in call-clobbered hard register @var{regno}. If
3630 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3631 returned. For most machines this macro need not be defined since GCC
3632 will select the smallest suitable mode.
3635 @node Function Entry
3636 @subsection Function Entry and Exit
3637 @cindex function entry and exit
3641 This section describes the macros that output function entry
3642 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
3644 @hook TARGET_ASM_FUNCTION_PROLOGUE
3646 @hook TARGET_ASM_FUNCTION_END_PROLOGUE
3648 @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
3650 @hook TARGET_ASM_FUNCTION_EPILOGUE
3654 @findex pretend_args_size
3655 @findex crtl->args.pretend_args_size
3656 A region of @code{crtl->args.pretend_args_size} bytes of
3657 uninitialized space just underneath the first argument arriving on the
3658 stack. (This may not be at the very start of the allocated stack region
3659 if the calling sequence has pushed anything else since pushing the stack
3660 arguments. But usually, on such machines, nothing else has been pushed
3661 yet, because the function prologue itself does all the pushing.) This
3662 region is used on machines where an argument may be passed partly in
3663 registers and partly in memory, and, in some cases to support the
3664 features in @code{<stdarg.h>}.
3667 An area of memory used to save certain registers used by the function.
3668 The size of this area, which may also include space for such things as
3669 the return address and pointers to previous stack frames, is
3670 machine-specific and usually depends on which registers have been used
3671 in the function. Machines with register windows often do not require
3675 A region of at least @var{size} bytes, possibly rounded up to an allocation
3676 boundary, to contain the local variables of the function. On some machines,
3677 this region and the save area may occur in the opposite order, with the
3678 save area closer to the top of the stack.
3681 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3682 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3683 @code{crtl->outgoing_args_size} bytes to be used for outgoing
3684 argument lists of the function. @xref{Stack Arguments}.
3687 @defmac EXIT_IGNORE_STACK
3688 Define this macro as a C expression that is nonzero if the return
3689 instruction or the function epilogue ignores the value of the stack
3690 pointer; in other words, if it is safe to delete an instruction to
3691 adjust the stack pointer before a return from the function. The
3694 Note that this macro's value is relevant only for functions for which
3695 frame pointers are maintained. It is never safe to delete a final
3696 stack adjustment in a function that has no frame pointer, and the
3697 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3700 @defmac EPILOGUE_USES (@var{regno})
3701 Define this macro as a C expression that is nonzero for registers that are
3702 used by the epilogue or the @samp{return} pattern. The stack and frame
3703 pointer registers are already assumed to be used as needed.
3706 @defmac EH_USES (@var{regno})
3707 Define this macro as a C expression that is nonzero for registers that are
3708 used by the exception handling mechanism, and so should be considered live
3709 on entry to an exception edge.
3712 @hook TARGET_ASM_OUTPUT_MI_THUNK
3714 @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
3717 @subsection Generating Code for Profiling
3718 @cindex profiling, code generation
3720 These macros will help you generate code for profiling.
3722 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
3723 A C statement or compound statement to output to @var{file} some
3724 assembler code to call the profiling subroutine @code{mcount}.
3727 The details of how @code{mcount} expects to be called are determined by
3728 your operating system environment, not by GCC@. To figure them out,
3729 compile a small program for profiling using the system's installed C
3730 compiler and look at the assembler code that results.
3732 Older implementations of @code{mcount} expect the address of a counter
3733 variable to be loaded into some register. The name of this variable is
3734 @samp{LP} followed by the number @var{labelno}, so you would generate
3735 the name using @samp{LP%d} in a @code{fprintf}.
3738 @defmac PROFILE_HOOK
3739 A C statement or compound statement to output to @var{file} some assembly
3740 code to call the profiling subroutine @code{mcount} even the target does
3741 not support profiling.
3744 @defmac NO_PROFILE_COUNTERS
3745 Define this macro to be an expression with a nonzero value if the
3746 @code{mcount} subroutine on your system does not need a counter variable
3747 allocated for each function. This is true for almost all modern
3748 implementations. If you define this macro, you must not use the
3749 @var{labelno} argument to @code{FUNCTION_PROFILER}.
3752 @defmac PROFILE_BEFORE_PROLOGUE
3753 Define this macro if the code for function profiling should come before
3754 the function prologue. Normally, the profiling code comes after.
3757 @hook TARGET_KEEP_LEAF_WHEN_PROFILED
3760 @subsection Permitting tail calls
3763 @hook TARGET_FUNCTION_OK_FOR_SIBCALL
3765 @hook TARGET_EXTRA_LIVE_ON_ENTRY
3767 @hook TARGET_SET_UP_BY_PROLOGUE
3769 @hook TARGET_WARN_FUNC_RETURN
3771 @node Stack Smashing Protection
3772 @subsection Stack smashing protection
3773 @cindex stack smashing protection
3775 @hook TARGET_STACK_PROTECT_GUARD
3777 @hook TARGET_STACK_PROTECT_FAIL
3779 @hook TARGET_SUPPORTS_SPLIT_STACK
3781 @node Miscellaneous Register Hooks
3782 @subsection Miscellaneous register hooks
3783 @cindex miscellaneous register hooks
3785 @hook TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
3788 @section Implementing the Varargs Macros
3789 @cindex varargs implementation
3791 GCC comes with an implementation of @code{<varargs.h>} and
3792 @code{<stdarg.h>} that work without change on machines that pass arguments
3793 on the stack. Other machines require their own implementations of
3794 varargs, and the two machine independent header files must have
3795 conditionals to include it.
3797 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
3798 the calling convention for @code{va_start}. The traditional
3799 implementation takes just one argument, which is the variable in which
3800 to store the argument pointer. The ISO implementation of
3801 @code{va_start} takes an additional second argument. The user is
3802 supposed to write the last named argument of the function here.
3804 However, @code{va_start} should not use this argument. The way to find
3805 the end of the named arguments is with the built-in functions described
3808 @defmac __builtin_saveregs ()
3809 Use this built-in function to save the argument registers in memory so
3810 that the varargs mechanism can access them. Both ISO and traditional
3811 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3812 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
3814 On some machines, @code{__builtin_saveregs} is open-coded under the
3815 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
3816 other machines, it calls a routine written in assembler language,
3817 found in @file{libgcc2.c}.
3819 Code generated for the call to @code{__builtin_saveregs} appears at the
3820 beginning of the function, as opposed to where the call to
3821 @code{__builtin_saveregs} is written, regardless of what the code is.
3822 This is because the registers must be saved before the function starts
3823 to use them for its own purposes.
3824 @c i rewrote the first sentence above to fix an overfull hbox. --mew
3828 @defmac __builtin_next_arg (@var{lastarg})
3829 This builtin returns the address of the first anonymous stack
3830 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
3831 returns the address of the location above the first anonymous stack
3832 argument. Use it in @code{va_start} to initialize the pointer for
3833 fetching arguments from the stack. Also use it in @code{va_start} to
3834 verify that the second parameter @var{lastarg} is the last named argument
3835 of the current function.
3838 @defmac __builtin_classify_type (@var{object})
3839 Since each machine has its own conventions for which data types are
3840 passed in which kind of register, your implementation of @code{va_arg}
3841 has to embody these conventions. The easiest way to categorize the
3842 specified data type is to use @code{__builtin_classify_type} together
3843 with @code{sizeof} and @code{__alignof__}.
3845 @code{__builtin_classify_type} ignores the value of @var{object},
3846 considering only its data type. It returns an integer describing what
3847 kind of type that is---integer, floating, pointer, structure, and so on.
3849 The file @file{typeclass.h} defines an enumeration that you can use to
3850 interpret the values of @code{__builtin_classify_type}.
3853 These machine description macros help implement varargs:
3855 @hook TARGET_EXPAND_BUILTIN_SAVEREGS
3857 @hook TARGET_SETUP_INCOMING_VARARGS
3859 @hook TARGET_STRICT_ARGUMENT_NAMING
3861 @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
3864 @section Trampolines for Nested Functions
3865 @cindex trampolines for nested functions
3866 @cindex nested functions, trampolines for
3868 A @dfn{trampoline} is a small piece of code that is created at run time
3869 when the address of a nested function is taken. It normally resides on
3870 the stack, in the stack frame of the containing function. These macros
3871 tell GCC how to generate code to allocate and initialize a
3874 The instructions in the trampoline must do two things: load a constant
3875 address into the static chain register, and jump to the real address of
3876 the nested function. On CISC machines such as the m68k, this requires
3877 two instructions, a move immediate and a jump. Then the two addresses
3878 exist in the trampoline as word-long immediate operands. On RISC
3879 machines, it is often necessary to load each address into a register in
3880 two parts. Then pieces of each address form separate immediate
3883 The code generated to initialize the trampoline must store the variable
3884 parts---the static chain value and the function address---into the
3885 immediate operands of the instructions. On a CISC machine, this is
3886 simply a matter of copying each address to a memory reference at the
3887 proper offset from the start of the trampoline. On a RISC machine, it
3888 may be necessary to take out pieces of the address and store them
3891 @hook TARGET_ASM_TRAMPOLINE_TEMPLATE
3893 @defmac TRAMPOLINE_SECTION
3894 Return the section into which the trampoline template is to be placed
3895 (@pxref{Sections}). The default value is @code{readonly_data_section}.
3898 @defmac TRAMPOLINE_SIZE
3899 A C expression for the size in bytes of the trampoline, as an integer.
3902 @defmac TRAMPOLINE_ALIGNMENT
3903 Alignment required for trampolines, in bits.
3905 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
3906 is used for aligning trampolines.
3909 @hook TARGET_TRAMPOLINE_INIT
3911 @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
3913 Implementing trampolines is difficult on many machines because they have
3914 separate instruction and data caches. Writing into a stack location
3915 fails to clear the memory in the instruction cache, so when the program
3916 jumps to that location, it executes the old contents.
3918 Here are two possible solutions. One is to clear the relevant parts of
3919 the instruction cache whenever a trampoline is set up. The other is to
3920 make all trampolines identical, by having them jump to a standard
3921 subroutine. The former technique makes trampoline execution faster; the
3922 latter makes initialization faster.
3924 To clear the instruction cache when a trampoline is initialized, define
3925 the following macro.
3927 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
3928 If defined, expands to a C expression clearing the @emph{instruction
3929 cache} in the specified interval. The definition of this macro would
3930 typically be a series of @code{asm} statements. Both @var{beg} and
3931 @var{end} are both pointer expressions.
3934 To use a standard subroutine, define the following macro. In addition,
3935 you must make sure that the instructions in a trampoline fill an entire
3936 cache line with identical instructions, or else ensure that the
3937 beginning of the trampoline code is always aligned at the same point in
3938 its cache line. Look in @file{m68k.h} as a guide.
3940 @defmac TRANSFER_FROM_TRAMPOLINE
3941 Define this macro if trampolines need a special subroutine to do their
3942 work. The macro should expand to a series of @code{asm} statements
3943 which will be compiled with GCC@. They go in a library function named
3944 @code{__transfer_from_trampoline}.
3946 If you need to avoid executing the ordinary prologue code of a compiled
3947 C function when you jump to the subroutine, you can do so by placing a
3948 special label of your own in the assembler code. Use one @code{asm}
3949 statement to generate an assembler label, and another to make the label
3950 global. Then trampolines can use that label to jump directly to your
3951 special assembler code.
3955 @section Implicit Calls to Library Routines
3956 @cindex library subroutine names
3957 @cindex @file{libgcc.a}
3959 @c prevent bad page break with this line
3960 Here is an explanation of implicit calls to library routines.
3962 @defmac DECLARE_LIBRARY_RENAMES
3963 This macro, if defined, should expand to a piece of C code that will get
3964 expanded when compiling functions for libgcc.a. It can be used to
3965 provide alternate names for GCC's internal library functions if there
3966 are ABI-mandated names that the compiler should provide.
3969 @findex set_optab_libfunc
3970 @findex init_one_libfunc
3971 @hook TARGET_INIT_LIBFUNCS
3973 @hook TARGET_LIBFUNC_GNU_PREFIX
3975 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
3976 This macro should return @code{true} if the library routine that
3977 implements the floating point comparison operator @var{comparison} in
3978 mode @var{mode} will return a boolean, and @var{false} if it will
3981 GCC's own floating point libraries return tristates from the
3982 comparison operators, so the default returns false always. Most ports
3983 don't need to define this macro.
3986 @defmac TARGET_LIB_INT_CMP_BIASED
3987 This macro should evaluate to @code{true} if the integer comparison
3988 functions (like @code{__cmpdi2}) return 0 to indicate that the first
3989 operand is smaller than the second, 1 to indicate that they are equal,
3990 and 2 to indicate that the first operand is greater than the second.
3991 If this macro evaluates to @code{false} the comparison functions return
3992 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
3993 in @file{libgcc.a}, you do not need to define this macro.
3996 @defmac TARGET_HAS_NO_HW_DIVIDE
3997 This macro should be defined if the target has no hardware divide
3998 instructions. If this macro is defined, GCC will use an algorithm which
3999 make use of simple logical and arithmetic operations for 64-bit
4000 division. If the macro is not defined, GCC will use an algorithm which
4001 make use of a 64-bit by 32-bit divide primitive.
4004 @cindex @code{EDOM}, implicit usage
4007 The value of @code{EDOM} on the target machine, as a C integer constant
4008 expression. If you don't define this macro, GCC does not attempt to
4009 deposit the value of @code{EDOM} into @code{errno} directly. Look in
4010 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
4013 If you do not define @code{TARGET_EDOM}, then compiled code reports
4014 domain errors by calling the library function and letting it report the
4015 error. If mathematical functions on your system use @code{matherr} when
4016 there is an error, then you should leave @code{TARGET_EDOM} undefined so
4017 that @code{matherr} is used normally.
4020 @cindex @code{errno}, implicit usage
4021 @defmac GEN_ERRNO_RTX
4022 Define this macro as a C expression to create an rtl expression that
4023 refers to the global ``variable'' @code{errno}. (On certain systems,
4024 @code{errno} may not actually be a variable.) If you don't define this
4025 macro, a reasonable default is used.
4028 @hook TARGET_LIBC_HAS_FUNCTION
4030 @defmac NEXT_OBJC_RUNTIME
4031 Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
4032 by default. This calling convention involves passing the object, the selector
4033 and the method arguments all at once to the method-lookup library function.
4034 This is the usual setting when targeting Darwin/Mac OS X systems, which have
4035 the NeXT runtime installed.
4037 If the macro is set to 0, the "GNU" Objective-C message sending convention
4038 will be used by default. This convention passes just the object and the
4039 selector to the method-lookup function, which returns a pointer to the method.
4041 In either case, it remains possible to select code-generation for the alternate
4042 scheme, by means of compiler command line switches.
4045 @node Addressing Modes
4046 @section Addressing Modes
4047 @cindex addressing modes
4049 @c prevent bad page break with this line
4050 This is about addressing modes.
4052 @defmac HAVE_PRE_INCREMENT
4053 @defmacx HAVE_PRE_DECREMENT
4054 @defmacx HAVE_POST_INCREMENT
4055 @defmacx HAVE_POST_DECREMENT
4056 A C expression that is nonzero if the machine supports pre-increment,
4057 pre-decrement, post-increment, or post-decrement addressing respectively.
4060 @defmac HAVE_PRE_MODIFY_DISP
4061 @defmacx HAVE_POST_MODIFY_DISP
4062 A C expression that is nonzero if the machine supports pre- or
4063 post-address side-effect generation involving constants other than
4064 the size of the memory operand.
4067 @defmac HAVE_PRE_MODIFY_REG
4068 @defmacx HAVE_POST_MODIFY_REG
4069 A C expression that is nonzero if the machine supports pre- or
4070 post-address side-effect generation involving a register displacement.
4073 @defmac CONSTANT_ADDRESS_P (@var{x})
4074 A C expression that is 1 if the RTX @var{x} is a constant which
4075 is a valid address. On most machines the default definition of
4076 @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
4077 is acceptable, but a few machines are more restrictive as to which
4078 constant addresses are supported.
4081 @defmac CONSTANT_P (@var{x})
4082 @code{CONSTANT_P}, which is defined by target-independent code,
4083 accepts integer-values expressions whose values are not explicitly
4084 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4085 expressions and @code{const} arithmetic expressions, in addition to
4086 @code{const_int} and @code{const_double} expressions.
4089 @defmac MAX_REGS_PER_ADDRESS
4090 A number, the maximum number of registers that can appear in a valid
4091 memory address. Note that it is up to you to specify a value equal to
4092 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
4096 @hook TARGET_LEGITIMATE_ADDRESS_P
4098 @defmac TARGET_MEM_CONSTRAINT
4099 A single character to be used instead of the default @code{'m'}
4100 character for general memory addresses. This defines the constraint
4101 letter which matches the memory addresses accepted by
4102 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
4103 support new address formats in your back end without changing the
4104 semantics of the @code{'m'} constraint. This is necessary in order to
4105 preserve functionality of inline assembly constructs using the
4106 @code{'m'} constraint.
4109 @defmac FIND_BASE_TERM (@var{x})
4110 A C expression to determine the base term of address @var{x},
4111 or to provide a simplified version of @var{x} from which @file{alias.c}
4112 can easily find the base term. This macro is used in only two places:
4113 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
4115 It is always safe for this macro to not be defined. It exists so
4116 that alias analysis can understand machine-dependent addresses.
4118 The typical use of this macro is to handle addresses containing
4119 a label_ref or symbol_ref within an UNSPEC@.
4122 @hook TARGET_LEGITIMIZE_ADDRESS
4124 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4125 A C compound statement that attempts to replace @var{x}, which is an address
4126 that needs reloading, with a valid memory address for an operand of mode
4127 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
4128 It is not necessary to define this macro, but it might be useful for
4129 performance reasons.
4131 For example, on the i386, it is sometimes possible to use a single
4132 reload register instead of two by reloading a sum of two pseudo
4133 registers into a register. On the other hand, for number of RISC
4134 processors offsets are limited so that often an intermediate address
4135 needs to be generated in order to address a stack slot. By defining
4136 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4137 generated for adjacent some stack slots can be made identical, and thus
4140 @emph{Note}: This macro should be used with caution. It is necessary
4141 to know something of how reload works in order to effectively use this,
4142 and it is quite easy to produce macros that build in too much knowledge
4143 of reload internals.
4145 @emph{Note}: This macro must be able to reload an address created by a
4146 previous invocation of this macro. If it fails to handle such addresses
4147 then the compiler may generate incorrect code or abort.
4150 The macro definition should use @code{push_reload} to indicate parts that
4151 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4152 suitable to be passed unaltered to @code{push_reload}.
4154 The code generated by this macro must not alter the substructure of
4155 @var{x}. If it transforms @var{x} into a more legitimate form, it
4156 should assign @var{x} (which will always be a C variable) a new value.
4157 This also applies to parts that you change indirectly by calling
4160 @findex strict_memory_address_p
4161 The macro definition may use @code{strict_memory_address_p} to test if
4162 the address has become legitimate.
4165 If you want to change only a part of @var{x}, one standard way of doing
4166 this is to use @code{copy_rtx}. Note, however, that it unshares only a
4167 single level of rtl. Thus, if the part to be changed is not at the
4168 top level, you'll need to replace first the top level.
4169 It is not necessary for this macro to come up with a legitimate
4170 address; but often a machine-dependent strategy can generate better code.
4173 @hook TARGET_MODE_DEPENDENT_ADDRESS_P
4175 @hook TARGET_LEGITIMATE_CONSTANT_P
4177 @hook TARGET_DELEGITIMIZE_ADDRESS
4179 @hook TARGET_CONST_NOT_OK_FOR_DEBUG_P
4181 @hook TARGET_CANNOT_FORCE_CONST_MEM
4183 @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
4185 @hook TARGET_USE_BLOCKS_FOR_DECL_P
4187 @hook TARGET_BUILTIN_RECIPROCAL
4189 @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
4191 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
4193 @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
4195 @hook TARGET_VECTORIZE_VEC_PERM_CONST_OK
4197 @hook TARGET_VECTORIZE_BUILTIN_CONVERSION
4199 @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
4201 @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
4203 @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
4205 @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_SIZES
4207 @hook TARGET_VECTORIZE_INIT_COST
4209 @hook TARGET_VECTORIZE_ADD_STMT_COST
4211 @hook TARGET_VECTORIZE_FINISH_COST
4213 @hook TARGET_VECTORIZE_DESTROY_COST_DATA
4215 @hook TARGET_VECTORIZE_BUILTIN_TM_LOAD
4217 @hook TARGET_VECTORIZE_BUILTIN_TM_STORE
4219 @hook TARGET_VECTORIZE_BUILTIN_GATHER
4221 @hook TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN
4223 @hook TARGET_SIMD_CLONE_ADJUST
4225 @hook TARGET_SIMD_CLONE_USABLE
4227 @node Anchored Addresses
4228 @section Anchored Addresses
4229 @cindex anchored addresses
4230 @cindex @option{-fsection-anchors}
4232 GCC usually addresses every static object as a separate entity.
4233 For example, if we have:
4237 int foo (void) @{ return a + b + c; @}
4240 the code for @code{foo} will usually calculate three separate symbolic
4241 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
4242 it would be better to calculate just one symbolic address and access
4243 the three variables relative to it. The equivalent pseudocode would
4249 register int *xr = &x;
4250 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
4254 (which isn't valid C). We refer to shared addresses like @code{x} as
4255 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
4257 The hooks below describe the target properties that GCC needs to know
4258 in order to make effective use of section anchors. It won't use
4259 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
4260 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
4262 @hook TARGET_MIN_ANCHOR_OFFSET
4264 @hook TARGET_MAX_ANCHOR_OFFSET
4266 @hook TARGET_ASM_OUTPUT_ANCHOR
4268 @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
4270 @node Condition Code
4271 @section Condition Code Status
4272 @cindex condition code status
4274 The macros in this section can be split in two families, according to the
4275 two ways of representing condition codes in GCC.
4277 The first representation is the so called @code{(cc0)} representation
4278 (@pxref{Jump Patterns}), where all instructions can have an implicit
4279 clobber of the condition codes. The second is the condition code
4280 register representation, which provides better schedulability for
4281 architectures that do have a condition code register, but on which
4282 most instructions do not affect it. The latter category includes
4285 The implicit clobbering poses a strong restriction on the placement of
4286 the definition and use of the condition code. In the past the definition
4287 and use were always adjacent. However, recent changes to support trapping
4288 arithmatic may result in the definition and user being in different blocks.
4289 Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally,
4290 the definition may be the source of exception handling edges.
4292 These restrictions can prevent important
4293 optimizations on some machines. For example, on the IBM RS/6000, there
4294 is a delay for taken branches unless the condition code register is set
4295 three instructions earlier than the conditional branch. The instruction
4296 scheduler cannot perform this optimization if it is not permitted to
4297 separate the definition and use of the condition code register.
4299 For this reason, it is possible and suggested to use a register to
4300 represent the condition code for new ports. If there is a specific
4301 condition code register in the machine, use a hard register. If the
4302 condition code or comparison result can be placed in any general register,
4303 or if there are multiple condition registers, use a pseudo register.
4304 Registers used to store the condition code value will usually have a mode
4305 that is in class @code{MODE_CC}.
4307 Alternatively, you can use @code{BImode} if the comparison operator is
4308 specified already in the compare instruction. In this case, you are not
4309 interested in most macros in this section.
4312 * CC0 Condition Codes:: Old style representation of condition codes.
4313 * MODE_CC Condition Codes:: Modern representation of condition codes.
4316 @node CC0 Condition Codes
4317 @subsection Representation of condition codes using @code{(cc0)}
4321 The file @file{conditions.h} defines a variable @code{cc_status} to
4322 describe how the condition code was computed (in case the interpretation of
4323 the condition code depends on the instruction that it was set by). This
4324 variable contains the RTL expressions on which the condition code is
4325 currently based, and several standard flags.
4327 Sometimes additional machine-specific flags must be defined in the machine
4328 description header file. It can also add additional machine-specific
4329 information by defining @code{CC_STATUS_MDEP}.
4331 @defmac CC_STATUS_MDEP
4332 C code for a data type which is used for declaring the @code{mdep}
4333 component of @code{cc_status}. It defaults to @code{int}.
4335 This macro is not used on machines that do not use @code{cc0}.
4338 @defmac CC_STATUS_MDEP_INIT
4339 A C expression to initialize the @code{mdep} field to ``empty''.
4340 The default definition does nothing, since most machines don't use
4341 the field anyway. If you want to use the field, you should probably
4342 define this macro to initialize it.
4344 This macro is not used on machines that do not use @code{cc0}.
4347 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
4348 A C compound statement to set the components of @code{cc_status}
4349 appropriately for an insn @var{insn} whose body is @var{exp}. It is
4350 this macro's responsibility to recognize insns that set the condition
4351 code as a byproduct of other activity as well as those that explicitly
4354 This macro is not used on machines that do not use @code{cc0}.
4356 If there are insns that do not set the condition code but do alter
4357 other machine registers, this macro must check to see whether they
4358 invalidate the expressions that the condition code is recorded as
4359 reflecting. For example, on the 68000, insns that store in address
4360 registers do not set the condition code, which means that usually
4361 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
4362 insns. But suppose that the previous insn set the condition code
4363 based on location @samp{a4@@(102)} and the current insn stores a new
4364 value in @samp{a4}. Although the condition code is not changed by
4365 this, it will no longer be true that it reflects the contents of
4366 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
4367 @code{cc_status} in this case to say that nothing is known about the
4368 condition code value.
4370 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
4371 with the results of peephole optimization: insns whose patterns are
4372 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
4373 constants which are just the operands. The RTL structure of these
4374 insns is not sufficient to indicate what the insns actually do. What
4375 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
4376 @code{CC_STATUS_INIT}.
4378 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
4379 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
4380 @samp{cc}. This avoids having detailed information about patterns in
4381 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
4384 @node MODE_CC Condition Codes
4385 @subsection Representation of condition codes using registers
4389 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4390 On many machines, the condition code may be produced by other instructions
4391 than compares, for example the branch can use directly the condition
4392 code set by a subtract instruction. However, on some machines
4393 when the condition code is set this way some bits (such as the overflow
4394 bit) are not set in the same way as a test instruction, so that a different
4395 branch instruction must be used for some conditional branches. When
4396 this happens, use the machine mode of the condition code register to
4397 record different formats of the condition code register. Modes can
4398 also be used to record which compare instruction (e.g. a signed or an
4399 unsigned comparison) produced the condition codes.
4401 If other modes than @code{CCmode} are required, add them to
4402 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
4403 a mode given an operand of a compare. This is needed because the modes
4404 have to be chosen not only during RTL generation but also, for example,
4405 by instruction combination. The result of @code{SELECT_CC_MODE} should
4406 be consistent with the mode used in the patterns; for example to support
4407 the case of the add on the SPARC discussed above, we have the pattern
4411 [(set (reg:CC_NOOV 0)
4413 (plus:SI (match_operand:SI 0 "register_operand" "%r")
4414 (match_operand:SI 1 "arith_operand" "rI"))
4421 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
4422 for comparisons whose argument is a @code{plus}:
4425 #define SELECT_CC_MODE(OP,X,Y) \
4426 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
4427 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
4428 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
4429 || GET_CODE (X) == NEG) \
4430 ? CC_NOOVmode : CCmode))
4433 Another reason to use modes is to retain information on which operands
4434 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
4437 You should define this macro if and only if you define extra CC modes
4438 in @file{@var{machine}-modes.def}.
4441 @hook TARGET_CANONICALIZE_COMPARISON
4443 @defmac REVERSIBLE_CC_MODE (@var{mode})
4444 A C expression whose value is one if it is always safe to reverse a
4445 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
4446 can ever return @var{mode} for a floating-point inequality comparison,
4447 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4449 You need not define this macro if it would always returns zero or if the
4450 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4451 For example, here is the definition used on the SPARC, where floating-point
4452 inequality comparisons are always given @code{CCFPEmode}:
4455 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
4459 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
4460 A C expression whose value is reversed condition code of the @var{code} for
4461 comparison done in CC_MODE @var{mode}. The macro is used only in case
4462 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
4463 machine has some non-standard way how to reverse certain conditionals. For
4464 instance in case all floating point conditions are non-trapping, compiler may
4465 freely convert unordered compares to ordered one. Then definition may look
4469 #define REVERSE_CONDITION(CODE, MODE) \
4470 ((MODE) != CCFPmode ? reverse_condition (CODE) \
4471 : reverse_condition_maybe_unordered (CODE))
4475 @hook TARGET_FIXED_CONDITION_CODE_REGS
4477 @hook TARGET_CC_MODES_COMPATIBLE
4480 @section Describing Relative Costs of Operations
4481 @cindex costs of instructions
4482 @cindex relative costs
4483 @cindex speed of instructions
4485 These macros let you describe the relative speed of various operations
4486 on the target machine.
4488 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
4489 A C expression for the cost of moving data of mode @var{mode} from a
4490 register in class @var{from} to one in class @var{to}. The classes are
4491 expressed using the enumeration values such as @code{GENERAL_REGS}. A
4492 value of 2 is the default; other values are interpreted relative to
4495 It is not required that the cost always equal 2 when @var{from} is the
4496 same as @var{to}; on some machines it is expensive to move between
4497 registers if they are not general registers.
4499 If reload sees an insn consisting of a single @code{set} between two
4500 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4501 classes returns a value of 2, reload does not check to ensure that the
4502 constraints of the insn are met. Setting a cost of other than 2 will
4503 allow reload to verify that the constraints are met. You should do this
4504 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4506 These macros are obsolete, new ports should use the target hook
4507 @code{TARGET_REGISTER_MOVE_COST} instead.
4510 @hook TARGET_REGISTER_MOVE_COST
4512 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4513 A C expression for the cost of moving data of mode @var{mode} between a
4514 register of class @var{class} and memory; @var{in} is zero if the value
4515 is to be written to memory, nonzero if it is to be read in. This cost
4516 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
4517 registers and memory is more expensive than between two registers, you
4518 should define this macro to express the relative cost.
4520 If you do not define this macro, GCC uses a default cost of 4 plus
4521 the cost of copying via a secondary reload register, if one is
4522 needed. If your machine requires a secondary reload register to copy
4523 between memory and a register of @var{class} but the reload mechanism is
4524 more complex than copying via an intermediate, define this macro to
4525 reflect the actual cost of the move.
4527 GCC defines the function @code{memory_move_secondary_cost} if
4528 secondary reloads are needed. It computes the costs due to copying via
4529 a secondary register. If your machine copies from memory using a
4530 secondary register in the conventional way but the default base value of
4531 4 is not correct for your machine, define this macro to add some other
4532 value to the result of that function. The arguments to that function
4533 are the same as to this macro.
4535 These macros are obsolete, new ports should use the target hook
4536 @code{TARGET_MEMORY_MOVE_COST} instead.
4539 @hook TARGET_MEMORY_MOVE_COST
4541 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
4542 A C expression for the cost of a branch instruction. A value of 1 is
4543 the default; other values are interpreted relative to that. Parameter
4544 @var{speed_p} is true when the branch in question should be optimized
4545 for speed. When it is false, @code{BRANCH_COST} should return a value
4546 optimal for code size rather than performance. @var{predictable_p} is
4547 true for well-predicted branches. On many architectures the
4548 @code{BRANCH_COST} can be reduced then.
4551 Here are additional macros which do not specify precise relative costs,
4552 but only that certain actions are more expensive than GCC would
4555 @defmac SLOW_BYTE_ACCESS
4556 Define this macro as a C expression which is nonzero if accessing less
4557 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
4558 faster than accessing a word of memory, i.e., if such access
4559 require more than one instruction or if there is no difference in cost
4560 between byte and (aligned) word loads.
4562 When this macro is not defined, the compiler will access a field by
4563 finding the smallest containing object; when it is defined, a fullword
4564 load will be used if alignment permits. Unless bytes accesses are
4565 faster than word accesses, using word accesses is preferable since it
4566 may eliminate subsequent memory access if subsequent accesses occur to
4567 other fields in the same word of the structure, but to different bytes.
4570 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
4571 Define this macro to be the value 1 if memory accesses described by the
4572 @var{mode} and @var{alignment} parameters have a cost many times greater
4573 than aligned accesses, for example if they are emulated in a trap
4576 When this macro is nonzero, the compiler will act as if
4577 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
4578 moves. This can cause significantly more instructions to be produced.
4579 Therefore, do not set this macro nonzero if unaligned accesses only add a
4580 cycle or two to the time for a memory access.
4582 If the value of this macro is always zero, it need not be defined. If
4583 this macro is defined, it should produce a nonzero value when
4584 @code{STRICT_ALIGNMENT} is nonzero.
4587 @defmac MOVE_RATIO (@var{speed})
4588 The threshold of number of scalar memory-to-memory move insns, @emph{below}
4589 which a sequence of insns should be generated instead of a
4590 string move insn or a library call. Increasing the value will always
4591 make code faster, but eventually incurs high cost in increased code size.
4593 Note that on machines where the corresponding move insn is a
4594 @code{define_expand} that emits a sequence of insns, this macro counts
4595 the number of such sequences.
4597 The parameter @var{speed} is true if the code is currently being
4598 optimized for speed rather than size.
4600 If you don't define this, a reasonable default is used.
4603 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
4604 A C expression used to determine whether @code{move_by_pieces} will be used to
4605 copy a chunk of memory, or whether some other block move mechanism
4606 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4607 than @code{MOVE_RATIO}.
4610 @defmac MOVE_MAX_PIECES
4611 A C expression used by @code{move_by_pieces} to determine the largest unit
4612 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
4615 @defmac CLEAR_RATIO (@var{speed})
4616 The threshold of number of scalar move insns, @emph{below} which a sequence
4617 of insns should be generated to clear memory instead of a string clear insn
4618 or a library call. Increasing the value will always make code faster, but
4619 eventually incurs high cost in increased code size.
4621 The parameter @var{speed} is true if the code is currently being
4622 optimized for speed rather than size.
4624 If you don't define this, a reasonable default is used.
4627 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
4628 A C expression used to determine whether @code{clear_by_pieces} will be used
4629 to clear a chunk of memory, or whether some other block clear mechanism
4630 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4631 than @code{CLEAR_RATIO}.
4634 @defmac SET_RATIO (@var{speed})
4635 The threshold of number of scalar move insns, @emph{below} which a sequence
4636 of insns should be generated to set memory to a constant value, instead of
4637 a block set insn or a library call.
4638 Increasing the value will always make code faster, but
4639 eventually incurs high cost in increased code size.
4641 The parameter @var{speed} is true if the code is currently being
4642 optimized for speed rather than size.
4644 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
4647 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
4648 A C expression used to determine whether @code{store_by_pieces} will be
4649 used to set a chunk of memory to a constant value, or whether some
4650 other mechanism will be used. Used by @code{__builtin_memset} when
4651 storing values other than constant zero.
4652 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4653 than @code{SET_RATIO}.
4656 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
4657 A C expression used to determine whether @code{store_by_pieces} will be
4658 used to set a chunk of memory to a constant string value, or whether some
4659 other mechanism will be used. Used by @code{__builtin_strcpy} when
4660 called with a constant source string.
4661 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
4662 than @code{MOVE_RATIO}.
4665 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
4666 A C expression used to determine whether a load postincrement is a good
4667 thing to use for a given mode. Defaults to the value of
4668 @code{HAVE_POST_INCREMENT}.
4671 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
4672 A C expression used to determine whether a load postdecrement is a good
4673 thing to use for a given mode. Defaults to the value of
4674 @code{HAVE_POST_DECREMENT}.
4677 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
4678 A C expression used to determine whether a load preincrement is a good
4679 thing to use for a given mode. Defaults to the value of
4680 @code{HAVE_PRE_INCREMENT}.
4683 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
4684 A C expression used to determine whether a load predecrement is a good
4685 thing to use for a given mode. Defaults to the value of
4686 @code{HAVE_PRE_DECREMENT}.
4689 @defmac USE_STORE_POST_INCREMENT (@var{mode})
4690 A C expression used to determine whether a store postincrement is a good
4691 thing to use for a given mode. Defaults to the value of
4692 @code{HAVE_POST_INCREMENT}.
4695 @defmac USE_STORE_POST_DECREMENT (@var{mode})
4696 A C expression used to determine whether a store postdecrement is a good
4697 thing to use for a given mode. Defaults to the value of
4698 @code{HAVE_POST_DECREMENT}.
4701 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
4702 This macro is used to determine whether a store preincrement is a good
4703 thing to use for a given mode. Defaults to the value of
4704 @code{HAVE_PRE_INCREMENT}.
4707 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
4708 This macro is used to determine whether a store predecrement is a good
4709 thing to use for a given mode. Defaults to the value of
4710 @code{HAVE_PRE_DECREMENT}.
4713 @defmac NO_FUNCTION_CSE
4714 Define this macro if it is as good or better to call a constant
4715 function address than to call an address kept in a register.
4718 @defmac LOGICAL_OP_NON_SHORT_CIRCUIT
4719 Define this macro if a non-short-circuit operation produced by
4720 @samp{fold_range_test ()} is optimal. This macro defaults to true if
4721 @code{BRANCH_COST} is greater than or equal to the value 2.
4724 @hook TARGET_RTX_COSTS
4726 @hook TARGET_ADDRESS_COST
4729 @section Adjusting the Instruction Scheduler
4731 The instruction scheduler may need a fair amount of machine-specific
4732 adjustment in order to produce good code. GCC provides several target
4733 hooks for this purpose. It is usually enough to define just a few of
4734 them: try the first ones in this list first.
4736 @hook TARGET_SCHED_ISSUE_RATE
4738 @hook TARGET_SCHED_VARIABLE_ISSUE
4740 @hook TARGET_SCHED_ADJUST_COST
4742 @hook TARGET_SCHED_ADJUST_PRIORITY
4744 @hook TARGET_SCHED_REORDER
4746 @hook TARGET_SCHED_REORDER2
4748 @hook TARGET_SCHED_MACRO_FUSION_P
4750 @hook TARGET_SCHED_MACRO_FUSION_PAIR_P
4752 @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
4754 @hook TARGET_SCHED_INIT
4756 @hook TARGET_SCHED_FINISH
4758 @hook TARGET_SCHED_INIT_GLOBAL
4760 @hook TARGET_SCHED_FINISH_GLOBAL
4762 @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
4764 @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
4766 @hook TARGET_SCHED_DFA_POST_CYCLE_INSN
4768 @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
4770 @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
4772 @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
4774 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
4776 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
4778 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
4780 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
4782 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
4784 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
4786 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
4788 @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
4790 @hook TARGET_SCHED_DFA_NEW_CYCLE
4792 @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
4794 @hook TARGET_SCHED_H_I_D_EXTENDED
4796 @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
4798 @hook TARGET_SCHED_INIT_SCHED_CONTEXT
4800 @hook TARGET_SCHED_SET_SCHED_CONTEXT
4802 @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
4804 @hook TARGET_SCHED_FREE_SCHED_CONTEXT
4806 @hook TARGET_SCHED_SPECULATE_INSN
4808 @hook TARGET_SCHED_NEEDS_BLOCK_P
4810 @hook TARGET_SCHED_GEN_SPEC_CHECK
4812 @hook TARGET_SCHED_SET_SCHED_FLAGS
4814 @hook TARGET_SCHED_SMS_RES_MII
4816 @hook TARGET_SCHED_DISPATCH
4818 @hook TARGET_SCHED_DISPATCH_DO
4820 @hook TARGET_SCHED_EXPOSED_PIPELINE
4822 @hook TARGET_SCHED_REASSOCIATION_WIDTH
4825 @section Dividing the Output into Sections (Texts, Data, @dots{})
4826 @c the above section title is WAY too long. maybe cut the part between
4827 @c the (...)? --mew 10feb93
4829 An object file is divided into sections containing different types of
4830 data. In the most common case, there are three sections: the @dfn{text
4831 section}, which holds instructions and read-only data; the @dfn{data
4832 section}, which holds initialized writable data; and the @dfn{bss
4833 section}, which holds uninitialized data. Some systems have other kinds
4836 @file{varasm.c} provides several well-known sections, such as
4837 @code{text_section}, @code{data_section} and @code{bss_section}.
4838 The normal way of controlling a @code{@var{foo}_section} variable
4839 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
4840 as described below. The macros are only read once, when @file{varasm.c}
4841 initializes itself, so their values must be run-time constants.
4842 They may however depend on command-line flags.
4844 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
4845 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
4846 to be string literals.
4848 Some assemblers require a different string to be written every time a
4849 section is selected. If your assembler falls into this category, you
4850 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
4851 @code{get_unnamed_section} to set up the sections.
4853 You must always create a @code{text_section}, either by defining
4854 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
4855 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
4856 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
4857 create a distinct @code{readonly_data_section}, the default is to
4858 reuse @code{text_section}.
4860 All the other @file{varasm.c} sections are optional, and are null
4861 if the target does not provide them.
4863 @defmac TEXT_SECTION_ASM_OP
4864 A C expression whose value is a string, including spacing, containing the
4865 assembler operation that should precede instructions and read-only data.
4866 Normally @code{"\t.text"} is right.
4869 @defmac HOT_TEXT_SECTION_NAME
4870 If defined, a C string constant for the name of the section containing most
4871 frequently executed functions of the program. If not defined, GCC will provide
4872 a default definition if the target supports named sections.
4875 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
4876 If defined, a C string constant for the name of the section containing unlikely
4877 executed functions in the program.
4880 @defmac DATA_SECTION_ASM_OP
4881 A C expression whose value is a string, including spacing, containing the
4882 assembler operation to identify the following data as writable initialized
4883 data. Normally @code{"\t.data"} is right.
4886 @defmac SDATA_SECTION_ASM_OP
4887 If defined, a C expression whose value is a string, including spacing,
4888 containing the assembler operation to identify the following data as
4889 initialized, writable small data.
4892 @defmac READONLY_DATA_SECTION_ASM_OP
4893 A C expression whose value is a string, including spacing, containing the
4894 assembler operation to identify the following data as read-only initialized
4898 @defmac BSS_SECTION_ASM_OP
4899 If defined, a C expression whose value is a string, including spacing,
4900 containing the assembler operation to identify the following data as
4901 uninitialized global data. If not defined, and
4902 @code{ASM_OUTPUT_ALIGNED_BSS} not defined,
4903 uninitialized global data will be output in the data section if
4904 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
4908 @defmac SBSS_SECTION_ASM_OP
4909 If defined, a C expression whose value is a string, including spacing,
4910 containing the assembler operation to identify the following data as
4911 uninitialized, writable small data.
4914 @defmac TLS_COMMON_ASM_OP
4915 If defined, a C expression whose value is a string containing the
4916 assembler operation to identify the following data as thread-local
4917 common data. The default is @code{".tls_common"}.
4920 @defmac TLS_SECTION_ASM_FLAG
4921 If defined, a C expression whose value is a character constant
4922 containing the flag used to mark a section as a TLS section. The
4923 default is @code{'T'}.
4926 @defmac INIT_SECTION_ASM_OP
4927 If defined, a C expression whose value is a string, including spacing,
4928 containing the assembler operation to identify the following data as
4929 initialization code. If not defined, GCC will assume such a section does
4930 not exist. This section has no corresponding @code{init_section}
4931 variable; it is used entirely in runtime code.
4934 @defmac FINI_SECTION_ASM_OP
4935 If defined, a C expression whose value is a string, including spacing,
4936 containing the assembler operation to identify the following data as
4937 finalization code. If not defined, GCC will assume such a section does
4938 not exist. This section has no corresponding @code{fini_section}
4939 variable; it is used entirely in runtime code.
4942 @defmac INIT_ARRAY_SECTION_ASM_OP
4943 If defined, a C expression whose value is a string, including spacing,
4944 containing the assembler operation to identify the following data as
4945 part of the @code{.init_array} (or equivalent) section. If not
4946 defined, GCC will assume such a section does not exist. Do not define
4947 both this macro and @code{INIT_SECTION_ASM_OP}.
4950 @defmac FINI_ARRAY_SECTION_ASM_OP
4951 If defined, a C expression whose value is a string, including spacing,
4952 containing the assembler operation to identify the following data as
4953 part of the @code{.fini_array} (or equivalent) section. If not
4954 defined, GCC will assume such a section does not exist. Do not define
4955 both this macro and @code{FINI_SECTION_ASM_OP}.
4958 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
4959 If defined, an ASM statement that switches to a different section
4960 via @var{section_op}, calls @var{function}, and switches back to
4961 the text section. This is used in @file{crtstuff.c} if
4962 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
4963 to initialization and finalization functions from the init and fini
4964 sections. By default, this macro uses a simple function call. Some
4965 ports need hand-crafted assembly code to avoid dependencies on
4966 registers initialized in the function prologue or to ensure that
4967 constant pools don't end up too far way in the text section.
4970 @defmac TARGET_LIBGCC_SDATA_SECTION
4971 If defined, a string which names the section into which small
4972 variables defined in crtstuff and libgcc should go. This is useful
4973 when the target has options for optimizing access to small data, and
4974 you want the crtstuff and libgcc routines to be conservative in what
4975 they expect of your application yet liberal in what your application
4976 expects. For example, for targets with a @code{.sdata} section (like
4977 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
4978 require small data support from your application, but use this macro
4979 to put small data into @code{.sdata} so that your application can
4980 access these variables whether it uses small data or not.
4983 @defmac FORCE_CODE_SECTION_ALIGN
4984 If defined, an ASM statement that aligns a code section to some
4985 arbitrary boundary. This is used to force all fragments of the
4986 @code{.init} and @code{.fini} sections to have to same alignment
4987 and thus prevent the linker from having to add any padding.
4990 @defmac JUMP_TABLES_IN_TEXT_SECTION
4991 Define this macro to be an expression with a nonzero value if jump
4992 tables (for @code{tablejump} insns) should be output in the text
4993 section, along with the assembler instructions. Otherwise, the
4994 readonly data section is used.
4996 This macro is irrelevant if there is no separate readonly data section.
4999 @hook TARGET_ASM_INIT_SECTIONS
5001 @hook TARGET_ASM_RELOC_RW_MASK
5003 @hook TARGET_ASM_SELECT_SECTION
5005 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
5006 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
5007 for @code{FUNCTION_DECL}s as well as for variables and constants.
5009 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
5010 function has been determined to be likely to be called, and nonzero if
5011 it is unlikely to be called.
5014 @hook TARGET_ASM_UNIQUE_SECTION
5016 @hook TARGET_ASM_FUNCTION_RODATA_SECTION
5018 @hook TARGET_ASM_MERGEABLE_RODATA_PREFIX
5020 @hook TARGET_ASM_TM_CLONE_TABLE_SECTION
5022 @hook TARGET_ASM_SELECT_RTX_SECTION
5024 @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
5026 @hook TARGET_ENCODE_SECTION_INFO
5028 @hook TARGET_STRIP_NAME_ENCODING
5030 @hook TARGET_IN_SMALL_DATA_P
5032 @hook TARGET_HAVE_SRODATA_SECTION
5034 @hook TARGET_PROFILE_BEFORE_PROLOGUE
5036 @hook TARGET_BINDS_LOCAL_P
5038 @hook TARGET_HAVE_TLS
5042 @section Position Independent Code
5043 @cindex position independent code
5046 This section describes macros that help implement generation of position
5047 independent code. Simply defining these macros is not enough to
5048 generate valid PIC; you must also add support to the hook
5049 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
5050 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
5051 must modify the definition of @samp{movsi} to do something appropriate
5052 when the source operand contains a symbolic address. You may also
5053 need to alter the handling of switch statements so that they use
5055 @c i rearranged the order of the macros above to try to force one of
5056 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
5058 @defmac PIC_OFFSET_TABLE_REGNUM
5059 The register number of the register used to address a table of static
5060 data addresses in memory. In some cases this register is defined by a
5061 processor's ``application binary interface'' (ABI)@. When this macro
5062 is defined, RTL is generated for this register once, as with the stack
5063 pointer and frame pointer registers. If this macro is not defined, it
5064 is up to the machine-dependent files to allocate such a register (if
5065 necessary). Note that this register must be fixed when in use (e.g.@:
5066 when @code{flag_pic} is true).
5069 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
5070 A C expression that is nonzero if the register defined by
5071 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined,
5072 the default is zero. Do not define
5073 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
5076 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
5077 A C expression that is nonzero if @var{x} is a legitimate immediate
5078 operand on the target machine when generating position independent code.
5079 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5080 check this. You can also assume @var{flag_pic} is true, so you need not
5081 check it either. You need not define this macro if all constants
5082 (including @code{SYMBOL_REF}) can be immediate operands when generating
5083 position independent code.
5086 @node Assembler Format
5087 @section Defining the Output Assembler Language
5089 This section describes macros whose principal purpose is to describe how
5090 to write instructions in assembler language---rather than what the
5094 * File Framework:: Structural information for the assembler file.
5095 * Data Output:: Output of constants (numbers, strings, addresses).
5096 * Uninitialized Data:: Output of uninitialized variables.
5097 * Label Output:: Output and generation of labels.
5098 * Initialization:: General principles of initialization
5099 and termination routines.
5100 * Macros for Initialization::
5101 Specific macros that control the handling of
5102 initialization and termination routines.
5103 * Instruction Output:: Output of actual instructions.
5104 * Dispatch Tables:: Output of jump tables.
5105 * Exception Region Output:: Output of exception region code.
5106 * Alignment Output:: Pseudo ops for alignment and skipping data.
5109 @node File Framework
5110 @subsection The Overall Framework of an Assembler File
5111 @cindex assembler format
5112 @cindex output of assembler code
5114 @c prevent bad page break with this line
5115 This describes the overall framework of an assembly file.
5117 @findex default_file_start
5118 @hook TARGET_ASM_FILE_START
5120 @hook TARGET_ASM_FILE_START_APP_OFF
5122 @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
5124 @hook TARGET_ASM_FILE_END
5126 @deftypefun void file_end_indicate_exec_stack ()
5127 Some systems use a common convention, the @samp{.note.GNU-stack}
5128 special section, to indicate whether or not an object file relies on
5129 the stack being executable. If your system uses this convention, you
5130 should define @code{TARGET_ASM_FILE_END} to this function. If you
5131 need to do other things in that hook, have your hook function call
5135 @hook TARGET_ASM_LTO_START
5137 @hook TARGET_ASM_LTO_END
5139 @hook TARGET_ASM_CODE_END
5141 @defmac ASM_COMMENT_START
5142 A C string constant describing how to begin a comment in the target
5143 assembler language. The compiler assumes that the comment will end at
5144 the end of the line.
5148 A C string constant for text to be output before each @code{asm}
5149 statement or group of consecutive ones. Normally this is
5150 @code{"#APP"}, which is a comment that has no effect on most
5151 assemblers but tells the GNU assembler that it must check the lines
5152 that follow for all valid assembler constructs.
5156 A C string constant for text to be output after each @code{asm}
5157 statement or group of consecutive ones. Normally this is
5158 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
5159 time-saving assumptions that are valid for ordinary compiler output.
5162 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5163 A C statement to output COFF information or DWARF debugging information
5164 which indicates that filename @var{name} is the current source file to
5165 the stdio stream @var{stream}.
5167 This macro need not be defined if the standard form of output
5168 for the file format in use is appropriate.
5171 @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
5173 @hook TARGET_ASM_OUTPUT_IDENT
5175 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5176 A C statement to output the string @var{string} to the stdio stream
5177 @var{stream}. If you do not call the function @code{output_quoted_string}
5178 in your config files, GCC will only call it to output filenames to
5179 the assembler source. So you can use it to canonicalize the format
5180 of the filename using this macro.
5183 @hook TARGET_ASM_NAMED_SECTION
5185 @hook TARGET_ASM_FUNCTION_SECTION
5187 @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
5189 @hook TARGET_HAVE_NAMED_SECTIONS
5190 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
5191 It must not be modified by command-line option processing.
5194 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
5195 @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
5197 @hook TARGET_SECTION_TYPE_FLAGS
5199 @hook TARGET_ASM_RECORD_GCC_SWITCHES
5201 @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
5205 @subsection Output of Data
5208 @hook TARGET_ASM_BYTE_OP
5210 @hook TARGET_ASM_INTEGER
5212 @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
5214 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5215 A C statement to output to the stdio stream @var{stream} an assembler
5216 instruction to assemble a string constant containing the @var{len}
5217 bytes at @var{ptr}. @var{ptr} will be a C expression of type
5218 @code{char *} and @var{len} a C expression of type @code{int}.
5220 If the assembler has a @code{.ascii} pseudo-op as found in the
5221 Berkeley Unix assembler, do not define the macro
5222 @code{ASM_OUTPUT_ASCII}.
5225 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
5226 A C statement to output word @var{n} of a function descriptor for
5227 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
5228 is defined, and is otherwise unused.
5231 @defmac CONSTANT_POOL_BEFORE_FUNCTION
5232 You may define this macro as a C expression. You should define the
5233 expression to have a nonzero value if GCC should output the constant
5234 pool for a function before the code for the function, or a zero value if
5235 GCC should output the constant pool after the function. If you do
5236 not define this macro, the usual case, GCC will output the constant
5237 pool before the function.
5240 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5241 A C statement to output assembler commands to define the start of the
5242 constant pool for a function. @var{funname} is a string giving
5243 the name of the function. Should the return type of the function
5244 be required, it can be obtained via @var{fundecl}. @var{size}
5245 is the size, in bytes, of the constant pool that will be written
5246 immediately after this call.
5248 If no constant-pool prefix is required, the usual case, this macro need
5252 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5253 A C statement (with or without semicolon) to output a constant in the
5254 constant pool, if it needs special treatment. (This macro need not do
5255 anything for RTL expressions that can be output normally.)
5257 The argument @var{file} is the standard I/O stream to output the
5258 assembler code on. @var{x} is the RTL expression for the constant to
5259 output, and @var{mode} is the machine mode (in case @var{x} is a
5260 @samp{const_int}). @var{align} is the required alignment for the value
5261 @var{x}; you should output an assembler directive to force this much
5264 The argument @var{labelno} is a number to use in an internal label for
5265 the address of this pool entry. The definition of this macro is
5266 responsible for outputting the label definition at the proper place.
5267 Here is how to do this:
5270 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
5273 When you output a pool entry specially, you should end with a
5274 @code{goto} to the label @var{jumpto}. This will prevent the same pool
5275 entry from being output a second time in the usual manner.
5277 You need not define this macro if it would do nothing.
5280 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5281 A C statement to output assembler commands to at the end of the constant
5282 pool for a function. @var{funname} is a string giving the name of the
5283 function. Should the return type of the function be required, you can
5284 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
5285 constant pool that GCC wrote immediately before this call.
5287 If no constant-pool epilogue is required, the usual case, you need not
5291 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
5292 Define this macro as a C expression which is nonzero if @var{C} is
5293 used as a logical line separator by the assembler. @var{STR} points
5294 to the position in the string where @var{C} was found; this can be used if
5295 a line separator uses multiple characters.
5297 If you do not define this macro, the default is that only
5298 the character @samp{;} is treated as a logical line separator.
5301 @hook TARGET_ASM_OPEN_PAREN
5303 These macros are provided by @file{real.h} for writing the definitions
5304 of @code{ASM_OUTPUT_DOUBLE} and the like:
5306 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5307 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5308 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5309 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
5310 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
5311 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
5312 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
5313 target's floating point representation, and store its bit pattern in
5314 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
5315 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
5316 simple @code{long int}. For the others, it should be an array of
5317 @code{long int}. The number of elements in this array is determined
5318 by the size of the desired target floating point data type: 32 bits of
5319 it go in each @code{long int} array element. Each array element holds
5320 32 bits of the result, even if @code{long int} is wider than 32 bits
5321 on the host machine.
5323 The array element values are designed so that you can print them out
5324 using @code{fprintf} in the order they should appear in the target
5328 @node Uninitialized Data
5329 @subsection Output of Uninitialized Variables
5331 Each of the macros in this section is used to do the whole job of
5332 outputting a single uninitialized variable.
5334 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5335 A C statement (sans semicolon) to output to the stdio stream
5336 @var{stream} the assembler definition of a common-label named
5337 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5338 is the size rounded up to whatever alignment the caller wants. It is
5339 possible that @var{size} may be zero, for instance if a struct with no
5340 other member than a zero-length array is defined. In this case, the
5341 backend must output a symbol definition that allocates at least one
5342 byte, both so that the address of the resulting object does not compare
5343 equal to any other, and because some object formats cannot even express
5344 the concept of a zero-sized common symbol, as that is how they represent
5345 an ordinary undefined external.
5347 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5348 output the name itself; before and after that, output the additional
5349 assembler syntax for defining the name, and a newline.
5351 This macro controls how the assembler definitions of uninitialized
5352 common global variables are output.
5355 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5356 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5357 separate, explicit argument. If you define this macro, it is used in
5358 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5359 handling the required alignment of the variable. The alignment is specified
5360 as the number of bits.
5363 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5364 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5365 variable to be output, if there is one, or @code{NULL_TREE} if there
5366 is no corresponding variable. If you define this macro, GCC will use it
5367 in place of both @code{ASM_OUTPUT_COMMON} and
5368 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
5369 the variable's decl in order to chose what to output.
5372 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5373 A C statement (sans semicolon) to output to the stdio stream
5374 @var{stream} the assembler definition of uninitialized global @var{decl} named
5375 @var{name} whose size is @var{size} bytes. The variable @var{alignment}
5376 is the alignment specified as the number of bits.
5378 Try to use function @code{asm_output_aligned_bss} defined in file
5379 @file{varasm.c} when defining this macro. If unable, use the expression
5380 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5381 before and after that, output the additional assembler syntax for defining
5382 the name, and a newline.
5384 There are two ways of handling global BSS@. One is to define this macro.
5385 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
5386 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
5387 You do not need to do both.
5389 Some languages do not have @code{common} data, and require a
5390 non-common form of global BSS in order to handle uninitialized globals
5391 efficiently. C++ is one example of this. However, if the target does
5392 not support global BSS, the front end may choose to make globals
5393 common in order to save space in the object file.
5396 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5397 A C statement (sans semicolon) to output to the stdio stream
5398 @var{stream} the assembler definition of a local-common-label named
5399 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
5400 is the size rounded up to whatever alignment the caller wants.
5402 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5403 output the name itself; before and after that, output the additional
5404 assembler syntax for defining the name, and a newline.
5406 This macro controls how the assembler definitions of uninitialized
5407 static variables are output.
5410 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5411 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5412 separate, explicit argument. If you define this macro, it is used in
5413 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5414 handling the required alignment of the variable. The alignment is specified
5415 as the number of bits.
5418 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5419 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
5420 variable to be output, if there is one, or @code{NULL_TREE} if there
5421 is no corresponding variable. If you define this macro, GCC will use it
5422 in place of both @code{ASM_OUTPUT_DECL} and
5423 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
5424 the variable's decl in order to chose what to output.
5428 @subsection Output and Generation of Labels
5430 @c prevent bad page break with this line
5431 This is about outputting labels.
5433 @findex assemble_name
5434 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5435 A C statement (sans semicolon) to output to the stdio stream
5436 @var{stream} the assembler definition of a label named @var{name}.
5437 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5438 output the name itself; before and after that, output the additional
5439 assembler syntax for defining the name, and a newline. A default
5440 definition of this macro is provided which is correct for most systems.
5443 @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
5444 A C statement (sans semicolon) to output to the stdio stream
5445 @var{stream} the assembler definition of a label named @var{name} of
5447 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5448 output the name itself; before and after that, output the additional
5449 assembler syntax for defining the name, and a newline. A default
5450 definition of this macro is provided which is correct for most systems.
5452 If this macro is not defined, then the function name is defined in the
5453 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5456 @findex assemble_name_raw
5457 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
5458 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
5459 to refer to a compiler-generated label. The default definition uses
5460 @code{assemble_name_raw}, which is like @code{assemble_name} except
5461 that it is more efficient.
5465 A C string containing the appropriate assembler directive to specify the
5466 size of a symbol, without any arguments. On systems that use ELF, the
5467 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
5468 systems, the default is not to define this macro.
5470 Define this macro only if it is correct to use the default definitions
5471 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
5472 for your system. If you need your own custom definitions of those
5473 macros, or if you do not need explicit symbol sizes at all, do not
5477 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
5478 A C statement (sans semicolon) to output to the stdio stream
5479 @var{stream} a directive telling the assembler that the size of the
5480 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
5481 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5485 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
5486 A C statement (sans semicolon) to output to the stdio stream
5487 @var{stream} a directive telling the assembler to calculate the size of
5488 the symbol @var{name} by subtracting its address from the current
5491 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5492 provided. The default assumes that the assembler recognizes a special
5493 @samp{.} symbol as referring to the current address, and can calculate
5494 the difference between this and another symbol. If your assembler does
5495 not recognize @samp{.} or cannot do calculations with it, you will need
5496 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
5499 @defmac NO_DOLLAR_IN_LABEL
5500 Define this macro if the assembler does not accept the character
5501 @samp{$} in label names. By default constructors and destructors in
5502 G++ have @samp{$} in the identifiers. If this macro is defined,
5503 @samp{.} is used instead.
5506 @defmac NO_DOT_IN_LABEL
5507 Define this macro if the assembler does not accept the character
5508 @samp{.} in label names. By default constructors and destructors in G++
5509 have names that use @samp{.}. If this macro is defined, these names
5510 are rewritten to avoid @samp{.}.
5514 A C string containing the appropriate assembler directive to specify the
5515 type of a symbol, without any arguments. On systems that use ELF, the
5516 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
5517 systems, the default is not to define this macro.
5519 Define this macro only if it is correct to use the default definition of
5520 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5521 custom definition of this macro, or if you do not need explicit symbol
5522 types at all, do not define this macro.
5525 @defmac TYPE_OPERAND_FMT
5526 A C string which specifies (using @code{printf} syntax) the format of
5527 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
5528 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
5529 the default is not to define this macro.
5531 Define this macro only if it is correct to use the default definition of
5532 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
5533 custom definition of this macro, or if you do not need explicit symbol
5534 types at all, do not define this macro.
5537 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
5538 A C statement (sans semicolon) to output to the stdio stream
5539 @var{stream} a directive telling the assembler that the type of the
5540 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
5541 that string is always either @samp{"function"} or @samp{"object"}, but
5542 you should not count on this.
5544 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
5545 definition of this macro is provided.
5548 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5549 A C statement (sans semicolon) to output to the stdio stream
5550 @var{stream} any text necessary for declaring the name @var{name} of a
5551 function which is being defined. This macro is responsible for
5552 outputting the label definition (perhaps using
5553 @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the
5554 @code{FUNCTION_DECL} tree node representing the function.
5556 If this macro is not defined, then the function name is defined in the
5557 usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
5559 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5563 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5564 A C statement (sans semicolon) to output to the stdio stream
5565 @var{stream} any text necessary for declaring the size of a function
5566 which is being defined. The argument @var{name} is the name of the
5567 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5568 representing the function.
5570 If this macro is not defined, then the function size is not defined.
5572 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5576 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5577 A C statement (sans semicolon) to output to the stdio stream
5578 @var{stream} any text necessary for declaring the name @var{name} of an
5579 initialized variable which is being defined. This macro must output the
5580 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
5581 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
5583 If this macro is not defined, then the variable name is defined in the
5584 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5586 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
5587 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
5590 @hook TARGET_ASM_DECLARE_CONSTANT_NAME
5592 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5593 A C statement (sans semicolon) to output to the stdio stream
5594 @var{stream} any text necessary for claiming a register @var{regno}
5595 for a global variable @var{decl} with name @var{name}.
5597 If you don't define this macro, that is equivalent to defining it to do
5601 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5602 A C statement (sans semicolon) to finish up declaring a variable name
5603 once the compiler has processed its initializer fully and thus has had a
5604 chance to determine the size of an array when controlled by an
5605 initializer. This is used on systems where it's necessary to declare
5606 something about the size of the object.
5608 If you don't define this macro, that is equivalent to defining it to do
5611 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
5612 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
5615 @hook TARGET_ASM_GLOBALIZE_LABEL
5617 @hook TARGET_ASM_GLOBALIZE_DECL_NAME
5619 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
5620 A C statement (sans semicolon) to output to the stdio stream
5621 @var{stream} some commands that will make the label @var{name} weak;
5622 that is, available for reference from other files but only used if
5623 no other definition is available. Use the expression
5624 @code{assemble_name (@var{stream}, @var{name})} to output the name
5625 itself; before and after that, output the additional assembler syntax
5626 for making that name weak, and a newline.
5628 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
5629 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
5633 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
5634 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
5635 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
5636 or variable decl. If @var{value} is not @code{NULL}, this C statement
5637 should output to the stdio stream @var{stream} assembler code which
5638 defines (equates) the weak symbol @var{name} to have the value
5639 @var{value}. If @var{value} is @code{NULL}, it should output commands
5640 to make @var{name} weak.
5643 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
5644 Outputs a directive that enables @var{name} to be used to refer to
5645 symbol @var{value} with weak-symbol semantics. @code{decl} is the
5646 declaration of @code{name}.
5649 @defmac SUPPORTS_WEAK
5650 A preprocessor constant expression which evaluates to true if the target
5651 supports weak symbols.
5653 If you don't define this macro, @file{defaults.h} provides a default
5654 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
5655 is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
5658 @defmac TARGET_SUPPORTS_WEAK
5659 A C expression which evaluates to true if the target supports weak symbols.
5661 If you don't define this macro, @file{defaults.h} provides a default
5662 definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define
5663 this macro if you want to control weak symbol support with a compiler
5664 flag such as @option{-melf}.
5667 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
5668 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5669 public symbol such that extra copies in multiple translation units will
5670 be discarded by the linker. Define this macro if your object file
5671 format provides support for this concept, such as the @samp{COMDAT}
5672 section flags in the Microsoft Windows PE/COFF format, and this support
5673 requires changes to @var{decl}, such as putting it in a separate section.
5676 @defmac SUPPORTS_ONE_ONLY
5677 A C expression which evaluates to true if the target supports one-only
5680 If you don't define this macro, @file{varasm.c} provides a default
5681 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5682 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
5683 you want to control one-only symbol support with a compiler flag, or if
5684 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5685 be emitted as one-only.
5688 @hook TARGET_ASM_ASSEMBLE_VISIBILITY
5690 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
5691 A C expression that evaluates to true if the target's linker expects
5692 that weak symbols do not appear in a static archive's table of contents.
5693 The default is @code{0}.
5695 Leaving weak symbols out of an archive's table of contents means that,
5696 if a symbol will only have a definition in one translation unit and
5697 will have undefined references from other translation units, that
5698 symbol should not be weak. Defining this macro to be nonzero will
5699 thus have the effect that certain symbols that would normally be weak
5700 (explicit template instantiations, and vtables for polymorphic classes
5701 with noninline key methods) will instead be nonweak.
5703 The C++ ABI requires this macro to be zero. Define this macro for
5704 targets where full C++ ABI compliance is impossible and where linker
5705 restrictions require weak symbols to be left out of a static archive's
5709 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5710 A C statement (sans semicolon) to output to the stdio stream
5711 @var{stream} any text necessary for declaring the name of an external
5712 symbol named @var{name} which is referenced in this compilation but
5713 not defined. The value of @var{decl} is the tree node for the
5716 This macro need not be defined if it does not need to output anything.
5717 The GNU assembler and most Unix assemblers don't require anything.
5720 @hook TARGET_ASM_EXTERNAL_LIBCALL
5722 @hook TARGET_ASM_MARK_DECL_PRESERVED
5724 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5725 A C statement (sans semicolon) to output to the stdio stream
5726 @var{stream} a reference in assembler syntax to a label named
5727 @var{name}. This should add @samp{_} to the front of the name, if that
5728 is customary on your operating system, as it is in most Berkeley Unix
5729 systems. This macro is used in @code{assemble_name}.
5732 @hook TARGET_MANGLE_ASSEMBLER_NAME
5734 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
5735 A C statement (sans semicolon) to output a reference to
5736 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
5737 will be used to output the name of the symbol. This macro may be used
5738 to modify the way a symbol is referenced depending on information
5739 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
5742 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
5743 A C statement (sans semicolon) to output a reference to @var{buf}, the
5744 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
5745 @code{assemble_name} will be used to output the name of the symbol.
5746 This macro is not used by @code{output_asm_label}, or the @code{%l}
5747 specifier that calls it; the intention is that this macro should be set
5748 when it is necessary to output a label differently when its address is
5752 @hook TARGET_ASM_INTERNAL_LABEL
5754 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
5755 A C statement to output to the stdio stream @var{stream} a debug info
5756 label whose name is made from the string @var{prefix} and the number
5757 @var{num}. This is useful for VLIW targets, where debug info labels
5758 may need to be treated differently than branch target labels. On some
5759 systems, branch target labels must be at the beginning of instruction
5760 bundles, but debug info labels can occur in the middle of instruction
5763 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
5767 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5768 A C statement to store into the string @var{string} a label whose name
5769 is made from the string @var{prefix} and the number @var{num}.
5771 This string, when output subsequently by @code{assemble_name}, should
5772 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
5773 with the same @var{prefix} and @var{num}.
5775 If the string begins with @samp{*}, then @code{assemble_name} will
5776 output the rest of the string unchanged. It is often convenient for
5777 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
5778 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5779 to output the string, and may change it. (Of course,
5780 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5781 you should know what it does on your machine.)
5784 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5785 A C expression to assign to @var{outvar} (which is a variable of type
5786 @code{char *}) a newly allocated string made from the string
5787 @var{name} and the number @var{number}, with some suitable punctuation
5788 added. Use @code{alloca} to get space for the string.
5790 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5791 produce an assembler label for an internal static variable whose name is
5792 @var{name}. Therefore, the string must be such as to result in valid
5793 assembler code. The argument @var{number} is different each time this
5794 macro is executed; it prevents conflicts between similarly-named
5795 internal static variables in different scopes.
5797 Ideally this string should not be a valid C identifier, to prevent any
5798 conflict with the user's own symbols. Most assemblers allow periods
5799 or percent signs in assembler symbols; putting at least one of these
5800 between the name and the number will suffice.
5802 If this macro is not defined, a default definition will be provided
5803 which is correct for most systems.
5806 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5807 A C statement to output to the stdio stream @var{stream} assembler code
5808 which defines (equates) the symbol @var{name} to have the value @var{value}.
5811 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5812 correct for most systems.
5815 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
5816 A C statement to output to the stdio stream @var{stream} assembler code
5817 which defines (equates) the symbol whose tree node is @var{decl_of_name}
5818 to have the value of the tree node @var{decl_of_value}. This macro will
5819 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
5820 the tree nodes are available.
5823 If @code{SET_ASM_OP} is defined, a default definition is provided which is
5824 correct for most systems.
5827 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
5828 A C statement that evaluates to true if the assembler code which defines
5829 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
5830 of the tree node @var{decl_of_value} should be emitted near the end of the
5831 current compilation unit. The default is to not defer output of defines.
5832 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
5833 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
5836 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5837 A C statement to output to the stdio stream @var{stream} assembler code
5838 which defines (equates) the weak symbol @var{name} to have the value
5839 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
5840 an undefined weak symbol.
5842 Define this macro if the target only supports weak aliases; define
5843 @code{ASM_OUTPUT_DEF} instead if possible.
5846 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5847 Define this macro to override the default assembler names used for
5848 Objective-C methods.
5850 The default name is a unique method number followed by the name of the
5851 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
5852 the category is also included in the assembler name (e.g.@:
5855 These names are safe on most systems, but make debugging difficult since
5856 the method's selector is not present in the name. Therefore, particular
5857 systems define other ways of computing names.
5859 @var{buf} is an expression of type @code{char *} which gives you a
5860 buffer in which to store the name; its length is as long as
5861 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
5862 50 characters extra.
5864 The argument @var{is_inst} specifies whether the method is an instance
5865 method or a class method; @var{class_name} is the name of the class;
5866 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
5867 in a category); and @var{sel_name} is the name of the selector.
5869 On systems where the assembler can handle quoted names, you can use this
5870 macro to provide more human-readable names.
5873 @node Initialization
5874 @subsection How Initialization Functions Are Handled
5875 @cindex initialization routines
5876 @cindex termination routines
5877 @cindex constructors, output of
5878 @cindex destructors, output of
5880 The compiled code for certain languages includes @dfn{constructors}
5881 (also called @dfn{initialization routines})---functions to initialize
5882 data in the program when the program is started. These functions need
5883 to be called before the program is ``started''---that is to say, before
5884 @code{main} is called.
5886 Compiling some languages generates @dfn{destructors} (also called
5887 @dfn{termination routines}) that should be called when the program
5890 To make the initialization and termination functions work, the compiler
5891 must output something in the assembler code to cause those functions to
5892 be called at the appropriate time. When you port the compiler to a new
5893 system, you need to specify how to do this.
5895 There are two major ways that GCC currently supports the execution of
5896 initialization and termination functions. Each way has two variants.
5897 Much of the structure is common to all four variations.
5899 @findex __CTOR_LIST__
5900 @findex __DTOR_LIST__
5901 The linker must build two lists of these functions---a list of
5902 initialization functions, called @code{__CTOR_LIST__}, and a list of
5903 termination functions, called @code{__DTOR_LIST__}.
5905 Each list always begins with an ignored function pointer (which may hold
5906 0, @minus{}1, or a count of the function pointers after it, depending on
5907 the environment). This is followed by a series of zero or more function
5908 pointers to constructors (or destructors), followed by a function
5909 pointer containing zero.
5911 Depending on the operating system and its executable file format, either
5912 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
5913 time and exit time. Constructors are called in reverse order of the
5914 list; destructors in forward order.
5916 The best way to handle static constructors works only for object file
5917 formats which provide arbitrarily-named sections. A section is set
5918 aside for a list of constructors, and another for a list of destructors.
5919 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
5920 object file that defines an initialization function also puts a word in
5921 the constructor section to point to that function. The linker
5922 accumulates all these words into one contiguous @samp{.ctors} section.
5923 Termination functions are handled similarly.
5925 This method will be chosen as the default by @file{target-def.h} if
5926 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
5927 support arbitrary sections, but does support special designated
5928 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
5929 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
5931 When arbitrary sections are available, there are two variants, depending
5932 upon how the code in @file{crtstuff.c} is called. On systems that
5933 support a @dfn{.init} section which is executed at program startup,
5934 parts of @file{crtstuff.c} are compiled into that section. The
5935 program is linked by the @command{gcc} driver like this:
5938 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
5941 The prologue of a function (@code{__init}) appears in the @code{.init}
5942 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
5943 for the function @code{__fini} in the @dfn{.fini} section. Normally these
5944 files are provided by the operating system or by the GNU C library, but
5945 are provided by GCC for a few targets.
5947 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
5948 compiled from @file{crtstuff.c}. They contain, among other things, code
5949 fragments within the @code{.init} and @code{.fini} sections that branch
5950 to routines in the @code{.text} section. The linker will pull all parts
5951 of a section together, which results in a complete @code{__init} function
5952 that invokes the routines we need at startup.
5954 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
5957 If no init section is available, when GCC compiles any function called
5958 @code{main} (or more accurately, any function designated as a program
5959 entry point by the language front end calling @code{expand_main_function}),
5960 it inserts a procedure call to @code{__main} as the first executable code
5961 after the function prologue. The @code{__main} function is defined
5962 in @file{libgcc2.c} and runs the global constructors.
5964 In file formats that don't support arbitrary sections, there are again
5965 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
5966 and an `a.out' format must be used. In this case,
5967 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
5968 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
5969 and with the address of the void function containing the initialization
5970 code as its value. The GNU linker recognizes this as a request to add
5971 the value to a @dfn{set}; the values are accumulated, and are eventually
5972 placed in the executable as a vector in the format described above, with
5973 a leading (ignored) count and a trailing zero element.
5974 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
5975 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
5976 the compilation of @code{main} to call @code{__main} as above, starting
5977 the initialization process.
5979 The last variant uses neither arbitrary sections nor the GNU linker.
5980 This is preferable when you want to do dynamic linking and when using
5981 file formats which the GNU linker does not support, such as `ECOFF'@. In
5982 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
5983 termination functions are recognized simply by their names. This requires
5984 an extra program in the linkage step, called @command{collect2}. This program
5985 pretends to be the linker, for use with GCC; it does its job by running
5986 the ordinary linker, but also arranges to include the vectors of
5987 initialization and termination functions. These functions are called
5988 via @code{__main} as described above. In order to use this method,
5989 @code{use_collect2} must be defined in the target in @file{config.gcc}.
5992 The following section describes the specific macros that control and
5993 customize the handling of initialization and termination functions.
5996 @node Macros for Initialization
5997 @subsection Macros Controlling Initialization Routines
5999 Here are the macros that control how the compiler handles initialization
6000 and termination functions:
6002 @defmac INIT_SECTION_ASM_OP
6003 If defined, a C string constant, including spacing, for the assembler
6004 operation to identify the following data as initialization code. If not
6005 defined, GCC will assume such a section does not exist. When you are
6006 using special sections for initialization and termination functions, this
6007 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
6008 run the initialization functions.
6011 @defmac HAS_INIT_SECTION
6012 If defined, @code{main} will not call @code{__main} as described above.
6013 This macro should be defined for systems that control start-up code
6014 on a symbol-by-symbol basis, such as OSF/1, and should not
6015 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
6018 @defmac LD_INIT_SWITCH
6019 If defined, a C string constant for a switch that tells the linker that
6020 the following symbol is an initialization routine.
6023 @defmac LD_FINI_SWITCH
6024 If defined, a C string constant for a switch that tells the linker that
6025 the following symbol is a finalization routine.
6028 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
6029 If defined, a C statement that will write a function that can be
6030 automatically called when a shared library is loaded. The function
6031 should call @var{func}, which takes no arguments. If not defined, and
6032 the object format requires an explicit initialization function, then a
6033 function called @code{_GLOBAL__DI} will be generated.
6035 This function and the following one are used by collect2 when linking a
6036 shared library that needs constructors or destructors, or has DWARF2
6037 exception tables embedded in the code.
6040 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
6041 If defined, a C statement that will write a function that can be
6042 automatically called when a shared library is unloaded. The function
6043 should call @var{func}, which takes no arguments. If not defined, and
6044 the object format requires an explicit finalization function, then a
6045 function called @code{_GLOBAL__DD} will be generated.
6048 @defmac INVOKE__main
6049 If defined, @code{main} will call @code{__main} despite the presence of
6050 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
6051 where the init section is not actually run automatically, but is still
6052 useful for collecting the lists of constructors and destructors.
6055 @defmac SUPPORTS_INIT_PRIORITY
6056 If nonzero, the C++ @code{init_priority} attribute is supported and the
6057 compiler should emit instructions to control the order of initialization
6058 of objects. If zero, the compiler will issue an error message upon
6059 encountering an @code{init_priority} attribute.
6062 @hook TARGET_HAVE_CTORS_DTORS
6064 @hook TARGET_ASM_CONSTRUCTOR
6066 @hook TARGET_ASM_DESTRUCTOR
6068 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
6069 generated for the generated object file will have static linkage.
6071 If your system uses @command{collect2} as the means of processing
6072 constructors, then that program normally uses @command{nm} to scan
6073 an object file for constructor functions to be called.
6075 On certain kinds of systems, you can define this macro to make
6076 @command{collect2} work faster (and, in some cases, make it work at all):
6078 @defmac OBJECT_FORMAT_COFF
6079 Define this macro if the system uses COFF (Common Object File Format)
6080 object files, so that @command{collect2} can assume this format and scan
6081 object files directly for dynamic constructor/destructor functions.
6083 This macro is effective only in a native compiler; @command{collect2} as
6084 part of a cross compiler always uses @command{nm} for the target machine.
6087 @defmac REAL_NM_FILE_NAME
6088 Define this macro as a C string constant containing the file name to use
6089 to execute @command{nm}. The default is to search the path normally for
6094 @command{collect2} calls @command{nm} to scan object files for static
6095 constructors and destructors and LTO info. By default, @option{-n} is
6096 passed. Define @code{NM_FLAGS} to a C string constant if other options
6097 are needed to get the same output format as GNU @command{nm -n}
6101 If your system supports shared libraries and has a program to list the
6102 dynamic dependencies of a given library or executable, you can define
6103 these macros to enable support for running initialization and
6104 termination functions in shared libraries:
6107 Define this macro to a C string constant containing the name of the program
6108 which lists dynamic dependencies, like @command{ldd} under SunOS 4.
6111 @defmac PARSE_LDD_OUTPUT (@var{ptr})
6112 Define this macro to be C code that extracts filenames from the output
6113 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
6114 of type @code{char *} that points to the beginning of a line of output
6115 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
6116 code must advance @var{ptr} to the beginning of the filename on that
6117 line. Otherwise, it must set @var{ptr} to @code{NULL}.
6120 @defmac SHLIB_SUFFIX
6121 Define this macro to a C string constant containing the default shared
6122 library extension of the target (e.g., @samp{".so"}). @command{collect2}
6123 strips version information after this suffix when generating global
6124 constructor and destructor names. This define is only needed on targets
6125 that use @command{collect2} to process constructors and destructors.
6128 @node Instruction Output
6129 @subsection Output of Assembler Instructions
6131 @c prevent bad page break with this line
6132 This describes assembler instruction output.
6134 @defmac REGISTER_NAMES
6135 A C initializer containing the assembler's names for the machine
6136 registers, each one as a C string constant. This is what translates
6137 register numbers in the compiler into assembler language.
6140 @defmac ADDITIONAL_REGISTER_NAMES
6141 If defined, a C initializer for an array of structures containing a name
6142 and a register number. This macro defines additional names for hard
6143 registers, thus allowing the @code{asm} option in declarations to refer
6144 to registers using alternate names.
6147 @defmac OVERLAPPING_REGISTER_NAMES
6148 If defined, a C initializer for an array of structures containing a
6149 name, a register number and a count of the number of consecutive
6150 machine registers the name overlaps. This macro defines additional
6151 names for hard registers, thus allowing the @code{asm} option in
6152 declarations to refer to registers using alternate names. Unlike
6153 @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
6154 register name implies multiple underlying registers.
6156 This macro should be used when it is important that a clobber in an
6157 @code{asm} statement clobbers all the underlying values implied by the
6158 register name. For example, on ARM, clobbering the double-precision
6159 VFP register ``d0'' implies clobbering both single-precision registers
6163 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6164 Define this macro if you are using an unusual assembler that
6165 requires different names for the machine instructions.
6167 The definition is a C statement or statements which output an
6168 assembler instruction opcode to the stdio stream @var{stream}. The
6169 macro-operand @var{ptr} is a variable of type @code{char *} which
6170 points to the opcode name in its ``internal'' form---the form that is
6171 written in the machine description. The definition should output the
6172 opcode name to @var{stream}, performing any translation you desire, and
6173 increment the variable @var{ptr} to point at the end of the opcode
6174 so that it will not be output twice.
6176 In fact, your macro definition may process less than the entire opcode
6177 name, or more than the opcode name; but if you want to process text
6178 that includes @samp{%}-sequences to substitute operands, you must take
6179 care of the substitution yourself. Just be sure to increment
6180 @var{ptr} over whatever text should not be output normally.
6182 @findex recog_data.operand
6183 If you need to look at the operand values, they can be found as the
6184 elements of @code{recog_data.operand}.
6186 If the macro definition does nothing, the instruction is output
6190 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6191 If defined, a C statement to be executed just prior to the output of
6192 assembler code for @var{insn}, to modify the extracted operands so
6193 they will be output differently.
6195 Here the argument @var{opvec} is the vector containing the operands
6196 extracted from @var{insn}, and @var{noperands} is the number of
6197 elements of the vector which contain meaningful data for this insn.
6198 The contents of this vector are what will be used to convert the insn
6199 template into assembler code, so you can change the assembler output
6200 by changing the contents of the vector.
6202 This macro is useful when various assembler syntaxes share a single
6203 file of instruction patterns; by defining this macro differently, you
6204 can cause a large class of instructions to be output differently (such
6205 as with rearranged operands). Naturally, variations in assembler
6206 syntax affecting individual insn patterns ought to be handled by
6207 writing conditional output routines in those patterns.
6209 If this macro is not defined, it is equivalent to a null statement.
6212 @hook TARGET_ASM_FINAL_POSTSCAN_INSN
6214 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6215 A C compound statement to output to stdio stream @var{stream} the
6216 assembler syntax for an instruction operand @var{x}. @var{x} is an
6219 @var{code} is a value that can be used to specify one of several ways
6220 of printing the operand. It is used when identical operands must be
6221 printed differently depending on the context. @var{code} comes from
6222 the @samp{%} specification that was used to request printing of the
6223 operand. If the specification was just @samp{%@var{digit}} then
6224 @var{code} is 0; if the specification was @samp{%@var{ltr}
6225 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6228 If @var{x} is a register, this macro should print the register's name.
6229 The names can be found in an array @code{reg_names} whose type is
6230 @code{char *[]}. @code{reg_names} is initialized from
6231 @code{REGISTER_NAMES}.
6233 When the machine description has a specification @samp{%@var{punct}}
6234 (a @samp{%} followed by a punctuation character), this macro is called
6235 with a null pointer for @var{x} and the punctuation character for
6239 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6240 A C expression which evaluates to true if @var{code} is a valid
6241 punctuation character for use in the @code{PRINT_OPERAND} macro. If
6242 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6243 punctuation characters (except for the standard one, @samp{%}) are used
6247 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6248 A C compound statement to output to stdio stream @var{stream} the
6249 assembler syntax for an instruction operand that is a memory reference
6250 whose address is @var{x}. @var{x} is an RTL expression.
6252 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
6253 On some machines, the syntax for a symbolic address depends on the
6254 section that the address refers to. On these machines, define the hook
6255 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
6256 @code{symbol_ref}, and then check for it here. @xref{Assembler
6260 @findex dbr_sequence_length
6261 @defmac DBR_OUTPUT_SEQEND (@var{file})
6262 A C statement, to be executed after all slot-filler instructions have
6263 been output. If necessary, call @code{dbr_sequence_length} to
6264 determine the number of slots filled in a sequence (zero if not
6265 currently outputting a sequence), to decide how many no-ops to output,
6268 Don't define this macro if it has nothing to do, but it is helpful in
6269 reading assembly output if the extent of the delay sequence is made
6270 explicit (e.g.@: with white space).
6273 @findex final_sequence
6274 Note that output routines for instructions with delay slots must be
6275 prepared to deal with not being output as part of a sequence
6276 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
6277 found.) The variable @code{final_sequence} is null when not
6278 processing a sequence, otherwise it contains the @code{sequence} rtx
6282 @defmac REGISTER_PREFIX
6283 @defmacx LOCAL_LABEL_PREFIX
6284 @defmacx USER_LABEL_PREFIX
6285 @defmacx IMMEDIATE_PREFIX
6286 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6287 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6288 @file{final.c}). These are useful when a single @file{md} file must
6289 support multiple assembler formats. In that case, the various @file{tm.h}
6290 files can define these macros differently.
6293 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
6294 If defined this macro should expand to a series of @code{case}
6295 statements which will be parsed inside the @code{switch} statement of
6296 the @code{asm_fprintf} function. This allows targets to define extra
6297 printf formats which may useful when generating their assembler
6298 statements. Note that uppercase letters are reserved for future
6299 generic extensions to asm_fprintf, and so are not available to target
6300 specific code. The output file is given by the parameter @var{file}.
6301 The varargs input pointer is @var{argptr} and the rest of the format
6302 string, starting the character after the one that is being switched
6303 upon, is pointed to by @var{format}.
6306 @defmac ASSEMBLER_DIALECT
6307 If your target supports multiple dialects of assembler language (such as
6308 different opcodes), define this macro as a C expression that gives the
6309 numeric index of the assembler language dialect to use, with zero as the
6312 If this macro is defined, you may use constructs of the form
6314 @samp{@{option0|option1|option2@dots{}@}}
6317 in the output templates of patterns (@pxref{Output Template}) or in the
6318 first argument of @code{asm_fprintf}. This construct outputs
6319 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
6320 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
6321 within these strings retain their usual meaning. If there are fewer
6322 alternatives within the braces than the value of
6323 @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
6324 to print curly braces or @samp{|} character in assembler output directly,
6325 @samp{%@{}, @samp{%@}} and @samp{%|} can be used.
6327 If you do not define this macro, the characters @samp{@{}, @samp{|} and
6328 @samp{@}} do not have any special meaning when used in templates or
6329 operands to @code{asm_fprintf}.
6331 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6332 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6333 the variations in assembler language syntax with that mechanism. Define
6334 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6335 if the syntax variant are larger and involve such things as different
6336 opcodes or operand order.
6339 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6340 A C expression to output to @var{stream} some assembler code
6341 which will push hard register number @var{regno} onto the stack.
6342 The code need not be optimal, since this macro is used only when
6346 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6347 A C expression to output to @var{stream} some assembler code
6348 which will pop hard register number @var{regno} off of the stack.
6349 The code need not be optimal, since this macro is used only when
6353 @node Dispatch Tables
6354 @subsection Output of Dispatch Tables
6356 @c prevent bad page break with this line
6357 This concerns dispatch tables.
6359 @cindex dispatch table
6360 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6361 A C statement to output to the stdio stream @var{stream} an assembler
6362 pseudo-instruction to generate a difference between two labels.
6363 @var{value} and @var{rel} are the numbers of two internal labels. The
6364 definitions of these labels are output using
6365 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
6366 way here. For example,
6369 fprintf (@var{stream}, "\t.word L%d-L%d\n",
6370 @var{value}, @var{rel})
6373 You must provide this macro on machines where the addresses in a
6374 dispatch table are relative to the table's own address. If defined, GCC
6375 will also use this macro on all machines when producing PIC@.
6376 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
6377 mode and flags can be read.
6380 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6381 This macro should be provided on machines where the addresses
6382 in a dispatch table are absolute.
6384 The definition should be a C statement to output to the stdio stream
6385 @var{stream} an assembler pseudo-instruction to generate a reference to
6386 a label. @var{value} is the number of an internal label whose
6387 definition is output using @code{(*targetm.asm_out.internal_label)}.
6391 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6395 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6396 Define this if the label before a jump-table needs to be output
6397 specially. The first three arguments are the same as for
6398 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
6399 jump-table which follows (a @code{jump_table_data} containing an
6400 @code{addr_vec} or @code{addr_diff_vec}).
6402 This feature is used on system V to output a @code{swbeg} statement
6405 If this macro is not defined, these labels are output with
6406 @code{(*targetm.asm_out.internal_label)}.
6409 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6410 Define this if something special must be output at the end of a
6411 jump-table. The definition should be a C statement to be executed
6412 after the assembler code for the table is written. It should write
6413 the appropriate code to stdio stream @var{stream}. The argument
6414 @var{table} is the jump-table insn, and @var{num} is the label-number
6415 of the preceding label.
6417 If this macro is not defined, nothing special is output at the end of
6421 @hook TARGET_ASM_EMIT_UNWIND_LABEL
6423 @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
6425 @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
6427 @hook TARGET_ASM_UNWIND_EMIT
6429 @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
6431 @node Exception Region Output
6432 @subsection Assembler Commands for Exception Regions
6434 @c prevent bad page break with this line
6436 This describes commands marking the start and the end of an exception
6439 @defmac EH_FRAME_SECTION_NAME
6440 If defined, a C string constant for the name of the section containing
6441 exception handling frame unwind information. If not defined, GCC will
6442 provide a default definition if the target supports named sections.
6443 @file{crtstuff.c} uses this macro to switch to the appropriate section.
6445 You should define this symbol if your target supports DWARF 2 frame
6446 unwind information and the default definition does not work.
6449 @defmac EH_FRAME_IN_DATA_SECTION
6450 If defined, DWARF 2 frame unwind information will be placed in the
6451 data section even though the target supports named sections. This
6452 might be necessary, for instance, if the system linker does garbage
6453 collection and sections cannot be marked as not to be collected.
6455 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
6459 @defmac EH_TABLES_CAN_BE_READ_ONLY
6460 Define this macro to 1 if your target is such that no frame unwind
6461 information encoding used with non-PIC code will ever require a
6462 runtime relocation, but the linker may not support merging read-only
6463 and read-write sections into a single read-write section.
6466 @defmac MASK_RETURN_ADDR
6467 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
6468 that it does not contain any extraneous set bits in it.
6471 @defmac DWARF2_UNWIND_INFO
6472 Define this macro to 0 if your target supports DWARF 2 frame unwind
6473 information, but it does not yet work with exception handling.
6474 Otherwise, if your target supports this information (if it defines
6475 @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
6476 GCC will provide a default definition of 1.
6479 @hook TARGET_EXCEPT_UNWIND_INFO
6480 This hook defines the mechanism that will be used for exception handling
6481 by the target. If the target has ABI specified unwind tables, the hook
6482 should return @code{UI_TARGET}. If the target is to use the
6483 @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
6484 should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind
6485 information, the hook should return @code{UI_DWARF2}.
6487 A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
6488 This may end up simplifying other parts of target-specific code. The
6489 default implementation of this hook never returns @code{UI_NONE}.
6491 Note that the value returned by this hook should be constant. It should
6492 not depend on anything except the command-line switches described by
6493 @var{opts}. In particular, the
6494 setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
6495 macros and builtin functions related to exception handling are set up
6496 depending on this setting.
6498 The default implementation of the hook first honors the
6499 @option{--enable-sjlj-exceptions} configure option, then
6500 @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If
6501 @code{DWARF2_UNWIND_INFO} depends on command-line options, the target
6502 must define this hook so that @var{opts} is used correctly.
6505 @hook TARGET_UNWIND_TABLES_DEFAULT
6506 This variable should be set to @code{true} if the target ABI requires unwinding
6507 tables even when exceptions are not used. It must not be modified by
6508 command-line option processing.
6511 @defmac DONT_USE_BUILTIN_SETJMP
6512 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
6513 should use the @code{setjmp}/@code{longjmp} functions from the C library
6514 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
6517 @defmac JMP_BUF_SIZE
6518 This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
6519 defined. Define this macro if the default size of @code{jmp_buf} buffer
6520 for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
6521 is not large enough, or if it is much too large.
6522 The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
6525 @defmac DWARF_CIE_DATA_ALIGNMENT
6526 This macro need only be defined if the target might save registers in the
6527 function prologue at an offset to the stack pointer that is not aligned to
6528 @code{UNITS_PER_WORD}. The definition should be the negative minimum
6529 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
6530 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
6531 the target supports DWARF 2 frame unwind information.
6534 @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
6536 @hook TARGET_DWARF_REGISTER_SPAN
6538 @hook TARGET_DWARF_FRAME_REG_MODE
6540 @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
6542 @hook TARGET_ASM_TTYPE
6544 @hook TARGET_ARM_EABI_UNWINDER
6546 @node Alignment Output
6547 @subsection Assembler Commands for Alignment
6549 @c prevent bad page break with this line
6550 This describes commands for alignment.
6552 @defmac JUMP_ALIGN (@var{label})
6553 The alignment (log base 2) to put in front of @var{label}, which is
6554 a common destination of jumps and has no fallthru incoming edge.
6556 This macro need not be defined if you don't want any special alignment
6557 to be done at such a time. Most machine descriptions do not currently
6560 Unless it's necessary to inspect the @var{label} parameter, it is better
6561 to set the variable @var{align_jumps} in the target's
6562 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6563 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
6566 @hook TARGET_ASM_JUMP_ALIGN_MAX_SKIP
6568 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
6569 The alignment (log base 2) to put in front of @var{label}, which follows
6572 This macro need not be defined if you don't want any special alignment
6573 to be done at such a time. Most machine descriptions do not currently
6577 @hook TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
6579 @defmac LOOP_ALIGN (@var{label})
6580 The alignment (log base 2) to put in front of @var{label} that heads
6581 a frequently executed basic block (usually the header of a loop).
6583 This macro need not be defined if you don't want any special alignment
6584 to be done at such a time. Most machine descriptions do not currently
6587 Unless it's necessary to inspect the @var{label} parameter, it is better
6588 to set the variable @code{align_loops} in the target's
6589 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6590 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
6593 @hook TARGET_ASM_LOOP_ALIGN_MAX_SKIP
6595 @defmac LABEL_ALIGN (@var{label})
6596 The alignment (log base 2) to put in front of @var{label}.
6597 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
6598 the maximum of the specified values is used.
6600 Unless it's necessary to inspect the @var{label} parameter, it is better
6601 to set the variable @code{align_labels} in the target's
6602 @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's
6603 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
6606 @hook TARGET_ASM_LABEL_ALIGN_MAX_SKIP
6608 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6609 A C statement to output to the stdio stream @var{stream} an assembler
6610 instruction to advance the location counter by @var{nbytes} bytes.
6611 Those bytes should be zero when loaded. @var{nbytes} will be a C
6612 expression of type @code{unsigned HOST_WIDE_INT}.
6615 @defmac ASM_NO_SKIP_IN_TEXT
6616 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6617 text section because it fails to put zeros in the bytes that are skipped.
6618 This is true on many Unix systems, where the pseudo--op to skip bytes
6619 produces no-op instructions rather than zeros when used in the text
6623 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6624 A C statement to output to the stdio stream @var{stream} an assembler
6625 command to advance the location counter to a multiple of 2 to the
6626 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
6629 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
6630 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
6631 for padding, if necessary.
6634 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6635 A C statement to output to the stdio stream @var{stream} an assembler
6636 command to advance the location counter to a multiple of 2 to the
6637 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6638 satisfy the alignment request. @var{power} and @var{max_skip} will be
6639 a C expression of type @code{int}.
6643 @node Debugging Info
6644 @section Controlling Debugging Information Format
6646 @c prevent bad page break with this line
6647 This describes how to specify debugging information.
6650 * All Debuggers:: Macros that affect all debugging formats uniformly.
6651 * DBX Options:: Macros enabling specific options in DBX format.
6652 * DBX Hooks:: Hook macros for varying DBX format.
6653 * File Names and DBX:: Macros controlling output of file names in DBX format.
6654 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
6655 * VMS Debug:: Macros for VMS debug format.
6659 @subsection Macros Affecting All Debugging Formats
6661 @c prevent bad page break with this line
6662 These macros affect all debugging formats.
6664 @defmac DBX_REGISTER_NUMBER (@var{regno})
6665 A C expression that returns the DBX register number for the compiler
6666 register number @var{regno}. In the default macro provided, the value
6667 of this expression will be @var{regno} itself. But sometimes there are
6668 some registers that the compiler knows about and DBX does not, or vice
6669 versa. In such cases, some register may need to have one number in the
6670 compiler and another for DBX@.
6672 If two registers have consecutive numbers inside GCC, and they can be
6673 used as a pair to hold a multiword value, then they @emph{must} have
6674 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6675 Otherwise, debuggers will be unable to access such a pair, because they
6676 expect register pairs to be consecutive in their own numbering scheme.
6678 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6679 does not preserve register pairs, then what you must do instead is
6680 redefine the actual register numbering scheme.
6683 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
6684 A C expression that returns the integer offset value for an automatic
6685 variable having address @var{x} (an RTL expression). The default
6686 computation assumes that @var{x} is based on the frame-pointer and
6687 gives the offset from the frame-pointer. This is required for targets
6688 that produce debugging output for DBX or COFF-style debugging output
6689 for SDB and allow the frame-pointer to be eliminated when the
6690 @option{-g} options is used.
6693 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6694 A C expression that returns the integer offset value for an argument
6695 having address @var{x} (an RTL expression). The nominal offset is
6699 @defmac PREFERRED_DEBUGGING_TYPE
6700 A C expression that returns the type of debugging output GCC should
6701 produce when the user specifies just @option{-g}. Define
6702 this if you have arranged for GCC to support more than one format of
6703 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
6704 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
6705 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
6707 When the user specifies @option{-ggdb}, GCC normally also uses the
6708 value of this macro to select the debugging output format, but with two
6709 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
6710 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
6711 defined, GCC uses @code{DBX_DEBUG}.
6713 The value of this macro only affects the default debugging output; the
6714 user can always get a specific type of output by using @option{-gstabs},
6715 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
6719 @subsection Specific Options for DBX Output
6721 @c prevent bad page break with this line
6722 These are specific options for DBX output.
6724 @defmac DBX_DEBUGGING_INFO
6725 Define this macro if GCC should produce debugging output for DBX
6726 in response to the @option{-g} option.
6729 @defmac XCOFF_DEBUGGING_INFO
6730 Define this macro if GCC should produce XCOFF format debugging output
6731 in response to the @option{-g} option. This is a variant of DBX format.
6734 @defmac DEFAULT_GDB_EXTENSIONS
6735 Define this macro to control whether GCC should by default generate
6736 GDB's extended version of DBX debugging information (assuming DBX-format
6737 debugging information is enabled at all). If you don't define the
6738 macro, the default is 1: always generate the extended information
6739 if there is any occasion to.
6742 @defmac DEBUG_SYMS_TEXT
6743 Define this macro if all @code{.stabs} commands should be output while
6744 in the text section.
6747 @defmac ASM_STABS_OP
6748 A C string constant, including spacing, naming the assembler pseudo op to
6749 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
6750 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
6751 applies only to DBX debugging information format.
6754 @defmac ASM_STABD_OP
6755 A C string constant, including spacing, naming the assembler pseudo op to
6756 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
6757 value is the current location. If you don't define this macro,
6758 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
6762 @defmac ASM_STABN_OP
6763 A C string constant, including spacing, naming the assembler pseudo op to
6764 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
6765 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
6766 macro applies only to DBX debugging information format.
6769 @defmac DBX_NO_XREFS
6770 Define this macro if DBX on your system does not support the construct
6771 @samp{xs@var{tagname}}. On some systems, this construct is used to
6772 describe a forward reference to a structure named @var{tagname}.
6773 On other systems, this construct is not supported at all.
6776 @defmac DBX_CONTIN_LENGTH
6777 A symbol name in DBX-format debugging information is normally
6778 continued (split into two separate @code{.stabs} directives) when it
6779 exceeds a certain length (by default, 80 characters). On some
6780 operating systems, DBX requires this splitting; on others, splitting
6781 must not be done. You can inhibit splitting by defining this macro
6782 with the value zero. You can override the default splitting-length by
6783 defining this macro as an expression for the length you desire.
6786 @defmac DBX_CONTIN_CHAR
6787 Normally continuation is indicated by adding a @samp{\} character to
6788 the end of a @code{.stabs} string when a continuation follows. To use
6789 a different character instead, define this macro as a character
6790 constant for the character you want to use. Do not define this macro
6791 if backslash is correct for your system.
6794 @defmac DBX_STATIC_STAB_DATA_SECTION
6795 Define this macro if it is necessary to go to the data section before
6796 outputting the @samp{.stabs} pseudo-op for a non-global static
6800 @defmac DBX_TYPE_DECL_STABS_CODE
6801 The value to use in the ``code'' field of the @code{.stabs} directive
6802 for a typedef. The default is @code{N_LSYM}.
6805 @defmac DBX_STATIC_CONST_VAR_CODE
6806 The value to use in the ``code'' field of the @code{.stabs} directive
6807 for a static variable located in the text section. DBX format does not
6808 provide any ``right'' way to do this. The default is @code{N_FUN}.
6811 @defmac DBX_REGPARM_STABS_CODE
6812 The value to use in the ``code'' field of the @code{.stabs} directive
6813 for a parameter passed in registers. DBX format does not provide any
6814 ``right'' way to do this. The default is @code{N_RSYM}.
6817 @defmac DBX_REGPARM_STABS_LETTER
6818 The letter to use in DBX symbol data to identify a symbol as a parameter
6819 passed in registers. DBX format does not customarily provide any way to
6820 do this. The default is @code{'P'}.
6823 @defmac DBX_FUNCTION_FIRST
6824 Define this macro if the DBX information for a function and its
6825 arguments should precede the assembler code for the function. Normally,
6826 in DBX format, the debugging information entirely follows the assembler
6830 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
6831 Define this macro, with value 1, if the value of a symbol describing
6832 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
6833 relative to the start of the enclosing function. Normally, GCC uses
6834 an absolute address.
6837 @defmac DBX_LINES_FUNCTION_RELATIVE
6838 Define this macro, with value 1, if the value of a symbol indicating
6839 the current line number (@code{N_SLINE}) should be relative to the
6840 start of the enclosing function. Normally, GCC uses an absolute address.
6843 @defmac DBX_USE_BINCL
6844 Define this macro if GCC should generate @code{N_BINCL} and
6845 @code{N_EINCL} stabs for included header files, as on Sun systems. This
6846 macro also directs GCC to output a type number as a pair of a file
6847 number and a type number within the file. Normally, GCC does not
6848 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6849 number for a type number.
6853 @subsection Open-Ended Hooks for DBX Format
6855 @c prevent bad page break with this line
6856 These are hooks for DBX format.
6858 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6859 A C statement to output DBX debugging information before code for line
6860 number @var{line} of the current source file to the stdio stream
6861 @var{stream}. @var{counter} is the number of time the macro was
6862 invoked, including the current invocation; it is intended to generate
6863 unique labels in the assembly output.
6865 This macro should not be defined if the default output is correct, or
6866 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
6869 @defmac NO_DBX_FUNCTION_END
6870 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
6871 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
6872 On those machines, define this macro to turn this feature off without
6873 disturbing the rest of the gdb extensions.
6876 @defmac NO_DBX_BNSYM_ENSYM
6877 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
6878 extension construct. On those machines, define this macro to turn this
6879 feature off without disturbing the rest of the gdb extensions.
6882 @node File Names and DBX
6883 @subsection File Names in DBX Format
6885 @c prevent bad page break with this line
6886 This describes file names in DBX format.
6888 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
6889 A C statement to output DBX debugging information to the stdio stream
6890 @var{stream}, which indicates that file @var{name} is the main source
6891 file---the file specified as the input file for compilation.
6892 This macro is called only once, at the beginning of compilation.
6894 This macro need not be defined if the standard form of output
6895 for DBX debugging information is appropriate.
6897 It may be necessary to refer to a label equal to the beginning of the
6898 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
6899 to do so. If you do this, you must also set the variable
6900 @var{used_ltext_label_name} to @code{true}.
6903 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
6904 Define this macro, with value 1, if GCC should not emit an indication
6905 of the current directory for compilation and current source language at
6906 the beginning of the file.
6909 @defmac NO_DBX_GCC_MARKER
6910 Define this macro, with value 1, if GCC should not emit an indication
6911 that this object file was compiled by GCC@. The default is to emit
6912 an @code{N_OPT} stab at the beginning of every source file, with
6913 @samp{gcc2_compiled.} for the string and value 0.
6916 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
6917 A C statement to output DBX debugging information at the end of
6918 compilation of the main source file @var{name}. Output should be
6919 written to the stdio stream @var{stream}.
6921 If you don't define this macro, nothing special is output at the end
6922 of compilation, which is correct for most machines.
6925 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
6926 Define this macro @emph{instead of} defining
6927 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
6928 the end of compilation is an @code{N_SO} stab with an empty string,
6929 whose value is the highest absolute text address in the file.
6934 @subsection Macros for SDB and DWARF Output
6936 @c prevent bad page break with this line
6937 Here are macros for SDB and DWARF output.
6939 @defmac SDB_DEBUGGING_INFO
6940 Define this macro if GCC should produce COFF-style debugging output
6941 for SDB in response to the @option{-g} option.
6944 @defmac DWARF2_DEBUGGING_INFO
6945 Define this macro if GCC should produce dwarf version 2 format
6946 debugging output in response to the @option{-g} option.
6948 @hook TARGET_DWARF_CALLING_CONVENTION
6950 To support optional call frame debugging information, you must also
6951 define @code{INCOMING_RETURN_ADDR_RTX} and either set
6952 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
6953 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
6954 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
6957 @defmac DWARF2_FRAME_INFO
6958 Define this macro to a nonzero value if GCC should always output
6959 Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO}
6960 (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
6961 exceptions are enabled, GCC will output this information not matter
6962 how you define @code{DWARF2_FRAME_INFO}.
6965 @hook TARGET_DEBUG_UNWIND_INFO
6967 @defmac DWARF2_ASM_LINE_DEBUG_INFO
6968 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
6969 line debug info sections. This will result in much more compact line number
6970 tables, and hence is desirable if it works.
6973 @hook TARGET_WANT_DEBUG_PUB_SECTIONS
6975 @hook TARGET_FORCE_AT_COMP_DIR
6977 @hook TARGET_DELAY_SCHED2
6979 @hook TARGET_DELAY_VARTRACK
6981 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
6982 A C statement to issue assembly directives that create a difference
6983 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
6986 @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
6987 A C statement to issue assembly directives that create a difference
6988 between the two given labels in system defined units, e.g. instruction
6989 slots on IA64 VMS, using an integer of the given size.
6992 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
6993 A C statement to issue assembly directives that create a
6994 section-relative reference to the given @var{label}, using an integer of the
6995 given @var{size}. The label is known to be defined in the given @var{section}.
6998 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
6999 A C statement to issue assembly directives that create a self-relative
7000 reference to the given @var{label}, using an integer of the given @var{size}.
7003 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
7004 A C statement to issue assembly directives that create a reference to
7005 the DWARF table identifier @var{label} from the current section. This
7006 is used on some systems to avoid garbage collecting a DWARF table which
7007 is referenced by a function.
7010 @hook TARGET_ASM_OUTPUT_DWARF_DTPREL
7012 @defmac PUT_SDB_@dots{}
7013 Define these macros to override the assembler syntax for the special
7014 SDB assembler directives. See @file{sdbout.c} for a list of these
7015 macros and their arguments. If the standard syntax is used, you need
7016 not define them yourself.
7020 Some assemblers do not support a semicolon as a delimiter, even between
7021 SDB assembler directives. In that case, define this macro to be the
7022 delimiter to use (usually @samp{\n}). It is not necessary to define
7023 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
7027 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
7028 Define this macro to allow references to unknown structure,
7029 union, or enumeration tags to be emitted. Standard COFF does not
7030 allow handling of unknown references, MIPS ECOFF has support for
7034 @defmac SDB_ALLOW_FORWARD_REFERENCES
7035 Define this macro to allow references to structure, union, or
7036 enumeration tags that have not yet been seen to be handled. Some
7037 assemblers choke if forward tags are used, while some require it.
7040 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
7041 A C statement to output SDB debugging information before code for line
7042 number @var{line} of the current source file to the stdio stream
7043 @var{stream}. The default is to emit an @code{.ln} directive.
7048 @subsection Macros for VMS Debug Format
7050 @c prevent bad page break with this line
7051 Here are macros for VMS debug format.
7053 @defmac VMS_DEBUGGING_INFO
7054 Define this macro if GCC should produce debugging output for VMS
7055 in response to the @option{-g} option. The default behavior for VMS
7056 is to generate minimal debug info for a traceback in the absence of
7057 @option{-g} unless explicitly overridden with @option{-g0}. This
7058 behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
7059 @code{TARGET_OPTION_OVERRIDE}.
7062 @node Floating Point
7063 @section Cross Compilation and Floating Point
7064 @cindex cross compilation and floating point
7065 @cindex floating point and cross compilation
7067 While all modern machines use twos-complement representation for integers,
7068 there are a variety of representations for floating point numbers. This
7069 means that in a cross-compiler the representation of floating point numbers
7070 in the compiled program may be different from that used in the machine
7071 doing the compilation.
7073 Because different representation systems may offer different amounts of
7074 range and precision, all floating point constants must be represented in
7075 the target machine's format. Therefore, the cross compiler cannot
7076 safely use the host machine's floating point arithmetic; it must emulate
7077 the target's arithmetic. To ensure consistency, GCC always uses
7078 emulation to work with floating point values, even when the host and
7079 target floating point formats are identical.
7081 The following macros are provided by @file{real.h} for the compiler to
7082 use. All parts of the compiler which generate or optimize
7083 floating-point calculations must use these macros. They may evaluate
7084 their operands more than once, so operands must not have side effects.
7086 @defmac REAL_VALUE_TYPE
7087 The C data type to be used to hold a floating point value in the target
7088 machine's format. Typically this is a @code{struct} containing an
7089 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
7093 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7094 Compares for equality the two values, @var{x} and @var{y}. If the target
7095 floating point format supports negative zeroes and/or NaNs,
7096 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
7097 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
7100 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7101 Tests whether @var{x} is less than @var{y}.
7104 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
7105 Truncates @var{x} to a signed integer, rounding toward zero.
7108 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
7109 Truncates @var{x} to an unsigned integer, rounding toward zero. If
7110 @var{x} is negative, returns zero.
7113 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
7114 Converts @var{string} into a floating point number in the target machine's
7115 representation for mode @var{mode}. This routine can handle both
7116 decimal and hexadecimal floating point constants, using the syntax
7117 defined by the C language for both.
7120 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
7121 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
7124 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
7125 Determines whether @var{x} represents infinity (positive or negative).
7128 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
7129 Determines whether @var{x} represents a ``NaN'' (not-a-number).
7132 @deftypefn Macro void REAL_ARITHMETIC (REAL_VALUE_TYPE @var{output}, enum tree_code @var{code}, REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
7133 Calculates an arithmetic operation on the two floating point values
7134 @var{x} and @var{y}, storing the result in @var{output} (which must be a
7137 The operation to be performed is specified by @var{code}. Only the
7138 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
7139 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
7141 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
7142 target's floating point format cannot represent infinity, it will call
7143 @code{abort}. Callers should check for this situation first, using
7144 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
7147 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
7148 Returns the negative of the floating point value @var{x}.
7151 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
7152 Returns the absolute value of @var{x}.
7155 @node Mode Switching
7156 @section Mode Switching Instructions
7157 @cindex mode switching
7158 The following macros control mode switching optimizations:
7160 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
7161 Define this macro if the port needs extra instructions inserted for mode
7162 switching in an optimizing compilation.
7164 For an example, the SH4 can perform both single and double precision
7165 floating point operations, but to perform a single precision operation,
7166 the FPSCR PR bit has to be cleared, while for a double precision
7167 operation, this bit has to be set. Changing the PR bit requires a general
7168 purpose register as a scratch register, hence these FPSCR sets have to
7169 be inserted before reload, i.e.@: you can't put this into instruction emitting
7170 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
7172 You can have multiple entities that are mode-switched, and select at run time
7173 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
7174 return nonzero for any @var{entity} that needs mode-switching.
7175 If you define this macro, you also have to define
7176 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
7177 @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
7178 @code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
7182 @defmac NUM_MODES_FOR_MODE_SWITCHING
7183 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7184 initializer for an array of integers. Each initializer element
7185 N refers to an entity that needs mode switching, and specifies the number
7186 of different modes that might need to be set for this entity.
7187 The position of the initializer in the initializer---starting counting at
7188 zero---determines the integer that is used to refer to the mode-switched
7190 In macros that take mode arguments / yield a mode result, modes are
7191 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
7192 switch is needed / supplied.
7195 @hook TARGET_MODE_EMIT
7197 @hook TARGET_MODE_NEEDED
7199 @hook TARGET_MODE_AFTER
7201 @hook TARGET_MODE_ENTRY
7203 @hook TARGET_MODE_EXIT
7205 @hook TARGET_MODE_PRIORITY
7207 @node Target Attributes
7208 @section Defining target-specific uses of @code{__attribute__}
7209 @cindex target attributes
7210 @cindex machine attributes
7211 @cindex attributes, target-specific
7213 Target-specific attributes may be defined for functions, data and types.
7214 These are described using the following target hooks; they also need to
7215 be documented in @file{extend.texi}.
7217 @hook TARGET_ATTRIBUTE_TABLE
7219 @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
7221 @hook TARGET_COMP_TYPE_ATTRIBUTES
7223 @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
7225 @hook TARGET_MERGE_TYPE_ATTRIBUTES
7227 @hook TARGET_MERGE_DECL_ATTRIBUTES
7229 @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
7231 @defmac TARGET_DECLSPEC
7232 Define this macro to a nonzero value if you want to treat
7233 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
7234 default, this behavior is enabled only for targets that define
7235 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
7236 of @code{__declspec} is via a built-in macro, but you should not rely
7237 on this implementation detail.
7240 @hook TARGET_INSERT_ATTRIBUTES
7242 @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
7244 @hook TARGET_OPTION_VALID_ATTRIBUTE_P
7246 @hook TARGET_OPTION_SAVE
7248 @hook TARGET_OPTION_RESTORE
7250 @hook TARGET_OPTION_PRINT
7252 @hook TARGET_OPTION_PRAGMA_PARSE
7254 @hook TARGET_OPTION_OVERRIDE
7256 @hook TARGET_OPTION_FUNCTION_VERSIONS
7258 @hook TARGET_CAN_INLINE_P
7261 @section Emulating TLS
7262 @cindex Emulated TLS
7264 For targets whose psABI does not provide Thread Local Storage via
7265 specific relocations and instruction sequences, an emulation layer is
7266 used. A set of target hooks allows this emulation layer to be
7267 configured for the requirements of a particular target. For instance
7268 the psABI may in fact specify TLS support in terms of an emulation
7271 The emulation layer works by creating a control object for every TLS
7272 object. To access the TLS object, a lookup function is provided
7273 which, when given the address of the control object, will return the
7274 address of the current thread's instance of the TLS object.
7276 @hook TARGET_EMUTLS_GET_ADDRESS
7278 @hook TARGET_EMUTLS_REGISTER_COMMON
7280 @hook TARGET_EMUTLS_VAR_SECTION
7282 @hook TARGET_EMUTLS_TMPL_SECTION
7284 @hook TARGET_EMUTLS_VAR_PREFIX
7286 @hook TARGET_EMUTLS_TMPL_PREFIX
7288 @hook TARGET_EMUTLS_VAR_FIELDS
7290 @hook TARGET_EMUTLS_VAR_INIT
7292 @hook TARGET_EMUTLS_VAR_ALIGN_FIXED
7294 @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
7296 @node MIPS Coprocessors
7297 @section Defining coprocessor specifics for MIPS targets.
7298 @cindex MIPS coprocessor-definition macros
7300 The MIPS specification allows MIPS implementations to have as many as 4
7301 coprocessors, each with as many as 32 private registers. GCC supports
7302 accessing these registers and transferring values between the registers
7303 and memory using asm-ized variables. For example:
7306 register unsigned int cp0count asm ("c0r1");
7312 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
7313 names may be added as described below, or the default names may be
7314 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
7316 Coprocessor registers are assumed to be epilogue-used; sets to them will
7317 be preserved even if it does not appear that the register is used again
7318 later in the function.
7320 Another note: according to the MIPS spec, coprocessor 1 (if present) is
7321 the FPU@. One accesses COP1 registers through standard mips
7322 floating-point support; they are not included in this mechanism.
7325 @section Parameters for Precompiled Header Validity Checking
7326 @cindex parameters, precompiled headers
7328 @hook TARGET_GET_PCH_VALIDITY
7330 @hook TARGET_PCH_VALID_P
7332 @hook TARGET_CHECK_PCH_TARGET_FLAGS
7334 @hook TARGET_PREPARE_PCH_SAVE
7337 @section C++ ABI parameters
7338 @cindex parameters, c++ abi
7340 @hook TARGET_CXX_GUARD_TYPE
7342 @hook TARGET_CXX_GUARD_MASK_BIT
7344 @hook TARGET_CXX_GET_COOKIE_SIZE
7346 @hook TARGET_CXX_COOKIE_HAS_SIZE
7348 @hook TARGET_CXX_IMPORT_EXPORT_CLASS
7350 @hook TARGET_CXX_CDTOR_RETURNS_THIS
7352 @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
7354 @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
7356 @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
7358 @hook TARGET_CXX_LIBRARY_RTTI_COMDAT
7360 @hook TARGET_CXX_USE_AEABI_ATEXIT
7362 @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
7364 @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
7366 @hook TARGET_CXX_DECL_MANGLING_CONTEXT
7368 @node Named Address Spaces
7369 @section Adding support for named address spaces
7370 @cindex named address spaces
7372 The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
7373 standards committee, @cite{Programming Languages - C - Extensions to
7374 support embedded processors}, specifies a syntax for embedded
7375 processors to specify alternate address spaces. You can configure a
7376 GCC port to support section 5.1 of the draft report to add support for
7377 address spaces other than the default address space. These address
7378 spaces are new keywords that are similar to the @code{volatile} and
7379 @code{const} type attributes.
7381 Pointers to named address spaces can have a different size than
7382 pointers to the generic address space.
7384 For example, the SPU port uses the @code{__ea} address space to refer
7385 to memory in the host processor, rather than memory local to the SPU
7386 processor. Access to memory in the @code{__ea} address space involves
7387 issuing DMA operations to move data between the host processor and the
7388 local processor memory address space. Pointers in the @code{__ea}
7389 address space are either 32 bits or 64 bits based on the
7390 @option{-mea32} or @option{-mea64} switches (native SPU pointers are
7393 Internally, address spaces are represented as a small integer in the
7394 range 0 to 15 with address space 0 being reserved for the generic
7397 To register a named address space qualifier keyword with the C front end,
7398 the target may call the @code{c_register_addr_space} routine. For example,
7399 the SPU port uses the following to declare @code{__ea} as the keyword for
7400 named address space #1:
7402 #define ADDR_SPACE_EA 1
7403 c_register_addr_space ("__ea", ADDR_SPACE_EA);
7406 @hook TARGET_ADDR_SPACE_POINTER_MODE
7408 @hook TARGET_ADDR_SPACE_ADDRESS_MODE
7410 @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
7412 @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
7414 @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
7416 @hook TARGET_ADDR_SPACE_SUBSET_P
7418 @hook TARGET_ADDR_SPACE_CONVERT
7421 @section Miscellaneous Parameters
7422 @cindex parameters, miscellaneous
7424 @c prevent bad page break with this line
7425 Here are several miscellaneous parameters.
7427 @defmac HAS_LONG_COND_BRANCH
7428 Define this boolean macro to indicate whether or not your architecture
7429 has conditional branches that can span all of memory. It is used in
7430 conjunction with an optimization that partitions hot and cold basic
7431 blocks into separate sections of the executable. If this macro is
7432 set to false, gcc will convert any conditional branches that attempt
7433 to cross between sections into unconditional branches or indirect jumps.
7436 @defmac HAS_LONG_UNCOND_BRANCH
7437 Define this boolean macro to indicate whether or not your architecture
7438 has unconditional branches that can span all of memory. It is used in
7439 conjunction with an optimization that partitions hot and cold basic
7440 blocks into separate sections of the executable. If this macro is
7441 set to false, gcc will convert any unconditional branches that attempt
7442 to cross between sections into indirect jumps.
7445 @defmac CASE_VECTOR_MODE
7446 An alias for a machine mode name. This is the machine mode that
7447 elements of a jump-table should have.
7450 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7451 Optional: return the preferred mode for an @code{addr_diff_vec}
7452 when the minimum and maximum offset are known. If you define this,
7453 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7454 To make this work, you also have to define @code{INSN_ALIGN} and
7455 make the alignment for @code{addr_diff_vec} explicit.
7456 The @var{body} argument is provided so that the offset_unsigned and scale
7457 flags can be updated.
7460 @defmac CASE_VECTOR_PC_RELATIVE
7461 Define this macro to be a C expression to indicate when jump-tables
7462 should contain relative addresses. You need not define this macro if
7463 jump-tables never contain relative addresses, or jump-tables should
7464 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
7468 @hook TARGET_CASE_VALUES_THRESHOLD
7470 @defmac WORD_REGISTER_OPERATIONS
7471 Define this macro if operations between registers with integral mode
7472 smaller than a word are always performed on the entire register.
7473 Most RISC machines have this property and most CISC machines do not.
7476 @defmac LOAD_EXTEND_OP (@var{mem_mode})
7477 Define this macro to be a C expression indicating when insns that read
7478 memory in @var{mem_mode}, an integral mode narrower than a word, set the
7479 bits outside of @var{mem_mode} to be either the sign-extension or the
7480 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
7481 of @var{mem_mode} for which the
7482 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7483 @code{UNKNOWN} for other modes.
7485 This macro is not called with @var{mem_mode} non-integral or with a width
7486 greater than or equal to @code{BITS_PER_WORD}, so you may return any
7487 value in this case. Do not define this macro if it would always return
7488 @code{UNKNOWN}. On machines where this macro is defined, you will normally
7489 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7491 You may return a non-@code{UNKNOWN} value even if for some hard registers
7492 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
7493 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
7494 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
7495 integral mode larger than this but not larger than @code{word_mode}.
7497 You must return @code{UNKNOWN} if for some hard registers that allow this
7498 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
7499 @code{word_mode}, but that they can change to another integral mode that
7500 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
7503 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
7504 Define this macro if loading short immediate values into registers sign
7508 @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
7511 The maximum number of bytes that a single instruction can move quickly
7512 between memory and registers or between two memory locations.
7515 @defmac MAX_MOVE_MAX
7516 The maximum number of bytes that a single instruction can move quickly
7517 between memory and registers or between two memory locations. If this
7518 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
7519 constant value that is the largest value that @code{MOVE_MAX} can have
7523 @defmac SHIFT_COUNT_TRUNCATED
7524 A C expression that is nonzero if on this machine the number of bits
7525 actually used for the count of a shift operation is equal to the number
7526 of bits needed to represent the size of the object being shifted. When
7527 this macro is nonzero, the compiler will assume that it is safe to omit
7528 a sign-extend, zero-extend, and certain bitwise `and' instructions that
7529 truncates the count of a shift operation. On machines that have
7530 instructions that act on bit-fields at variable positions, which may
7531 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7532 also enables deletion of truncations of the values that serve as
7533 arguments to bit-field instructions.
7535 If both types of instructions truncate the count (for shifts) and
7536 position (for bit-field operations), or if no variable-position bit-field
7537 instructions exist, you should define this macro.
7539 However, on some machines, such as the 80386 and the 680x0, truncation
7540 only applies to shift operations and not the (real or pretended)
7541 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7542 such machines. Instead, add patterns to the @file{md} file that include
7543 the implied truncation of the shift instructions.
7545 You need not define this macro if it would always have the value of zero.
7548 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
7549 @hook TARGET_SHIFT_TRUNCATION_MASK
7551 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
7552 A C expression which is nonzero if on this machine it is safe to
7553 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
7554 bits (where @var{outprec} is smaller than @var{inprec}) by merely
7555 operating on it as if it had only @var{outprec} bits.
7557 On many machines, this expression can be 1.
7559 @c rearranged this, removed the phrase "it is reported that". this was
7560 @c to fix an overfull hbox. --mew 10feb93
7561 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
7562 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
7563 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
7564 such cases may improve things.
7567 @hook TARGET_MODE_REP_EXTENDED
7569 @defmac STORE_FLAG_VALUE
7570 A C expression describing the value returned by a comparison operator
7571 with an integral mode and stored by a store-flag instruction
7572 (@samp{cstore@var{mode}4}) when the condition is true. This description must
7573 apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
7574 comparison operators whose results have a @code{MODE_INT} mode.
7576 A value of 1 or @minus{}1 means that the instruction implementing the
7577 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
7578 and 0 when the comparison is false. Otherwise, the value indicates
7579 which bits of the result are guaranteed to be 1 when the comparison is
7580 true. This value is interpreted in the mode of the comparison
7581 operation, which is given by the mode of the first operand in the
7582 @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of
7583 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
7586 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
7587 generate code that depends only on the specified bits. It can also
7588 replace comparison operators with equivalent operations if they cause
7589 the required bits to be set, even if the remaining bits are undefined.
7590 For example, on a machine whose comparison operators return an
7591 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7592 @samp{0x80000000}, saying that just the sign bit is relevant, the
7596 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7603 (ashift:SI @var{x} (const_int @var{n}))
7607 where @var{n} is the appropriate shift count to move the bit being
7608 tested into the sign bit.
7610 There is no way to describe a machine that always sets the low-order bit
7611 for a true value, but does not guarantee the value of any other bits,
7612 but we do not know of any machine that has such an instruction. If you
7613 are trying to port GCC to such a machine, include an instruction to
7614 perform a logical-and of the result with 1 in the pattern for the
7615 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
7617 Often, a machine will have multiple instructions that obtain a value
7618 from a comparison (or the condition codes). Here are rules to guide the
7619 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7624 Use the shortest sequence that yields a valid definition for
7625 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
7626 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7627 comparison operators to do so because there may be opportunities to
7628 combine the normalization with other operations.
7631 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
7632 slightly preferred on machines with expensive jumps and 1 preferred on
7636 As a second choice, choose a value of @samp{0x80000001} if instructions
7637 exist that set both the sign and low-order bits but do not define the
7641 Otherwise, use a value of @samp{0x80000000}.
7644 Many machines can produce both the value chosen for
7645 @code{STORE_FLAG_VALUE} and its negation in the same number of
7646 instructions. On those machines, you should also define a pattern for
7647 those cases, e.g., one matching
7650 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7653 Some machines can also perform @code{and} or @code{plus} operations on
7654 condition code values with less instructions than the corresponding
7655 @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those
7656 machines, define the appropriate patterns. Use the names @code{incscc}
7657 and @code{decscc}, respectively, for the patterns which perform
7658 @code{plus} or @code{minus} operations on condition code values. See
7659 @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to
7660 find such instruction sequences on other machines.
7662 If this macro is not defined, the default value, 1, is used. You need
7663 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7664 instructions, or if the value generated by these instructions is 1.
7667 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
7668 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
7669 returned when comparison operators with floating-point results are true.
7670 Define this macro on machines that have comparison operations that return
7671 floating-point values. If there are no such operations, do not define
7675 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
7676 A C expression that gives a rtx representing the nonzero true element
7677 for vector comparisons. The returned rtx should be valid for the inner
7678 mode of @var{mode} which is guaranteed to be a vector mode. Define
7679 this macro on machines that have vector comparison operations that
7680 return a vector result. If there are no such operations, do not define
7681 this macro. Typically, this macro is defined as @code{const1_rtx} or
7682 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
7683 the compiler optimizing such vector comparison operations for the
7687 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7688 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7689 A C expression that indicates whether the architecture defines a value
7690 for @code{clz} or @code{ctz} with a zero operand.
7691 A result of @code{0} indicates the value is undefined.
7692 If the value is defined for only the RTL expression, the macro should
7693 evaluate to @code{1}; if the value applies also to the corresponding optab
7694 entry (which is normally the case if it expands directly into
7695 the corresponding RTL), then the macro should evaluate to @code{2}.
7696 In the cases where the value is defined, @var{value} should be set to
7699 If this macro is not defined, the value of @code{clz} or
7700 @code{ctz} at zero is assumed to be undefined.
7702 This macro must be defined if the target's expansion for @code{ffs}
7703 relies on a particular value to get correct results. Otherwise it
7704 is not necessary, though it may be used to optimize some corner cases, and
7705 to provide a default expansion for the @code{ffs} optab.
7707 Note that regardless of this macro the ``definedness'' of @code{clz}
7708 and @code{ctz} at zero do @emph{not} extend to the builtin functions
7709 visible to the user. Thus one may be free to adjust the value at will
7710 to match the target expansion of these operations without fear of
7715 An alias for the machine mode for pointers. On most machines, define
7716 this to be the integer mode corresponding to the width of a hardware
7717 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7718 On some machines you must define this to be one of the partial integer
7719 modes, such as @code{PSImode}.
7721 The width of @code{Pmode} must be at least as large as the value of
7722 @code{POINTER_SIZE}. If it is not equal, you must define the macro
7723 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7727 @defmac FUNCTION_MODE
7728 An alias for the machine mode used for memory references to functions
7729 being called, in @code{call} RTL expressions. On most CISC machines,
7730 where an instruction can begin at any byte address, this should be
7731 @code{QImode}. On most RISC machines, where all instructions have fixed
7732 size and alignment, this should be a mode with the same size and alignment
7733 as the machine instruction words - typically @code{SImode} or @code{HImode}.
7736 @defmac STDC_0_IN_SYSTEM_HEADERS
7737 In normal operation, the preprocessor expands @code{__STDC__} to the
7738 constant 1, to signify that GCC conforms to ISO Standard C@. On some
7739 hosts, like Solaris, the system compiler uses a different convention,
7740 where @code{__STDC__} is normally 0, but is 1 if the user specifies
7741 strict conformance to the C Standard.
7743 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
7744 convention when processing system header files, but when processing user
7745 files @code{__STDC__} will always expand to 1.
7748 @hook TARGET_C_PREINCLUDE
7750 @hook TARGET_CXX_IMPLICIT_EXTERN_C
7752 @defmac NO_IMPLICIT_EXTERN_C
7753 Define this macro if the system header files support C++ as well as C@.
7754 This macro inhibits the usual method of using system header files in
7755 C++, which is to pretend that the file's contents are enclosed in
7756 @samp{extern "C" @{@dots{}@}}.
7761 @defmac REGISTER_TARGET_PRAGMAS ()
7762 Define this macro if you want to implement any target-specific pragmas.
7763 If defined, it is a C expression which makes a series of calls to
7764 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
7765 for each pragma. The macro may also do any
7766 setup required for the pragmas.
7768 The primary reason to define this macro is to provide compatibility with
7769 other compilers for the same target. In general, we discourage
7770 definition of target-specific pragmas for GCC@.
7772 If the pragma can be implemented by attributes then you should consider
7773 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
7775 Preprocessor macros that appear on pragma lines are not expanded. All
7776 @samp{#pragma} directives that do not match any registered pragma are
7777 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
7780 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7781 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7783 Each call to @code{c_register_pragma} or
7784 @code{c_register_pragma_with_expansion} establishes one pragma. The
7785 @var{callback} routine will be called when the preprocessor encounters a
7789 #pragma [@var{space}] @var{name} @dots{}
7792 @var{space} is the case-sensitive namespace of the pragma, or
7793 @code{NULL} to put the pragma in the global namespace. The callback
7794 routine receives @var{pfile} as its first argument, which can be passed
7795 on to cpplib's functions if necessary. You can lex tokens after the
7796 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
7797 callback will be silently ignored. The end of the line is indicated by
7798 a token of type @code{CPP_EOF}. Macro expansion occurs on the
7799 arguments of pragmas registered with
7800 @code{c_register_pragma_with_expansion} but not on the arguments of
7801 pragmas registered with @code{c_register_pragma}.
7803 Note that the use of @code{pragma_lex} is specific to the C and C++
7804 compilers. It will not work in the Java or Fortran compilers, or any
7805 other language compilers for that matter. Thus if @code{pragma_lex} is going
7806 to be called from target-specific code, it must only be done so when
7807 building the C and C++ compilers. This can be done by defining the
7808 variables @code{c_target_objs} and @code{cxx_target_objs} in the
7809 target entry in the @file{config.gcc} file. These variables should name
7810 the target-specific, language-specific object file which contains the
7811 code that uses @code{pragma_lex}. Note it will also be necessary to add a
7812 rule to the makefile fragment pointed to by @code{tmake_file} that shows
7813 how to build this object file.
7816 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
7817 Define this macro if macros should be expanded in the
7818 arguments of @samp{#pragma pack}.
7821 @defmac TARGET_DEFAULT_PACK_STRUCT
7822 If your target requires a structure packing default other than 0 (meaning
7823 the machine default), define this macro to the necessary value (in bytes).
7824 This must be a value that would also be valid to use with
7825 @samp{#pragma pack()} (that is, a small power of two).
7828 @defmac DOLLARS_IN_IDENTIFIERS
7829 Define this macro to control use of the character @samp{$} in
7830 identifier names for the C family of languages. 0 means @samp{$} is
7831 not allowed by default; 1 means it is allowed. 1 is the default;
7832 there is no need to define this macro in that case.
7835 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
7836 Define this macro as a C expression that is nonzero if it is safe for the
7837 delay slot scheduler to place instructions in the delay slot of @var{insn},
7838 even if they appear to use a resource set or clobbered in @var{insn}.
7839 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
7840 every @code{call_insn} has this behavior. On machines where some @code{insn}
7841 or @code{jump_insn} is really a function call and hence has this behavior,
7842 you should define this macro.
7844 You need not define this macro if it would always return zero.
7847 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
7848 Define this macro as a C expression that is nonzero if it is safe for the
7849 delay slot scheduler to place instructions in the delay slot of @var{insn},
7850 even if they appear to set or clobber a resource referenced in @var{insn}.
7851 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
7852 some @code{insn} or @code{jump_insn} is really a function call and its operands
7853 are registers whose use is actually in the subroutine it calls, you should
7854 define this macro. Doing so allows the delay slot scheduler to move
7855 instructions which copy arguments into the argument registers into the delay
7858 You need not define this macro if it would always return zero.
7861 @defmac MULTIPLE_SYMBOL_SPACES
7862 Define this macro as a C expression that is nonzero if, in some cases,
7863 global symbols from one translation unit may not be bound to undefined
7864 symbols in another translation unit without user intervention. For
7865 instance, under Microsoft Windows symbols must be explicitly imported
7866 from shared libraries (DLLs).
7868 You need not define this macro if it would always evaluate to zero.
7871 @hook TARGET_MD_ASM_CLOBBERS
7873 @defmac MATH_LIBRARY
7874 Define this macro as a C string constant for the linker argument to link
7875 in the system math library, minus the initial @samp{"-l"}, or
7876 @samp{""} if the target does not have a
7877 separate math library.
7879 You need only define this macro if the default of @samp{"m"} is wrong.
7882 @defmac LIBRARY_PATH_ENV
7883 Define this macro as a C string constant for the environment variable that
7884 specifies where the linker should look for libraries.
7886 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
7890 @defmac TARGET_POSIX_IO
7891 Define this macro if the target supports the following POSIX@ file
7892 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
7893 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
7894 to use file locking when exiting a program, which avoids race conditions
7895 if the program has forked. It will also create directories at run-time
7896 for cross-profiling.
7899 @defmac MAX_CONDITIONAL_EXECUTE
7901 A C expression for the maximum number of instructions to execute via
7902 conditional execution instructions instead of a branch. A value of
7903 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
7904 1 if it does use cc0.
7907 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
7908 Used if the target needs to perform machine-dependent modifications on the
7909 conditionals used for turning basic blocks into conditionally executed code.
7910 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
7911 contains information about the currently processed blocks. @var{true_expr}
7912 and @var{false_expr} are the tests that are used for converting the
7913 then-block and the else-block, respectively. Set either @var{true_expr} or
7914 @var{false_expr} to a null pointer if the tests cannot be converted.
7917 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
7918 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
7919 if-statements into conditions combined by @code{and} and @code{or} operations.
7920 @var{bb} contains the basic block that contains the test that is currently
7921 being processed and about to be turned into a condition.
7924 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
7925 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
7926 be converted to conditional execution format. @var{ce_info} points to
7927 a data structure, @code{struct ce_if_block}, which contains information
7928 about the currently processed blocks.
7931 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
7932 A C expression to perform any final machine dependent modifications in
7933 converting code to conditional execution. The involved basic blocks
7934 can be found in the @code{struct ce_if_block} structure that is pointed
7935 to by @var{ce_info}.
7938 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
7939 A C expression to cancel any machine dependent modifications in
7940 converting code to conditional execution. The involved basic blocks
7941 can be found in the @code{struct ce_if_block} structure that is pointed
7942 to by @var{ce_info}.
7945 @defmac IFCVT_MACHDEP_INIT (@var{ce_info})
7946 A C expression to initialize any machine specific data for if-conversion
7947 of the if-block in the @code{struct ce_if_block} structure that is pointed
7948 to by @var{ce_info}.
7951 @hook TARGET_MACHINE_DEPENDENT_REORG
7953 @hook TARGET_INIT_BUILTINS
7955 @hook TARGET_BUILTIN_DECL
7957 @hook TARGET_EXPAND_BUILTIN
7959 @hook TARGET_RESOLVE_OVERLOADED_BUILTIN
7961 @hook TARGET_FOLD_BUILTIN
7963 @hook TARGET_GIMPLE_FOLD_BUILTIN
7965 @hook TARGET_COMPARE_VERSION_PRIORITY
7967 @hook TARGET_GET_FUNCTION_VERSIONS_DISPATCHER
7969 @hook TARGET_GENERATE_VERSION_DISPATCHER_BODY
7971 @hook TARGET_CAN_USE_DOLOOP_P
7973 @hook TARGET_INVALID_WITHIN_DOLOOP
7975 @hook TARGET_LEGITIMATE_COMBINED_INSN
7977 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
7979 Take a branch insn in @var{branch1} and another in @var{branch2}.
7980 Return true if redirecting @var{branch1} to the destination of
7981 @var{branch2} is possible.
7983 On some targets, branches may have a limited range. Optimizing the
7984 filling of delay slots can result in branches being redirected, and this
7985 may in turn cause a branch offset to overflow.
7988 @hook TARGET_CAN_FOLLOW_JUMP
7990 @hook TARGET_COMMUTATIVE_P
7992 @hook TARGET_ALLOCATE_INITIAL_VALUE
7994 @hook TARGET_UNSPEC_MAY_TRAP_P
7996 @hook TARGET_SET_CURRENT_FUNCTION
7998 @defmac TARGET_OBJECT_SUFFIX
7999 Define this macro to be a C string representing the suffix for object
8000 files on your target machine. If you do not define this macro, GCC will
8001 use @samp{.o} as the suffix for object files.
8004 @defmac TARGET_EXECUTABLE_SUFFIX
8005 Define this macro to be a C string representing the suffix to be
8006 automatically added to executable files on your target machine. If you
8007 do not define this macro, GCC will use the null string as the suffix for
8011 @defmac COLLECT_EXPORT_LIST
8012 If defined, @code{collect2} will scan the individual object files
8013 specified on its command line and create an export list for the linker.
8014 Define this macro for systems like AIX, where the linker discards
8015 object files that are not referenced from @code{main} and uses export
8019 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
8020 Define this macro to a C expression representing a variant of the
8021 method call @var{mdecl}, if Java Native Interface (JNI) methods
8022 must be invoked differently from other methods on your target.
8023 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
8024 the @code{stdcall} calling convention and this macro is then
8025 defined as this expression:
8028 build_type_attribute_variant (@var{mdecl},
8030 (get_identifier ("stdcall"),
8035 @hook TARGET_CANNOT_MODIFY_JUMPS_P
8037 @hook TARGET_BRANCH_TARGET_REGISTER_CLASS
8039 @hook TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED
8041 @hook TARGET_HAVE_CONDITIONAL_EXECUTION
8043 @hook TARGET_LOOP_UNROLL_ADJUST
8045 @defmac POWI_MAX_MULTS
8046 If defined, this macro is interpreted as a signed integer C expression
8047 that specifies the maximum number of floating point multiplications
8048 that should be emitted when expanding exponentiation by an integer
8049 constant inline. When this value is defined, exponentiation requiring
8050 more than this number of multiplications is implemented by calling the
8051 system library's @code{pow}, @code{powf} or @code{powl} routines.
8052 The default value places no upper bound on the multiplication count.
8055 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8056 This target hook should register any extra include files for the
8057 target. The parameter @var{stdinc} indicates if normal include files
8058 are present. The parameter @var{sysroot} is the system root directory.
8059 The parameter @var{iprefix} is the prefix for the gcc directory.
8062 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8063 This target hook should register any extra include files for the
8064 target before any standard headers. The parameter @var{stdinc}
8065 indicates if normal include files are present. The parameter
8066 @var{sysroot} is the system root directory. The parameter
8067 @var{iprefix} is the prefix for the gcc directory.
8070 @deftypefn Macro void TARGET_OPTF (char *@var{path})
8071 This target hook should register special include paths for the target.
8072 The parameter @var{path} is the include to register. On Darwin
8073 systems, this is used for Framework includes, which have semantics
8074 that are different from @option{-I}.
8077 @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
8078 This target macro returns @code{true} if it is safe to use a local alias
8079 for a virtual function @var{fndecl} when constructing thunks,
8080 @code{false} otherwise. By default, the macro returns @code{true} for all
8081 functions, if a target supports aliases (i.e.@: defines
8082 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
8085 @defmac TARGET_FORMAT_TYPES
8086 If defined, this macro is the name of a global variable containing
8087 target-specific format checking information for the @option{-Wformat}
8088 option. The default is to have no target-specific format checks.
8091 @defmac TARGET_N_FORMAT_TYPES
8092 If defined, this macro is the number of entries in
8093 @code{TARGET_FORMAT_TYPES}.
8096 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
8097 If defined, this macro is the name of a global variable containing
8098 target-specific format overrides for the @option{-Wformat} option. The
8099 default is to have no target-specific format overrides. If defined,
8100 @code{TARGET_FORMAT_TYPES} must be defined, too.
8103 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
8104 If defined, this macro specifies the number of entries in
8105 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
8108 @defmac TARGET_OVERRIDES_FORMAT_INIT
8109 If defined, this macro specifies the optional initialization
8110 routine for target specific customizations of the system printf
8111 and scanf formatter settings.
8114 @hook TARGET_RELAXED_ORDERING
8116 @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
8118 @hook TARGET_INVALID_CONVERSION
8120 @hook TARGET_INVALID_UNARY_OP
8122 @hook TARGET_INVALID_BINARY_OP
8124 @hook TARGET_INVALID_PARAMETER_TYPE
8126 @hook TARGET_INVALID_RETURN_TYPE
8128 @hook TARGET_PROMOTED_TYPE
8130 @hook TARGET_CONVERT_TO_TYPE
8132 @defmac TARGET_USE_JCR_SECTION
8133 This macro determines whether to use the JCR section to register Java
8134 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
8135 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
8139 This macro determines the size of the objective C jump buffer for the
8140 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
8143 @defmac LIBGCC2_UNWIND_ATTRIBUTE
8144 Define this macro if any target-specific attributes need to be attached
8145 to the functions in @file{libgcc} that provide low-level support for
8146 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
8147 and the associated definitions of those functions.
8150 @hook TARGET_UPDATE_STACK_BOUNDARY
8152 @hook TARGET_GET_DRAP_RTX
8154 @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
8156 @hook TARGET_CONST_ANCHOR
8158 @hook TARGET_ASAN_SHADOW_OFFSET
8160 @hook TARGET_MEMMODEL_CHECK
8162 @hook TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
8164 @hook TARGET_HAS_IFUNC_P
8166 @hook TARGET_ATOMIC_ALIGN_FOR_MODE
8168 @hook TARGET_ATOMIC_ASSIGN_EXPAND_FENV
8170 @defmac TARGET_SUPPORTS_WIDE_INT
8172 On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
8173 objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
8174 to indicate that large integers are stored in
8175 @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows
8176 very large integer constants to be represented. @code{CONST_DOUBLE}
8177 is limited to twice the size of the host's @code{HOST_WIDE_INT}
8180 Converting a port mostly requires looking for the places where
8181 @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
8182 code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i
8183 const_double"} at the port level gets you to 95% of the changes that
8184 need to be made. There are a few places that require a deeper look.
8188 There is no equivalent to @code{hval} and @code{lval} for
8189 @code{CONST_WIDE_INT}s. This would be difficult to express in the md
8190 language since there are a variable number of elements.
8192 Most ports only check that @code{hval} is either 0 or -1 to see if the
8193 value is small. As mentioned above, this will no longer be necessary
8194 since small constants are always @code{CONST_INT}. Of course there
8195 are still a few exceptions, the alpha's constraint used by the zap
8196 instruction certainly requires careful examination by C code.
8197 However, all the current code does is pass the hval and lval to C
8198 code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
8199 not really a large change.
8202 Because there is no standard template that ports use to materialize
8203 constants, there is likely to be some futzing that is unique to each
8207 The rtx costs may have to be adjusted to properly account for larger
8208 constants that are represented as @code{CONST_WIDE_INT}.
8211 All and all it does not take long to convert ports that the
8212 maintainer is familiar with.