xref: /netbsd/external/gpl3/gcc/dist/gcc/doc/tm.texi.in (revision f0fbc68b)
1@c Copyright (C) 1988-2022 Free Software Foundation, Inc.
2@c This is part of the GCC manual.
3@c For copying conditions, see the file gcc.texi.
4
5@node Target Macros
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
11
12In addition to the file @file{@var{machine}.md}, a machine description
13includes a C header file conventionally given the name
14@file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
15The header file defines numerous macros that convey the information
16about 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
20source file defines a variable @code{targetm}, which is a structure
21containing pointers to functions and data relating to the target
22machine.  @file{@var{machine}.c} should also contain their definitions,
23if they are not defined elsewhere in GCC, and other functions called
24through the macros defined in the @file{.h} file.
25
26@menu
27* Target Structure::    The @code{targetm} variable.
28* Driver::              Controlling how the driver runs the compilation passes.
29* Run-time Target::     Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
30* Per-Function Data::   Defining data structures for per-function information.
31* Storage Layout::      Defining sizes and alignments of data.
32* Type Layout::         Defining sizes and properties of basic user data types.
33* Registers::           Naming and describing the hardware registers.
34* Register Classes::    Defining the classes of hardware registers.
35* Stack and Calling::   Defining which way the stack grows and by how much.
36* Varargs::             Defining the varargs macros.
37* Trampolines::         Code set up at run time to enter a nested function.
38* Library Calls::       Controlling how library routines are implicitly called.
39* Addressing Modes::    Defining addressing modes valid for memory operands.
40* Anchored Addresses::  Defining how @option{-fsection-anchors} should work.
41* Condition Code::      Defining how insns update the condition code.
42* Costs::               Defining relative costs of different operations.
43* Scheduling::          Adjusting the behavior of the instruction scheduler.
44* Sections::            Dividing storage into text, data, and other sections.
45* PIC::                 Macros for position independent code.
46* Assembler Format::    Defining how to write insns and pseudo-ops to output.
47* Debugging Info::      Defining the format of debugging output.
48* Floating Point::      Handling floating point for cross-compilers.
49* Mode Switching::      Insertion of mode-switching instructions.
50* Target Attributes::   Defining target-specific uses of @code{__attribute__}.
51* Emulated TLS::        Emulated TLS support.
52* MIPS Coprocessors::   MIPS coprocessor support and how to customize it.
53* PCH Target::          Validity checking for precompiled headers.
54* C++ ABI::             Controlling C++ ABI changes.
55* D Language and ABI::  Controlling D ABI changes.
56* Named Address Spaces:: Adding support for named address spaces
57* Misc::                Everything else.
58@end menu
59
60@node Target Structure
61@section The Global @code{targetm} Variable
62@cindex target hooks
63@cindex target functions
64
65@deftypevar {struct gcc_target} targetm
66The target @file{.c} file must define the global @code{targetm} variable
67which contains pointers to functions and data relating to the target
68machine.  The variable is declared in @file{target.h};
69@file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
70used to initialize the variable, and macros for the default initializers
71for elements of the structure.  The @file{.c} file should override those
72macros for which the default definition is inappropriate.  For example:
73@smallexample
74#include "target.h"
75#include "target-def.h"
76
77/* @r{Initialize the GCC target structure.}  */
78
79#undef TARGET_COMP_TYPE_ATTRIBUTES
80#define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
81
82struct gcc_target targetm = TARGET_INITIALIZER;
83@end smallexample
84@end deftypevar
85
86Where a macro should be defined in the @file{.c} file in this manner to
87form part of the @code{targetm} structure, it is documented below as a
88``Target Hook'' with a prototype.  Many macros will change in future
89from being defined in the @file{.h} file to being part of the
90@code{targetm} structure.
91
92Similarly, there is a @code{targetcm} variable for hooks that are
93specific to front ends for C-family languages, documented as ``C
94Target Hook''.  This is declared in @file{c-family/c-target.h}, the
95initializer @code{TARGETCM_INITIALIZER} in
96@file{c-family/c-target-def.h}.  If targets initialize @code{targetcm}
97themselves, they should set @code{target_has_targetcm=yes} in
98@file{config.gcc}; otherwise a default definition is used.
99
100Similarly, there is a @code{targetm_common} variable for hooks that
101are shared between the compiler driver and the compilers proper,
102documented as ``Common Target Hook''.  This is declared in
103@file{common/common-target.h}, the initializer
104@code{TARGETM_COMMON_INITIALIZER} in
105@file{common/common-target-def.h}.  If targets initialize
106@code{targetm_common} themselves, they should set
107@code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a
108default definition is used.
109
110Similarly, there is a @code{targetdm} variable for hooks that are
111specific to the D language front end, documented as ``D Target Hook''.
112This is declared in @file{d/d-target.h}, the initializer
113@code{TARGETDM_INITIALIZER} in @file{d/d-target-def.h}.  If targets
114initialize @code{targetdm} themselves, they should set
115@code{target_has_targetdm=yes} in @file{config.gcc}; otherwise a default
116definition is used.
117
118@node Driver
119@section Controlling the Compilation Driver, @file{gcc}
120@cindex driver
121@cindex controlling the compilation driver
122
123@c prevent bad page break with this line
124You can control the compilation driver.
125
126@defmac DRIVER_SELF_SPECS
127A list of specs for the driver itself.  It should be a suitable
128initializer for an array of strings, with no surrounding braces.
129
130The driver applies these specs to its own command line between loading
131default @file{specs} files (but not command-line specified ones) and
132choosing the multilib directory or running any subcommands.  It
133applies them in the order given, so each spec can depend on the
134options added by earlier ones.  It is also possible to remove options
135using @samp{%<@var{option}} in the usual way.
136
137This macro can be useful when a port has several interdependent target
138options.  It provides a way of standardizing the command line so
139that the other specs are easier to write.
140
141Do not define this macro if it does not need to do anything.
142@end defmac
143
144@defmac OPTION_DEFAULT_SPECS
145A list of specs used to support configure-time default options (i.e.@:
146@option{--with} options) in the driver.  It should be a suitable initializer
147for an array of structures, each containing two strings, without the
148outermost pair of surrounding braces.
149
150The first item in the pair is the name of the default.  This must match
151the code in @file{config.gcc} for the target.  The second item is a spec
152to apply if a default with this name was specified.  The string
153@samp{%(VALUE)} in the spec will be replaced by the value of the default
154everywhere it occurs.
155
156The driver will apply these specs to its own command line between loading
157default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
158the same mechanism as @code{DRIVER_SELF_SPECS}.
159
160Do not define this macro if it does not need to do anything.
161@end defmac
162
163@defmac CPP_SPEC
164A C string constant that tells the GCC driver program options to
165pass to CPP@.  It can also specify how to translate options you
166give to GCC into options for GCC to pass to the CPP@.
167
168Do not define this macro if it does not need to do anything.
169@end defmac
170
171@defmac CPLUSPLUS_CPP_SPEC
172This macro is just like @code{CPP_SPEC}, but is used for C++, rather
173than C@.  If you do not define this macro, then the value of
174@code{CPP_SPEC} (if any) will be used instead.
175@end defmac
176
177@defmac CC1_SPEC
178A C string constant that tells the GCC driver program options to
179pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
180front ends.
181It can also specify how to translate options you give to GCC into options
182for GCC to pass to front ends.
183
184Do not define this macro if it does not need to do anything.
185@end defmac
186
187@defmac CC1PLUS_SPEC
188A C string constant that tells the GCC driver program options to
189pass to @code{cc1plus}.  It can also specify how to translate options you
190give to GCC into options for GCC to pass to the @code{cc1plus}.
191
192Do not define this macro if it does not need to do anything.
193Note that everything defined in CC1_SPEC is already passed to
194@code{cc1plus} so there is no need to duplicate the contents of
195CC1_SPEC in CC1PLUS_SPEC@.
196@end defmac
197
198@defmac ASM_SPEC
199A C string constant that tells the GCC driver program options to
200pass to the assembler.  It can also specify how to translate options
201you give to GCC into options for GCC to pass to the assembler.
202See the file @file{sun3.h} for an example of this.
203
204Do not define this macro if it does not need to do anything.
205@end defmac
206
207@defmac ASM_FINAL_SPEC
208A C string constant that tells the GCC driver program how to
209run any programs which cleanup after the normal assembler.
210Normally, this is not needed.  See the file @file{mips.h} for
211an example of this.
212
213Do not define this macro if it does not need to do anything.
214@end defmac
215
216@defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
217Define this macro, with no value, if the driver should give the assembler
218an argument consisting of a single dash, @option{-}, to instruct it to
219read from its standard input (which will be a pipe connected to the
220output of the compiler proper).  This argument is given after any
221@option{-o} option specifying the name of the output file.
222
223If you do not define this macro, the assembler is assumed to read its
224standard input if given no non-option arguments.  If your assembler
225cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
226see @file{mips.h} for instance.
227@end defmac
228
229@defmac LINK_SPEC
230A C string constant that tells the GCC driver program options to
231pass to the linker.  It can also specify how to translate options you
232give to GCC into options for GCC to pass to the linker.
233
234Do not define this macro if it does not need to do anything.
235@end defmac
236
237@defmac LIB_SPEC
238Another C string constant used much like @code{LINK_SPEC}.  The difference
239between the two is that @code{LIB_SPEC} is used at the end of the
240command given to the linker.
241
242If this macro is not defined, a default is provided that
243loads the standard C library from the usual place.  See @file{gcc.cc}.
244@end defmac
245
246@defmac LIBGCC_SPEC
247Another C string constant that tells the GCC driver program
248how and when to place a reference to @file{libgcc.a} into the
249linker command line.  This constant is placed both before and after
250the value of @code{LIB_SPEC}.
251
252If this macro is not defined, the GCC driver provides a default that
253passes the string @option{-lgcc} to the linker.
254@end defmac
255
256@defmac REAL_LIBGCC_SPEC
257By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
258@code{LIBGCC_SPEC} is not directly used by the driver program but is
259instead modified to refer to different versions of @file{libgcc.a}
260depending on the values of the command line flags @option{-static},
261@option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}.  On
262targets where these modifications are inappropriate, define
263@code{REAL_LIBGCC_SPEC} instead.  @code{REAL_LIBGCC_SPEC} tells the
264driver how to place a reference to @file{libgcc} on the link command
265line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
266@end defmac
267
268@defmac USE_LD_AS_NEEDED
269A macro that controls the modifications to @code{LIBGCC_SPEC}
270mentioned in @code{REAL_LIBGCC_SPEC}.  If nonzero, a spec will be
271generated that uses @option{--as-needed} or equivalent options and the
272shared @file{libgcc} in place of the
273static exception handler library, when linking without any of
274@code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
275@end defmac
276
277@defmac LINK_EH_SPEC
278If defined, this C string constant is added to @code{LINK_SPEC}.
279When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
280the modifications to @code{LIBGCC_SPEC} mentioned in
281@code{REAL_LIBGCC_SPEC}.
282@end defmac
283
284@defmac STARTFILE_SPEC
285Another C string constant used much like @code{LINK_SPEC}.  The
286difference between the two is that @code{STARTFILE_SPEC} is used at
287the very beginning of the command given to the linker.
288
289If this macro is not defined, a default is provided that loads the
290standard C startup file from the usual place.  See @file{gcc.cc}.
291@end defmac
292
293@defmac ENDFILE_SPEC
294Another C string constant used much like @code{LINK_SPEC}.  The
295difference between the two is that @code{ENDFILE_SPEC} is used at
296the very end of the command given to the linker.
297
298Do not define this macro if it does not need to do anything.
299@end defmac
300
301@defmac THREAD_MODEL_SPEC
302GCC @code{-v} will print the thread model GCC was configured to use.
303However, this doesn't work on platforms that are multilibbed on thread
304models, such as AIX 4.3.  On such platforms, define
305@code{THREAD_MODEL_SPEC} such that it evaluates to a string without
306blanks that names one of the recognized thread models.  @code{%*}, the
307default value of this macro, will expand to the value of
308@code{thread_file} set in @file{config.gcc}.
309@end defmac
310
311@defmac SYSROOT_SUFFIX_SPEC
312Define this macro to add a suffix to the target sysroot when GCC is
313configured with a sysroot.  This will cause GCC to search for usr/lib,
314et al, within sysroot+suffix.
315@end defmac
316
317@defmac SYSROOT_HEADERS_SUFFIX_SPEC
318Define this macro to add a headers_suffix to the target sysroot when
319GCC is configured with a sysroot.  This will cause GCC to pass the
320updated sysroot+headers_suffix to CPP, causing it to search for
321usr/include, et al, within sysroot+headers_suffix.
322@end defmac
323
324@defmac EXTRA_SPECS
325Define this macro to provide additional specifications to put in the
326@file{specs} file that can be used in various specifications like
327@code{CC1_SPEC}.
328
329The definition should be an initializer for an array of structures,
330containing a string constant, that defines the specification name, and a
331string constant that provides the specification.
332
333Do not define this macro if it does not need to do anything.
334
335@code{EXTRA_SPECS} is useful when an architecture contains several
336related targets, which have various @code{@dots{}_SPECS} which are similar
337to each other, and the maintainer would like one central place to keep
338these definitions.
339
340For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
341define either @code{_CALL_SYSV} when the System V calling sequence is
342used or @code{_CALL_AIX} when the older AIX-based calling sequence is
343used.
344
345The @file{config/rs6000/rs6000.h} target file defines:
346
347@smallexample
348#define EXTRA_SPECS \
349  @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
350
351#define CPP_SYS_DEFAULT ""
352@end smallexample
353
354The @file{config/rs6000/sysv.h} target file defines:
355@smallexample
356#undef CPP_SPEC
357#define CPP_SPEC \
358"%@{posix: -D_POSIX_SOURCE @} \
359%@{mcall-sysv: -D_CALL_SYSV @} \
360%@{!mcall-sysv: %(cpp_sysv_default) @} \
361%@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
362
363#undef CPP_SYSV_DEFAULT
364#define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
365@end smallexample
366
367while the @file{config/rs6000/eabiaix.h} target file defines
368@code{CPP_SYSV_DEFAULT} as:
369
370@smallexample
371#undef CPP_SYSV_DEFAULT
372#define CPP_SYSV_DEFAULT "-D_CALL_AIX"
373@end smallexample
374@end defmac
375
376@defmac LINK_LIBGCC_SPECIAL_1
377Define this macro if the driver program should find the library
378@file{libgcc.a}.  If you do not define this macro, the driver program will pass
379the argument @option{-lgcc} to tell the linker to do the search.
380@end defmac
381
382@defmac LINK_GCC_C_SEQUENCE_SPEC
383The sequence in which libgcc and libc are specified to the linker.
384By default this is @code{%G %L %G}.
385@end defmac
386
387@defmac POST_LINK_SPEC
388Define this macro to add additional steps to be executed after linker.
389The default value of this macro is empty string.
390@end defmac
391
392@defmac LINK_COMMAND_SPEC
393A C string constant giving the complete command line need to execute the
394linker.  When you do this, you will need to update your port each time a
395change is made to the link command line within @file{gcc.cc}.  Therefore,
396define this macro only if you need to completely redefine the command
397line for invoking the linker and there is no other way to accomplish
398the effect you need.  Overriding this macro may be avoidable by overriding
399@code{LINK_GCC_C_SEQUENCE_SPEC} instead.
400@end defmac
401
402@hook TARGET_ALWAYS_STRIP_DOTDOT
403
404@defmac MULTILIB_DEFAULTS
405Define this macro as a C expression for the initializer of an array of
406string to tell the driver program which options are defaults for this
407target and thus do not need to be handled specially when using
408@code{MULTILIB_OPTIONS}.
409
410Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
411the target makefile fragment or if none of the options listed in
412@code{MULTILIB_OPTIONS} are set by default.
413@xref{Target Fragment}.
414@end defmac
415
416@defmac RELATIVE_PREFIX_NOT_LINKDIR
417Define this macro to tell @command{gcc} that it should only translate
418a @option{-B} prefix into a @option{-L} linker option if the prefix
419indicates an absolute file name.
420@end defmac
421
422@defmac MD_EXEC_PREFIX
423If defined, this macro is an additional prefix to try after
424@code{STANDARD_EXEC_PREFIX}.  @code{MD_EXEC_PREFIX} is not searched
425when the compiler is built as a cross
426compiler.  If you define @code{MD_EXEC_PREFIX}, then be sure to add it
427to the list of directories used to find the assembler in @file{configure.ac}.
428@end defmac
429
430@defmac STANDARD_STARTFILE_PREFIX
431Define this macro as a C string constant if you wish to override the
432standard choice of @code{libdir} as the default prefix to
433try when searching for startup files such as @file{crt0.o}.
434@code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
435is built as a cross compiler.
436@end defmac
437
438@defmac STANDARD_STARTFILE_PREFIX_1
439Define this macro as a C string constant if you wish to override the
440standard choice of @code{/lib} as a prefix to try after the default prefix
441when searching for startup files such as @file{crt0.o}.
442@code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
443is built as a cross compiler.
444@end defmac
445
446@defmac STANDARD_STARTFILE_PREFIX_2
447Define this macro as a C string constant if you wish to override the
448standard choice of @code{/lib} as yet another prefix to try after the
449default prefix when searching for startup files such as @file{crt0.o}.
450@code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
451is built as a cross compiler.
452@end defmac
453
454@defmac MD_STARTFILE_PREFIX
455If defined, this macro supplies an additional prefix to try after the
456standard prefixes.  @code{MD_EXEC_PREFIX} is not searched when the
457compiler is built as a cross compiler.
458@end defmac
459
460@defmac MD_STARTFILE_PREFIX_1
461If defined, this macro supplies yet another prefix to try after the
462standard prefixes.  It is not searched when the compiler is built as a
463cross compiler.
464@end defmac
465
466@defmac INIT_ENVIRONMENT
467Define this macro as a C string constant if you wish to set environment
468variables for programs called by the driver, such as the assembler and
469loader.  The driver passes the value of this macro to @code{putenv} to
470initialize the necessary environment variables.
471@end defmac
472
473@defmac LOCAL_INCLUDE_DIR
474Define this macro as a C string constant if you wish to override the
475standard choice of @file{/usr/local/include} as the default prefix to
476try when searching for local header files.  @code{LOCAL_INCLUDE_DIR}
477comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in
478@file{config.gcc}, normally @file{/usr/include}) in the search order.
479
480Cross compilers do not search either @file{/usr/local/include} or its
481replacement.
482@end defmac
483
484@defmac NATIVE_SYSTEM_HEADER_COMPONENT
485The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}.
486See @code{INCLUDE_DEFAULTS}, below, for the description of components.
487If you do not define this macro, no component is used.
488@end defmac
489
490@defmac INCLUDE_DEFAULTS
491Define this macro if you wish to override the entire default search path
492for include files.  For a native compiler, the default search path
493usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
494@code{GPLUSPLUS_INCLUDE_DIR}, and
495@code{NATIVE_SYSTEM_HEADER_DIR}.  In addition, @code{GPLUSPLUS_INCLUDE_DIR}
496and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
497and specify private search areas for GCC@.  The directory
498@code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
499
500The definition should be an initializer for an array of structures.
501Each array element should have four elements: the directory name (a
502string constant), the component name (also a string constant), a flag
503for C++-only directories,
504and a flag showing that the includes in the directory don't need to be
505wrapped in @code{extern @samp{C}} when compiling C++.  Mark the end of
506the array with a null element.
507
508The component name denotes what GNU package the include file is part of,
509if any, in all uppercase letters.  For example, it might be @samp{GCC}
510or @samp{BINUTILS}.  If the package is part of a vendor-supplied
511operating system, code the component name as @samp{0}.
512
513For example, here is the definition used for VAX/VMS:
514
515@smallexample
516#define INCLUDE_DEFAULTS \
517@{                                       \
518  @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@},   \
519  @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@},    \
520  @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@},  \
521  @{ ".", 0, 0, 0@},                      \
522  @{ 0, 0, 0, 0@}                         \
523@}
524@end smallexample
525@end defmac
526
527Here is the order of prefixes tried for exec files:
528
529@enumerate
530@item
531Any prefixes specified by the user with @option{-B}.
532
533@item
534The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
535is not set and the compiler has not been installed in the configure-time
536@var{prefix}, the location in which the compiler has actually been installed.
537
538@item
539The directories specified by the environment variable @code{COMPILER_PATH}.
540
541@item
542The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
543in the configured-time @var{prefix}.
544
545@item
546The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
547
548@item
549The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
550
551@item
552The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
553compiler.
554@end enumerate
555
556Here is the order of prefixes tried for startfiles:
557
558@enumerate
559@item
560Any prefixes specified by the user with @option{-B}.
561
562@item
563The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
564value based on the installed toolchain location.
565
566@item
567The directories specified by the environment variable @code{LIBRARY_PATH}
568(or port-specific name; native only, cross compilers do not use this).
569
570@item
571The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
572in the configured @var{prefix} or this is a native compiler.
573
574@item
575The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
576
577@item
578The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
579compiler.
580
581@item
582The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
583native compiler, or we have a target system root.
584
585@item
586The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
587native compiler, or we have a target system root.
588
589@item
590The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
591If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
592the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
593
594@item
595The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
596compiler, or we have a target system root. The default for this macro is
597@file{/lib/}.
598
599@item
600The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
601compiler, or we have a target system root. The default for this macro is
602@file{/usr/lib/}.
603@end enumerate
604
605@node Run-time Target
606@section Run-time Target Specification
607@cindex run-time target specification
608@cindex predefined macros
609@cindex target specifications
610
611@c prevent bad page break with this line
612Here are run-time target specifications.
613
614@defmac TARGET_CPU_CPP_BUILTINS ()
615This function-like macro expands to a block of code that defines
616built-in preprocessor macros and assertions for the target CPU, using
617the functions @code{builtin_define}, @code{builtin_define_std} and
618@code{builtin_assert}.  When the front end
619calls this macro it provides a trailing semicolon, and since it has
620finished command line option processing your code can use those
621results freely.
622
623@code{builtin_assert} takes a string in the form you pass to the
624command-line option @option{-A}, such as @code{cpu=mips}, and creates
625the assertion.  @code{builtin_define} takes a string in the form
626accepted by option @option{-D} and unconditionally defines the macro.
627
628@code{builtin_define_std} takes a string representing the name of an
629object-like macro.  If it doesn't lie in the user's namespace,
630@code{builtin_define_std} defines it unconditionally.  Otherwise, it
631defines a version with two leading underscores, and another version
632with two leading and trailing underscores, and defines the original
633only if an ISO standard was not requested on the command line.  For
634example, passing @code{unix} defines @code{__unix}, @code{__unix__}
635and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
636@code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
637defines only @code{_ABI64}.
638
639You can also test for the C dialect being compiled.  The variable
640@code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
641or @code{clk_objective_c}.  Note that if we are preprocessing
642assembler, this variable will be @code{clk_c} but the function-like
643macro @code{preprocessing_asm_p()} will return true, so you might want
644to check for that first.  If you need to check for strict ANSI, the
645variable @code{flag_iso} can be used.  The function-like macro
646@code{preprocessing_trad_p()} can be used to check for traditional
647preprocessing.
648@end defmac
649
650@defmac TARGET_OS_CPP_BUILTINS ()
651Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
652and is used for the target operating system instead.
653@end defmac
654
655@defmac TARGET_OBJFMT_CPP_BUILTINS ()
656Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
657and is used for the target object format.  @file{elfos.h} uses this
658macro to define @code{__ELF__}, so you probably do not need to define
659it yourself.
660@end defmac
661
662@deftypevar {extern int} target_flags
663This variable is declared in @file{options.h}, which is included before
664any target-specific headers.
665@end deftypevar
666
667@hook TARGET_DEFAULT_TARGET_FLAGS
668This variable specifies the initial value of @code{target_flags}.
669Its default setting is 0.
670@end deftypevr
671
672@cindex optional hardware or system features
673@cindex features, optional, in system conventions
674
675@hook TARGET_HANDLE_OPTION
676This hook is called whenever the user specifies one of the
677target-specific options described by the @file{.opt} definition files
678(@pxref{Options}).  It has the opportunity to do some option-specific
679processing and should return true if the option is valid.  The default
680definition does nothing but return true.
681
682@var{decoded} specifies the option and its arguments.  @var{opts} and
683@var{opts_set} are the @code{gcc_options} structures to be used for
684storing option state, and @var{loc} is the location at which the
685option was passed (@code{UNKNOWN_LOCATION} except for options passed
686via attributes).
687@end deftypefn
688
689@hook TARGET_HANDLE_C_OPTION
690This target hook is called whenever the user specifies one of the
691target-specific C language family options described by the @file{.opt}
692definition files(@pxref{Options}).  It has the opportunity to do some
693option-specific processing and should return true if the option is
694valid.  The arguments are like for @code{TARGET_HANDLE_OPTION}.  The
695default definition does nothing but return false.
696
697In general, you should use @code{TARGET_HANDLE_OPTION} to handle
698options.  However, if processing an option requires routines that are
699only available in the C (and related language) front ends, then you
700should use @code{TARGET_HANDLE_C_OPTION} instead.
701@end deftypefn
702
703@hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT
704
705@hook TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE
706
707@hook TARGET_OBJC_DECLARE_CLASS_DEFINITION
708
709@hook TARGET_STRING_OBJECT_REF_TYPE_P
710
711@hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG
712
713@hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE
714
715@defmac C_COMMON_OVERRIDE_OPTIONS
716This is similar to the @code{TARGET_OPTION_OVERRIDE} hook
717but is only used in the C
718language frontends (C, Objective-C, C++, Objective-C++) and so can be
719used to alter option flag variables which only exist in those
720frontends.
721@end defmac
722
723@hook TARGET_OPTION_OPTIMIZATION_TABLE
724Some machines may desire to change what optimizations are performed for
725various optimization levels.   This variable, if defined, describes
726options to enable at particular sets of optimization levels.  These
727options are processed once
728just after the optimization level is determined and before the remainder
729of the command options have been parsed, so may be overridden by other
730options passed explicitly.
731
732This processing is run once at program startup and when the optimization
733options are changed via @code{#pragma GCC optimize} or by using the
734@code{optimize} attribute.
735@end deftypevr
736
737@hook TARGET_OPTION_INIT_STRUCT
738
739@defmac SWITCHABLE_TARGET
740Some targets need to switch between substantially different subtargets
741during compilation.  For example, the MIPS target has one subtarget for
742the traditional MIPS architecture and another for MIPS16.  Source code
743can switch between these two subarchitectures using the @code{mips16}
744and @code{nomips16} attributes.
745
746Such subtargets can differ in things like the set of available
747registers, the set of available instructions, the costs of various
748operations, and so on.  GCC caches a lot of this type of information
749in global variables, and recomputing them for each subtarget takes a
750significant amount of time.  The compiler therefore provides a facility
751for maintaining several versions of the global variables and quickly
752switching between them; see @file{target-globals.h} for details.
753
754Define this macro to 1 if your target needs this facility.  The default
755is 0.
756@end defmac
757
758@hook TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P
759
760@node Per-Function Data
761@section Defining data structures for per-function information.
762@cindex per-function data
763@cindex data structures
764
765If the target needs to store information on a per-function basis, GCC
766provides a macro and a couple of variables to allow this.  Note, just
767using statics to store the information is a bad idea, since GCC supports
768nested functions, so you can be halfway through encoding one function
769when another one comes along.
770
771GCC defines a data structure called @code{struct function} which
772contains all of the data specific to an individual function.  This
773structure contains a field called @code{machine} whose type is
774@code{struct machine_function *}, which can be used by targets to point
775to their own specific data.
776
777If a target needs per-function specific data it should define the type
778@code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
779This macro should be used to initialize the function pointer
780@code{init_machine_status}.  This pointer is explained below.
781
782One typical use of per-function, target specific data is to create an
783RTX to hold the register containing the function's return address.  This
784RTX can then be used to implement the @code{__builtin_return_address}
785function, for level 0.
786
787Note---earlier implementations of GCC used a single data area to hold
788all of the per-function information.  Thus when processing of a nested
789function began the old per-function data had to be pushed onto a
790stack, and when the processing was finished, it had to be popped off the
791stack.  GCC used to provide function pointers called
792@code{save_machine_status} and @code{restore_machine_status} to handle
793the saving and restoring of the target specific information.  Since the
794single data area approach is no longer used, these pointers are no
795longer supported.
796
797@defmac INIT_EXPANDERS
798Macro called to initialize any target specific information.  This macro
799is called once per function, before generation of any RTL has begun.
800The intention of this macro is to allow the initialization of the
801function pointer @code{init_machine_status}.
802@end defmac
803
804@deftypevar {void (*)(struct function *)} init_machine_status
805If this function pointer is non-@code{NULL} it will be called once per
806function, before function compilation starts, in order to allow the
807target to perform any target specific initialization of the
808@code{struct function} structure.  It is intended that this would be
809used to initialize the @code{machine} of that structure.
810
811@code{struct machine_function} structures are expected to be freed by GC@.
812Generally, any memory that they reference must be allocated by using
813GC allocation, including the structure itself.
814@end deftypevar
815
816@node Storage Layout
817@section Storage Layout
818@cindex storage layout
819
820Note that the definitions of the macros in this table which are sizes or
821alignments measured in bits do not need to be constant.  They can be C
822expressions that refer to static variables, such as the @code{target_flags}.
823@xref{Run-time Target}.
824
825@defmac BITS_BIG_ENDIAN
826Define this macro to have the value 1 if the most significant bit in a
827byte has the lowest number; otherwise define it to have the value zero.
828This means that bit-field instructions count from the most significant
829bit.  If the machine has no bit-field instructions, then this must still
830be defined, but it doesn't matter which value it is defined to.  This
831macro need not be a constant.
832
833This macro does not affect the way structure fields are packed into
834bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
835@end defmac
836
837@defmac BYTES_BIG_ENDIAN
838Define this macro to have the value 1 if the most significant byte in a
839word has the lowest number.  This macro need not be a constant.
840@end defmac
841
842@defmac WORDS_BIG_ENDIAN
843Define this macro to have the value 1 if, in a multiword object, the
844most significant word has the lowest number.  This applies to both
845memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the
846order of words in memory is not the same as the order in registers.  This
847macro need not be a constant.
848@end defmac
849
850@defmac REG_WORDS_BIG_ENDIAN
851On some machines, the order of words in a multiword object differs between
852registers in memory.  In such a situation, define this macro to describe
853the order of words in a register.  The macro @code{WORDS_BIG_ENDIAN} controls
854the order of words in memory.
855@end defmac
856
857@defmac FLOAT_WORDS_BIG_ENDIAN
858Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
859@code{TFmode} floating point numbers are stored in memory with the word
860containing the sign bit at the lowest address; otherwise define it to
861have the value 0.  This macro need not be a constant.
862
863You need not define this macro if the ordering is the same as for
864multi-word integers.
865@end defmac
866
867@defmac BITS_PER_WORD
868Number of bits in a word.  If you do not define this macro, the default
869is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
870@end defmac
871
872@defmac MAX_BITS_PER_WORD
873Maximum number of bits in a word.  If this is undefined, the default is
874@code{BITS_PER_WORD}.  Otherwise, it is the constant value that is the
875largest value that @code{BITS_PER_WORD} can have at run-time.
876@end defmac
877
878@defmac UNITS_PER_WORD
879Number of storage units in a word; normally the size of a general-purpose
880register, a power of two from 1 or 8.
881@end defmac
882
883@defmac MIN_UNITS_PER_WORD
884Minimum number of units in a word.  If this is undefined, the default is
885@code{UNITS_PER_WORD}.  Otherwise, it is the constant value that is the
886smallest value that @code{UNITS_PER_WORD} can have at run-time.
887@end defmac
888
889@defmac POINTER_SIZE
890Width of a pointer, in bits.  You must specify a value no wider than the
891width of @code{Pmode}.  If it is not equal to the width of @code{Pmode},
892you must define @code{POINTERS_EXTEND_UNSIGNED}.  If you do not specify
893a value the default is @code{BITS_PER_WORD}.
894@end defmac
895
896@defmac POINTERS_EXTEND_UNSIGNED
897A C expression that determines how pointers should be extended from
898@code{ptr_mode} to either @code{Pmode} or @code{word_mode}.  It is
899greater than zero if pointers should be zero-extended, zero if they
900should be sign-extended, and negative if some other sort of conversion
901is needed.  In the last case, the extension is done by the target's
902@code{ptr_extend} instruction.
903
904You need not define this macro if the @code{ptr_mode}, @code{Pmode}
905and @code{word_mode} are all the same width.
906@end defmac
907
908@defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
909A macro to update @var{m} and @var{unsignedp} when an object whose type
910is @var{type} and which has the specified mode and signedness is to be
911stored in a register.  This macro is only called when @var{type} is a
912scalar type.
913
914On most RISC machines, which only have operations that operate on a full
915register, define this macro to set @var{m} to @code{word_mode} if
916@var{m} is an integer mode narrower than @code{BITS_PER_WORD}.  In most
917cases, only integer modes should be widened because wider-precision
918floating-point operations are usually more expensive than their narrower
919counterparts.
920
921For most machines, the macro definition does not change @var{unsignedp}.
922However, some machines, have instructions that preferentially handle
923either signed or unsigned quantities of certain modes.  For example, on
924the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
925sign-extend the result to 64 bits.  On such machines, set
926@var{unsignedp} according to which kind of extension is more efficient.
927
928Do not define this macro if it would never modify @var{m}.
929@end defmac
930
931@hook TARGET_C_EXCESS_PRECISION
932Return a value, with the same meaning as the C99 macro
933@code{FLT_EVAL_METHOD} that describes which excess precision should be
934applied.
935
936@hook TARGET_PROMOTE_FUNCTION_MODE
937
938@defmac PARM_BOUNDARY
939Normal alignment required for function parameters on the stack, in
940bits.  All stack parameters receive at least this much alignment
941regardless of data type.  On most machines, this is the same as the
942size of an integer.
943@end defmac
944
945@defmac STACK_BOUNDARY
946Define this macro to the minimum alignment enforced by hardware for the
947stack pointer on this machine.  The definition is a C expression for the
948desired alignment (measured in bits).  This value is used as a default
949if @code{PREFERRED_STACK_BOUNDARY} is not defined.  On most machines,
950this should be the same as @code{PARM_BOUNDARY}.
951@end defmac
952
953@defmac PREFERRED_STACK_BOUNDARY
954Define this macro if you wish to preserve a certain alignment for the
955stack pointer, greater than what the hardware enforces.  The definition
956is a C expression for the desired alignment (measured in bits).  This
957macro must evaluate to a value equal to or larger than
958@code{STACK_BOUNDARY}.
959@end defmac
960
961@defmac INCOMING_STACK_BOUNDARY
962Define this macro if the incoming stack boundary may be different
963from @code{PREFERRED_STACK_BOUNDARY}.  This macro must evaluate
964to a value equal to or larger than @code{STACK_BOUNDARY}.
965@end defmac
966
967@defmac FUNCTION_BOUNDARY
968Alignment required for a function entry point, in bits.
969@end defmac
970
971@defmac BIGGEST_ALIGNMENT
972Biggest alignment that any data type can require on this machine, in
973bits.  Note that this is not the biggest alignment that is supported,
974just the biggest alignment that, when violated, may cause a fault.
975@end defmac
976
977@hook TARGET_ABSOLUTE_BIGGEST_ALIGNMENT
978
979@defmac MALLOC_ABI_ALIGNMENT
980Alignment, in bits, a C conformant malloc implementation has to
981provide.  If not defined, the default value is @code{BITS_PER_WORD}.
982@end defmac
983
984@defmac ATTRIBUTE_ALIGNED_VALUE
985Alignment used by the @code{__attribute__ ((aligned))} construct.  If
986not defined, the default value is @code{BIGGEST_ALIGNMENT}.
987@end defmac
988
989@defmac MINIMUM_ATOMIC_ALIGNMENT
990If defined, the smallest alignment, in bits, that can be given to an
991object that can be referenced in one operation, without disturbing any
992nearby object.  Normally, this is @code{BITS_PER_UNIT}, but may be larger
993on machines that don't have byte or half-word store operations.
994@end defmac
995
996@defmac BIGGEST_FIELD_ALIGNMENT
997Biggest alignment that any structure or union field can require on this
998machine, in bits.  If defined, this overrides @code{BIGGEST_ALIGNMENT} for
999structure and union fields only, unless the field alignment has been set
1000by the @code{__attribute__ ((aligned (@var{n})))} construct.
1001@end defmac
1002
1003@defmac ADJUST_FIELD_ALIGN (@var{field}, @var{type}, @var{computed})
1004An expression for the alignment of a structure field @var{field} of
1005type @var{type} if the alignment computed in the usual way (including
1006applying of @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1007alignment) is @var{computed}.  It overrides alignment only if the
1008field alignment has not been set by the
1009@code{__attribute__ ((aligned (@var{n})))} construct.  Note that @var{field}
1010may be @code{NULL_TREE} in case we just query for the minimum alignment
1011of a field of type @var{type} in structure context.
1012@end defmac
1013
1014@defmac MAX_STACK_ALIGNMENT
1015Biggest stack alignment guaranteed by the backend.  Use this macro
1016to specify the maximum alignment of a variable on stack.
1017
1018If not defined, the default value is @code{STACK_BOUNDARY}.
1019
1020@c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1021@c But the fix for PR 32893 indicates that we can only guarantee
1022@c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1023@c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1024@end defmac
1025
1026@defmac MAX_OFILE_ALIGNMENT
1027Biggest alignment supported by the object file format of this machine.
1028Use this macro to limit the alignment which can be specified using the
1029@code{__attribute__ ((aligned (@var{n})))} construct for functions and
1030objects with static storage duration.  The alignment of automatic
1031objects may exceed the object file format maximum up to the maximum
1032supported by GCC.  If not defined, the default value is
1033@code{BIGGEST_ALIGNMENT}.
1034
1035On systems that use ELF, the default (in @file{config/elfos.h}) is
1036the largest supported 32-bit ELF section alignment representable on
1037a 32-bit host e.g.@: @samp{(((uint64_t) 1 << 28) * 8)}.
1038On 32-bit ELF the largest supported section alignment in bits is
1039@samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1040@end defmac
1041
1042@hook TARGET_LOWER_LOCAL_DECL_ALIGNMENT
1043
1044@hook TARGET_STATIC_RTX_ALIGNMENT
1045
1046@defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1047If defined, a C expression to compute the alignment for a variable in
1048the static store.  @var{type} is the data type, and @var{basic-align} is
1049the alignment that the object would ordinarily have.  The value of this
1050macro is used instead of that alignment to align the object.
1051
1052If this macro is not defined, then @var{basic-align} is used.
1053
1054@findex strcpy
1055One use of this macro is to increase alignment of medium-size data to
1056make it all fit in fewer cache lines.  Another is to cause character
1057arrays to be word-aligned so that @code{strcpy} calls that copy
1058constants to character arrays can be done inline.
1059@end defmac
1060
1061@defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align})
1062Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates
1063some alignment increase, instead of optimization only purposes.  E.g.@
1064AMD x86-64 psABI says that variables with array type larger than 15 bytes
1065must be aligned to 16 byte boundaries.
1066
1067If this macro is not defined, then @var{basic-align} is used.
1068@end defmac
1069
1070@hook TARGET_CONSTANT_ALIGNMENT
1071
1072@defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1073If defined, a C expression to compute the alignment for a variable in
1074the local store.  @var{type} is the data type, and @var{basic-align} is
1075the alignment that the object would ordinarily have.  The value of this
1076macro is used instead of that alignment to align the object.
1077
1078If this macro is not defined, then @var{basic-align} is used.
1079
1080One use of this macro is to increase alignment of medium-size data to
1081make it all fit in fewer cache lines.
1082
1083If the value of this macro has a type, it should be an unsigned type.
1084@end defmac
1085
1086@hook TARGET_VECTOR_ALIGNMENT
1087
1088@defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1089If defined, a C expression to compute the alignment for stack slot.
1090@var{type} is the data type, @var{mode} is the widest mode available,
1091and @var{basic-align} is the alignment that the slot would ordinarily
1092have.  The value of this macro is used instead of that alignment to
1093align the slot.
1094
1095If this macro is not defined, then @var{basic-align} is used when
1096@var{type} is @code{NULL}.  Otherwise, @code{LOCAL_ALIGNMENT} will
1097be used.
1098
1099This macro is to set alignment of stack slot to the maximum alignment
1100of all possible modes which the slot may have.
1101
1102If the value of this macro has a type, it should be an unsigned type.
1103@end defmac
1104
1105@defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1106If defined, a C expression to compute the alignment for a local
1107variable @var{decl}.
1108
1109If this macro is not defined, then
1110@code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1111is used.
1112
1113One use of this macro is to increase alignment of medium-size data to
1114make it all fit in fewer cache lines.
1115
1116If the value of this macro has a type, it should be an unsigned type.
1117@end defmac
1118
1119@defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1120If defined, a C expression to compute the minimum required alignment
1121for dynamic stack realignment purposes for @var{exp} (a type or decl),
1122@var{mode}, assuming normal alignment @var{align}.
1123
1124If this macro is not defined, then @var{align} will be used.
1125@end defmac
1126
1127@defmac EMPTY_FIELD_BOUNDARY
1128Alignment in bits to be given to a structure bit-field that follows an
1129empty field such as @code{int : 0;}.
1130
1131If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1132@end defmac
1133
1134@defmac STRUCTURE_SIZE_BOUNDARY
1135Number of bits which any structure or union's size must be a multiple of.
1136Each structure or union's size is rounded up to a multiple of this.
1137
1138If you do not define this macro, the default is the same as
1139@code{BITS_PER_UNIT}.
1140@end defmac
1141
1142@defmac STRICT_ALIGNMENT
1143Define this macro to be the value 1 if instructions will fail to work
1144if given data not on the nominal alignment.  If instructions will merely
1145go slower in that case, define this macro as 0.
1146@end defmac
1147
1148@defmac PCC_BITFIELD_TYPE_MATTERS
1149Define this if you wish to imitate the way many other C compilers handle
1150alignment of bit-fields and the structures that contain them.
1151
1152The behavior is that the type written for a named bit-field (@code{int},
1153@code{short}, or other integer type) imposes an alignment for the entire
1154structure, as if the structure really did contain an ordinary field of
1155that type.  In addition, the bit-field is placed within the structure so
1156that it would fit within such a field, not crossing a boundary for it.
1157
1158Thus, on most machines, a named bit-field whose type is written as
1159@code{int} would not cross a four-byte boundary, and would force
1160four-byte alignment for the whole structure.  (The alignment used may
1161not be four bytes; it is controlled by the other alignment parameters.)
1162
1163An unnamed bit-field will not affect the alignment of the containing
1164structure.
1165
1166If the macro is defined, its definition should be a C expression;
1167a nonzero value for the expression enables this behavior.
1168
1169Note that if this macro is not defined, or its value is zero, some
1170bit-fields may cross more than one alignment boundary.  The compiler can
1171support such references if there are @samp{insv}, @samp{extv}, and
1172@samp{extzv} insns that can directly reference memory.
1173
1174The other known way of making bit-fields work is to define
1175@code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1176Then every structure can be accessed with fullwords.
1177
1178Unless the machine has bit-field instructions or you define
1179@code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1180@code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1181
1182If your aim is to make GCC use the same conventions for laying out
1183bit-fields as are used by another compiler, here is how to investigate
1184what the other compiler does.  Compile and run this program:
1185
1186@smallexample
1187struct foo1
1188@{
1189  char x;
1190  char :0;
1191  char y;
1192@};
1193
1194struct foo2
1195@{
1196  char x;
1197  int :0;
1198  char y;
1199@};
1200
1201main ()
1202@{
1203  printf ("Size of foo1 is %d\n",
1204          sizeof (struct foo1));
1205  printf ("Size of foo2 is %d\n",
1206          sizeof (struct foo2));
1207  exit (0);
1208@}
1209@end smallexample
1210
1211If this prints 2 and 5, then the compiler's behavior is what you would
1212get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1213@end defmac
1214
1215@defmac BITFIELD_NBYTES_LIMITED
1216Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1217to aligning a bit-field within the structure.
1218@end defmac
1219
1220@hook TARGET_ALIGN_ANON_BITFIELD
1221
1222@hook TARGET_NARROW_VOLATILE_BITFIELD
1223
1224@hook TARGET_MEMBER_TYPE_FORCES_BLK
1225
1226@defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1227Define this macro as an expression for the alignment of a type (given
1228by @var{type} as a tree node) if the alignment computed in the usual
1229way is @var{computed} and the alignment explicitly specified was
1230@var{specified}.
1231
1232The default is to use @var{specified} if it is larger; otherwise, use
1233the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1234@end defmac
1235
1236@defmac MAX_FIXED_MODE_SIZE
1237An integer expression for the size in bits of the largest integer
1238machine mode that should actually be used.  All integer machine modes of
1239this size or smaller can be used for structures and unions with the
1240appropriate sizes.  If this macro is undefined, @code{GET_MODE_BITSIZE
1241(DImode)} is assumed.
1242@end defmac
1243
1244@defmac STACK_SAVEAREA_MODE (@var{save_level})
1245If defined, an expression of type @code{machine_mode} that
1246specifies the mode of the save area operand of a
1247@code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1248@var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1249@code{SAVE_NONLOCAL} and selects which of the three named patterns is
1250having its mode specified.
1251
1252You need not define this macro if it always returns @code{Pmode}.  You
1253would most commonly define this macro if the
1254@code{save_stack_@var{level}} patterns need to support both a 32- and a
125564-bit mode.
1256@end defmac
1257
1258@defmac STACK_SIZE_MODE
1259If defined, an expression of type @code{machine_mode} that
1260specifies the mode of the size increment operand of an
1261@code{allocate_stack} named pattern (@pxref{Standard Names}).
1262
1263You need not define this macro if it always returns @code{word_mode}.
1264You would most commonly define this macro if the @code{allocate_stack}
1265pattern needs to support both a 32- and a 64-bit mode.
1266@end defmac
1267
1268@hook TARGET_LIBGCC_CMP_RETURN_MODE
1269
1270@hook TARGET_LIBGCC_SHIFT_COUNT_MODE
1271
1272@hook TARGET_UNWIND_WORD_MODE
1273
1274@hook TARGET_MS_BITFIELD_LAYOUT_P
1275
1276@hook TARGET_DECIMAL_FLOAT_SUPPORTED_P
1277
1278@hook TARGET_FIXED_POINT_SUPPORTED_P
1279
1280@hook TARGET_EXPAND_TO_RTL_HOOK
1281
1282@hook TARGET_INSTANTIATE_DECLS
1283
1284@hook TARGET_MANGLE_TYPE
1285
1286@node Type Layout
1287@section Layout of Source Language Data Types
1288
1289These macros define the sizes and other characteristics of the standard
1290basic data types used in programs being compiled.  Unlike the macros in
1291the previous section, these apply to specific features of C and related
1292languages, rather than to fundamental aspects of storage layout.
1293
1294@defmac INT_TYPE_SIZE
1295A C expression for the size in bits of the type @code{int} on the
1296target machine.  If you don't define this, the default is one word.
1297@end defmac
1298
1299@defmac SHORT_TYPE_SIZE
1300A C expression for the size in bits of the type @code{short} on the
1301target machine.  If you don't define this, the default is half a word.
1302(If this would be less than one storage unit, it is rounded up to one
1303unit.)
1304@end defmac
1305
1306@defmac LONG_TYPE_SIZE
1307A C expression for the size in bits of the type @code{long} on the
1308target machine.  If you don't define this, the default is one word.
1309@end defmac
1310
1311@defmac ADA_LONG_TYPE_SIZE
1312On some machines, the size used for the Ada equivalent of the type
1313@code{long} by a native Ada compiler differs from that used by C@.  In
1314that situation, define this macro to be a C expression to be used for
1315the size of that type.  If you don't define this, the default is the
1316value of @code{LONG_TYPE_SIZE}.
1317@end defmac
1318
1319@defmac LONG_LONG_TYPE_SIZE
1320A C expression for the size in bits of the type @code{long long} on the
1321target machine.  If you don't define this, the default is two
1322words.  If you want to support GNU Ada on your machine, the value of this
1323macro must be at least 64.
1324@end defmac
1325
1326@defmac CHAR_TYPE_SIZE
1327A C expression for the size in bits of the type @code{char} on the
1328target machine.  If you don't define this, the default is
1329@code{BITS_PER_UNIT}.
1330@end defmac
1331
1332@defmac BOOL_TYPE_SIZE
1333A C expression for the size in bits of the C++ type @code{bool} and
1334C99 type @code{_Bool} on the target machine.  If you don't define
1335this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1336@end defmac
1337
1338@defmac FLOAT_TYPE_SIZE
1339A C expression for the size in bits of the type @code{float} on the
1340target machine.  If you don't define this, the default is one word.
1341@end defmac
1342
1343@defmac DOUBLE_TYPE_SIZE
1344A C expression for the size in bits of the type @code{double} on the
1345target machine.  If you don't define this, the default is two
1346words.
1347@end defmac
1348
1349@defmac LONG_DOUBLE_TYPE_SIZE
1350A C expression for the size in bits of the type @code{long double} on
1351the target machine.  If you don't define this, the default is two
1352words.
1353@end defmac
1354
1355@defmac SHORT_FRACT_TYPE_SIZE
1356A C expression for the size in bits of the type @code{short _Fract} on
1357the target machine.  If you don't define this, the default is
1358@code{BITS_PER_UNIT}.
1359@end defmac
1360
1361@defmac FRACT_TYPE_SIZE
1362A C expression for the size in bits of the type @code{_Fract} on
1363the target machine.  If you don't define this, the default is
1364@code{BITS_PER_UNIT * 2}.
1365@end defmac
1366
1367@defmac LONG_FRACT_TYPE_SIZE
1368A C expression for the size in bits of the type @code{long _Fract} on
1369the target machine.  If you don't define this, the default is
1370@code{BITS_PER_UNIT * 4}.
1371@end defmac
1372
1373@defmac LONG_LONG_FRACT_TYPE_SIZE
1374A C expression for the size in bits of the type @code{long long _Fract} on
1375the target machine.  If you don't define this, the default is
1376@code{BITS_PER_UNIT * 8}.
1377@end defmac
1378
1379@defmac SHORT_ACCUM_TYPE_SIZE
1380A C expression for the size in bits of the type @code{short _Accum} on
1381the target machine.  If you don't define this, the default is
1382@code{BITS_PER_UNIT * 2}.
1383@end defmac
1384
1385@defmac ACCUM_TYPE_SIZE
1386A C expression for the size in bits of the type @code{_Accum} on
1387the target machine.  If you don't define this, the default is
1388@code{BITS_PER_UNIT * 4}.
1389@end defmac
1390
1391@defmac LONG_ACCUM_TYPE_SIZE
1392A C expression for the size in bits of the type @code{long _Accum} on
1393the target machine.  If you don't define this, the default is
1394@code{BITS_PER_UNIT * 8}.
1395@end defmac
1396
1397@defmac LONG_LONG_ACCUM_TYPE_SIZE
1398A C expression for the size in bits of the type @code{long long _Accum} on
1399the target machine.  If you don't define this, the default is
1400@code{BITS_PER_UNIT * 16}.
1401@end defmac
1402
1403@defmac LIBGCC2_GNU_PREFIX
1404This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target
1405hook and should be defined if that hook is overriden to be true.  It
1406causes function names in libgcc to be changed to use a @code{__gnu_}
1407prefix for their name rather than the default @code{__}.  A port which
1408uses this macro should also arrange to use @file{t-gnu-prefix} in
1409the libgcc @file{config.host}.
1410@end defmac
1411
1412@defmac WIDEST_HARDWARE_FP_SIZE
1413A C expression for the size in bits of the widest floating-point format
1414supported by the hardware.  If you define this macro, you must specify a
1415value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1416If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1417is the default.
1418@end defmac
1419
1420@defmac DEFAULT_SIGNED_CHAR
1421An expression whose value is 1 or 0, according to whether the type
1422@code{char} should be signed or unsigned by default.  The user can
1423always override this default with the options @option{-fsigned-char}
1424and @option{-funsigned-char}.
1425@end defmac
1426
1427@hook TARGET_DEFAULT_SHORT_ENUMS
1428
1429@defmac SIZE_TYPE
1430A C expression for a string describing the name of the data type to use
1431for size values.  The typedef name @code{size_t} is defined using the
1432contents of the string.
1433
1434The string can contain more than one keyword.  If so, separate them with
1435spaces, and write first any length keyword, then @code{unsigned} if
1436appropriate, and finally @code{int}.  The string must exactly match one
1437of the data type names defined in the function
1438@code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.cc}.
1439You may not omit @code{int} or change the order---that would cause the
1440compiler to crash on startup.
1441
1442If you don't define this macro, the default is @code{"long unsigned
1443int"}.
1444@end defmac
1445
1446@defmac SIZETYPE
1447GCC defines internal types (@code{sizetype}, @code{ssizetype},
1448@code{bitsizetype} and @code{sbitsizetype}) for expressions
1449dealing with size.  This macro is a C expression for a string describing
1450the name of the data type from which the precision of @code{sizetype}
1451is extracted.
1452
1453The string has the same restrictions as @code{SIZE_TYPE} string.
1454
1455If you don't define this macro, the default is @code{SIZE_TYPE}.
1456@end defmac
1457
1458@defmac PTRDIFF_TYPE
1459A C expression for a string describing the name of the data type to use
1460for the result of subtracting two pointers.  The typedef name
1461@code{ptrdiff_t} is defined using the contents of the string.  See
1462@code{SIZE_TYPE} above for more information.
1463
1464If you don't define this macro, the default is @code{"long int"}.
1465@end defmac
1466
1467@defmac WCHAR_TYPE
1468A C expression for a string describing the name of the data type to use
1469for wide characters.  The typedef name @code{wchar_t} is defined using
1470the contents of the string.  See @code{SIZE_TYPE} above for more
1471information.
1472
1473If you don't define this macro, the default is @code{"int"}.
1474@end defmac
1475
1476@defmac WCHAR_TYPE_SIZE
1477A C expression for the size in bits of the data type for wide
1478characters.  This is used in @code{cpp}, which cannot make use of
1479@code{WCHAR_TYPE}.
1480@end defmac
1481
1482@defmac WINT_TYPE
1483A C expression for a string describing the name of the data type to
1484use for wide characters passed to @code{printf} and returned from
1485@code{getwc}.  The typedef name @code{wint_t} is defined using the
1486contents of the string.  See @code{SIZE_TYPE} above for more
1487information.
1488
1489If you don't define this macro, the default is @code{"unsigned int"}.
1490@end defmac
1491
1492@defmac INTMAX_TYPE
1493A C expression for a string describing the name of the data type that
1494can represent any value of any standard or extended signed integer type.
1495The typedef name @code{intmax_t} is defined using the contents of the
1496string.  See @code{SIZE_TYPE} above for more information.
1497
1498If you don't define this macro, the default is the first of
1499@code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1500much precision as @code{long long int}.
1501@end defmac
1502
1503@defmac UINTMAX_TYPE
1504A C expression for a string describing the name of the data type that
1505can represent any value of any standard or extended unsigned integer
1506type.  The typedef name @code{uintmax_t} is defined using the contents
1507of the string.  See @code{SIZE_TYPE} above for more information.
1508
1509If you don't define this macro, the default is the first of
1510@code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1511unsigned int"} that has as much precision as @code{long long unsigned
1512int}.
1513@end defmac
1514
1515@defmac SIG_ATOMIC_TYPE
1516@defmacx INT8_TYPE
1517@defmacx INT16_TYPE
1518@defmacx INT32_TYPE
1519@defmacx INT64_TYPE
1520@defmacx UINT8_TYPE
1521@defmacx UINT16_TYPE
1522@defmacx UINT32_TYPE
1523@defmacx UINT64_TYPE
1524@defmacx INT_LEAST8_TYPE
1525@defmacx INT_LEAST16_TYPE
1526@defmacx INT_LEAST32_TYPE
1527@defmacx INT_LEAST64_TYPE
1528@defmacx UINT_LEAST8_TYPE
1529@defmacx UINT_LEAST16_TYPE
1530@defmacx UINT_LEAST32_TYPE
1531@defmacx UINT_LEAST64_TYPE
1532@defmacx INT_FAST8_TYPE
1533@defmacx INT_FAST16_TYPE
1534@defmacx INT_FAST32_TYPE
1535@defmacx INT_FAST64_TYPE
1536@defmacx UINT_FAST8_TYPE
1537@defmacx UINT_FAST16_TYPE
1538@defmacx UINT_FAST32_TYPE
1539@defmacx UINT_FAST64_TYPE
1540@defmacx INTPTR_TYPE
1541@defmacx UINTPTR_TYPE
1542C expressions for the standard types @code{sig_atomic_t},
1543@code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1544@code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1545@code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1546@code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1547@code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1548@code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1549@code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1550@code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}.  See
1551@code{SIZE_TYPE} above for more information.
1552
1553If any of these macros evaluates to a null pointer, the corresponding
1554type is not supported; if GCC is configured to provide
1555@code{<stdint.h>} in such a case, the header provided may not conform
1556to C99, depending on the type in question.  The defaults for all of
1557these macros are null pointers.
1558@end defmac
1559
1560@defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1561The C++ compiler represents a pointer-to-member-function with a struct
1562that looks like:
1563
1564@smallexample
1565  struct @{
1566    union @{
1567      void (*fn)();
1568      ptrdiff_t vtable_index;
1569    @};
1570    ptrdiff_t delta;
1571  @};
1572@end smallexample
1573
1574@noindent
1575The C++ compiler must use one bit to indicate whether the function that
1576will be called through a pointer-to-member-function is virtual.
1577Normally, we assume that the low-order bit of a function pointer must
1578always be zero.  Then, by ensuring that the vtable_index is odd, we can
1579distinguish which variant of the union is in use.  But, on some
1580platforms function pointers can be odd, and so this doesn't work.  In
1581that case, we use the low-order bit of the @code{delta} field, and shift
1582the remainder of the @code{delta} field to the left.
1583
1584GCC will automatically make the right selection about where to store
1585this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1586However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1587set such that functions always start at even addresses, but the lowest
1588bit of pointers to functions indicate whether the function at that
1589address is in ARM or Thumb mode.  If this is the case of your
1590architecture, you should define this macro to
1591@code{ptrmemfunc_vbit_in_delta}.
1592
1593In general, you should not have to define this macro.  On architectures
1594in which function addresses are always even, according to
1595@code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1596@code{ptrmemfunc_vbit_in_pfn}.
1597@end defmac
1598
1599@defmac TARGET_VTABLE_USES_DESCRIPTORS
1600Normally, the C++ compiler uses function pointers in vtables.  This
1601macro allows the target to change to use ``function descriptors''
1602instead.  Function descriptors are found on targets for whom a
1603function pointer is actually a small data structure.  Normally the
1604data structure consists of the actual code address plus a data
1605pointer to which the function's data is relative.
1606
1607If vtables are used, the value of this macro should be the number
1608of words that the function descriptor occupies.
1609@end defmac
1610
1611@defmac TARGET_VTABLE_ENTRY_ALIGN
1612By default, the vtable entries are void pointers, the so the alignment
1613is the same as pointer alignment.  The value of this macro specifies
1614the alignment of the vtable entry in bits.  It should be defined only
1615when special alignment is necessary. */
1616@end defmac
1617
1618@defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1619There are a few non-descriptor entries in the vtable at offsets below
1620zero.  If these entries must be padded (say, to preserve the alignment
1621specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1622of words in each data entry.
1623@end defmac
1624
1625@node Registers
1626@section Register Usage
1627@cindex register usage
1628
1629This section explains how to describe what registers the target machine
1630has, and how (in general) they can be used.
1631
1632The description of which registers a specific instruction can use is
1633done with register classes; see @ref{Register Classes}.  For information
1634on using registers to access a stack frame, see @ref{Frame Registers}.
1635For passing values in registers, see @ref{Register Arguments}.
1636For returning values in registers, see @ref{Scalar Return}.
1637
1638@menu
1639* Register Basics::             Number and kinds of registers.
1640* Allocation Order::            Order in which registers are allocated.
1641* Values in Registers::         What kinds of values each reg can hold.
1642* Leaf Functions::              Renumbering registers for leaf functions.
1643* Stack Registers::             Handling a register stack such as 80387.
1644@end menu
1645
1646@node Register Basics
1647@subsection Basic Characteristics of Registers
1648
1649@c prevent bad page break with this line
1650Registers have various characteristics.
1651
1652@defmac FIRST_PSEUDO_REGISTER
1653Number of hardware registers known to the compiler.  They receive
1654numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1655pseudo register's number really is assigned the number
1656@code{FIRST_PSEUDO_REGISTER}.
1657@end defmac
1658
1659@defmac FIXED_REGISTERS
1660@cindex fixed register
1661An initializer that says which registers are used for fixed purposes
1662all throughout the compiled code and are therefore not available for
1663general allocation.  These would include the stack pointer, the frame
1664pointer (except on machines where that can be used as a general
1665register when no frame pointer is needed), the program counter on
1666machines where that is considered one of the addressable registers,
1667and any other numbered register with a standard use.
1668
1669This information is expressed as a sequence of numbers, separated by
1670commas and surrounded by braces.  The @var{n}th number is 1 if
1671register @var{n} is fixed, 0 otherwise.
1672
1673The table initialized from this macro, and the table initialized by
1674the following one, may be overridden at run time either automatically,
1675by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1676the user with the command options @option{-ffixed-@var{reg}},
1677@option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1678@end defmac
1679
1680@defmac CALL_USED_REGISTERS
1681@cindex call-used register
1682@cindex call-clobbered register
1683@cindex call-saved register
1684Like @code{FIXED_REGISTERS} but has 1 for each register that is
1685clobbered (in general) by function calls as well as for fixed
1686registers.  This macro therefore identifies the registers that are not
1687available for general allocation of values that must live across
1688function calls.
1689
1690If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1691automatically saves it on function entry and restores it on function
1692exit, if the register is used within the function.
1693
1694Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS}
1695must be defined.  Modern ports should define @code{CALL_REALLY_USED_REGISTERS}.
1696@end defmac
1697
1698@defmac CALL_REALLY_USED_REGISTERS
1699@cindex call-used register
1700@cindex call-clobbered register
1701@cindex call-saved register
1702Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1703that the entire set of @code{FIXED_REGISTERS} be included.
1704(@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1705
1706Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS}
1707must be defined.  Modern ports should define @code{CALL_REALLY_USED_REGISTERS}.
1708@end defmac
1709
1710@cindex call-used register
1711@cindex call-clobbered register
1712@cindex call-saved register
1713@hook TARGET_FNTYPE_ABI
1714
1715@hook TARGET_INSN_CALLEE_ABI
1716
1717@cindex call-used register
1718@cindex call-clobbered register
1719@cindex call-saved register
1720@hook TARGET_HARD_REGNO_CALL_PART_CLOBBERED
1721
1722@hook TARGET_GET_MULTILIB_ABI_NAME
1723
1724@findex fixed_regs
1725@findex call_used_regs
1726@findex global_regs
1727@findex reg_names
1728@findex reg_class_contents
1729@hook TARGET_CONDITIONAL_REGISTER_USAGE
1730
1731@defmac INCOMING_REGNO (@var{out})
1732Define this macro if the target machine has register windows.  This C
1733expression returns the register number as seen by the called function
1734corresponding to the register number @var{out} as seen by the calling
1735function.  Return @var{out} if register number @var{out} is not an
1736outbound register.
1737@end defmac
1738
1739@defmac OUTGOING_REGNO (@var{in})
1740Define this macro if the target machine has register windows.  This C
1741expression returns the register number as seen by the calling function
1742corresponding to the register number @var{in} as seen by the called
1743function.  Return @var{in} if register number @var{in} is not an inbound
1744register.
1745@end defmac
1746
1747@defmac LOCAL_REGNO (@var{regno})
1748Define this macro if the target machine has register windows.  This C
1749expression returns true if the register is call-saved but is in the
1750register window.  Unlike most call-saved registers, such registers
1751need not be explicitly restored on function exit or during non-local
1752gotos.
1753@end defmac
1754
1755@defmac PC_REGNUM
1756If the program counter has a register number, define this as that
1757register number.  Otherwise, do not define it.
1758@end defmac
1759
1760@node Allocation Order
1761@subsection Order of Allocation of Registers
1762@cindex order of register allocation
1763@cindex register allocation order
1764
1765@c prevent bad page break with this line
1766Registers are allocated in order.
1767
1768@defmac REG_ALLOC_ORDER
1769If defined, an initializer for a vector of integers, containing the
1770numbers of hard registers in the order in which GCC should prefer
1771to use them (from most preferred to least).
1772
1773If this macro is not defined, registers are used lowest numbered first
1774(all else being equal).
1775
1776One use of this macro is on machines where the highest numbered
1777registers must always be saved and the save-multiple-registers
1778instruction supports only sequences of consecutive registers.  On such
1779machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
1780the highest numbered allocable register first.
1781@end defmac
1782
1783@defmac ADJUST_REG_ALLOC_ORDER
1784A C statement (sans semicolon) to choose the order in which to allocate
1785hard registers for pseudo-registers local to a basic block.
1786
1787Store the desired register order in the array @code{reg_alloc_order}.
1788Element 0 should be the register to allocate first; element 1, the next
1789register; and so on.
1790
1791The macro body should not assume anything about the contents of
1792@code{reg_alloc_order} before execution of the macro.
1793
1794On most machines, it is not necessary to define this macro.
1795@end defmac
1796
1797@defmac HONOR_REG_ALLOC_ORDER
1798Normally, IRA tries to estimate the costs for saving a register in the
1799prologue and restoring it in the epilogue.  This discourages it from
1800using call-saved registers.  If a machine wants to ensure that IRA
1801allocates registers in the order given by REG_ALLOC_ORDER even if some
1802call-saved registers appear earlier than call-used ones, then define this
1803macro as a C expression to nonzero. Default is 0.
1804@end defmac
1805
1806@defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
1807In some case register allocation order is not enough for the
1808Integrated Register Allocator (@acronym{IRA}) to generate a good code.
1809If this macro is defined, it should return a floating point value
1810based on @var{regno}.  The cost of using @var{regno} for a pseudo will
1811be increased by approximately the pseudo's usage frequency times the
1812value returned by this macro.  Not defining this macro is equivalent
1813to having it always return @code{0.0}.
1814
1815On most machines, it is not necessary to define this macro.
1816@end defmac
1817
1818@node Values in Registers
1819@subsection How Values Fit in Registers
1820
1821This section discusses the macros that describe which kinds of values
1822(specifically, which machine modes) each register can hold, and how many
1823consecutive registers are needed for a given mode.
1824
1825@hook TARGET_HARD_REGNO_NREGS
1826
1827@defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
1828A C expression that is nonzero if a value of mode @var{mode}, stored
1829in memory, ends with padding that causes it to take up more space than
1830in registers starting at register number @var{regno} (as determined by
1831multiplying GCC's notion of the size of the register when containing
1832this mode by the number of registers returned by
1833@code{TARGET_HARD_REGNO_NREGS}).  By default this is zero.
1834
1835For example, if a floating-point value is stored in three 32-bit
1836registers but takes up 128 bits in memory, then this would be
1837nonzero.
1838
1839This macros only needs to be defined if there are cases where
1840@code{subreg_get_info}
1841would otherwise wrongly determine that a @code{subreg} can be
1842represented by an offset to the register number, when in fact such a
1843@code{subreg} would contain some of the padding not stored in
1844registers and so not be representable.
1845@end defmac
1846
1847@defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
1848For values of @var{regno} and @var{mode} for which
1849@code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
1850returning the greater number of registers required to hold the value
1851including any padding.  In the example above, the value would be four.
1852@end defmac
1853
1854@defmac REGMODE_NATURAL_SIZE (@var{mode})
1855Define this macro if the natural size of registers that hold values
1856of mode @var{mode} is not the word size.  It is a C expression that
1857should give the natural size in bytes for the specified mode.  It is
1858used by the register allocator to try to optimize its results.  This
1859happens for example on SPARC 64-bit where the natural size of
1860floating-point registers is still 32-bit.
1861@end defmac
1862
1863@hook TARGET_HARD_REGNO_MODE_OK
1864
1865@defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
1866A C expression that is nonzero if it is OK to rename a hard register
1867@var{from} to another hard register @var{to}.
1868
1869One common use of this macro is to prevent renaming of a register to
1870another register that is not saved by a prologue in an interrupt
1871handler.
1872
1873The default is always nonzero.
1874@end defmac
1875
1876@hook TARGET_MODES_TIEABLE_P
1877
1878@hook TARGET_HARD_REGNO_SCRATCH_OK
1879
1880@defmac AVOID_CCMODE_COPIES
1881Define this macro if the compiler should avoid copies to/from @code{CCmode}
1882registers.  You should only define this macro if support for copying to/from
1883@code{CCmode} is incomplete.
1884@end defmac
1885
1886@node Leaf Functions
1887@subsection Handling Leaf Functions
1888
1889@cindex leaf functions
1890@cindex functions, leaf
1891On some machines, a leaf function (i.e., one which makes no calls) can run
1892more efficiently if it does not make its own register window.  Often this
1893means it is required to receive its arguments in the registers where they
1894are passed by the caller, instead of the registers where they would
1895normally arrive.
1896
1897The special treatment for leaf functions generally applies only when
1898other conditions are met; for example, often they may use only those
1899registers for its own variables and temporaries.  We use the term ``leaf
1900function'' to mean a function that is suitable for this special
1901handling, so that functions with no calls are not necessarily ``leaf
1902functions''.
1903
1904GCC assigns register numbers before it knows whether the function is
1905suitable for leaf function treatment.  So it needs to renumber the
1906registers in order to output a leaf function.  The following macros
1907accomplish this.
1908
1909@defmac LEAF_REGISTERS
1910Name of a char vector, indexed by hard register number, which
1911contains 1 for a register that is allowable in a candidate for leaf
1912function treatment.
1913
1914If leaf function treatment involves renumbering the registers, then the
1915registers marked here should be the ones before renumbering---those that
1916GCC would ordinarily allocate.  The registers which will actually be
1917used in the assembler code, after renumbering, should not be marked with 1
1918in this vector.
1919
1920Define this macro only if the target machine offers a way to optimize
1921the treatment of leaf functions.
1922@end defmac
1923
1924@defmac LEAF_REG_REMAP (@var{regno})
1925A C expression whose value is the register number to which @var{regno}
1926should be renumbered, when a function is treated as a leaf function.
1927
1928If @var{regno} is a register number which should not appear in a leaf
1929function before renumbering, then the expression should yield @minus{}1, which
1930will cause the compiler to abort.
1931
1932Define this macro only if the target machine offers a way to optimize the
1933treatment of leaf functions, and registers need to be renumbered to do
1934this.
1935@end defmac
1936
1937@findex current_function_is_leaf
1938@findex current_function_uses_only_leaf_regs
1939@code{TARGET_ASM_FUNCTION_PROLOGUE} and
1940@code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
1941specially.  They can test the C variable @code{current_function_is_leaf}
1942which is nonzero for leaf functions.  @code{current_function_is_leaf} is
1943set prior to local register allocation and is valid for the remaining
1944compiler passes.  They can also test the C variable
1945@code{current_function_uses_only_leaf_regs} which is nonzero for leaf
1946functions which only use leaf registers.
1947@code{current_function_uses_only_leaf_regs} is valid after all passes
1948that modify the instructions have been run and is only useful if
1949@code{LEAF_REGISTERS} is defined.
1950@c changed this to fix overfull.  ALSO:  why the "it" at the beginning
1951@c of the next paragraph?!  --mew 2feb93
1952
1953@node Stack Registers
1954@subsection Registers That Form a Stack
1955
1956There are special features to handle computers where some of the
1957``registers'' form a stack.  Stack registers are normally written by
1958pushing onto the stack, and are numbered relative to the top of the
1959stack.
1960
1961Currently, GCC can only handle one group of stack-like registers, and
1962they must be consecutively numbered.  Furthermore, the existing
1963support for stack-like registers is specific to the 80387 floating
1964point coprocessor.  If you have a new architecture that uses
1965stack-like registers, you will need to do substantial work on
1966@file{reg-stack.cc} and write your machine description to cooperate
1967with it, as well as defining these macros.
1968
1969@defmac STACK_REGS
1970Define this if the machine has any stack-like registers.
1971@end defmac
1972
1973@defmac STACK_REG_COVER_CLASS
1974This is a cover class containing the stack registers.  Define this if
1975the machine has any stack-like registers.
1976@end defmac
1977
1978@defmac FIRST_STACK_REG
1979The number of the first stack-like register.  This one is the top
1980of the stack.
1981@end defmac
1982
1983@defmac LAST_STACK_REG
1984The number of the last stack-like register.  This one is the bottom of
1985the stack.
1986@end defmac
1987
1988@node Register Classes
1989@section Register Classes
1990@cindex register class definitions
1991@cindex class definitions, register
1992
1993On many machines, the numbered registers are not all equivalent.
1994For example, certain registers may not be allowed for indexed addressing;
1995certain registers may not be allowed in some instructions.  These machine
1996restrictions are described to the compiler using @dfn{register classes}.
1997
1998You define a number of register classes, giving each one a name and saying
1999which of the registers belong to it.  Then you can specify register classes
2000that are allowed as operands to particular instruction patterns.
2001
2002@findex ALL_REGS
2003@findex NO_REGS
2004In general, each register will belong to several classes.  In fact, one
2005class must be named @code{ALL_REGS} and contain all the registers.  Another
2006class must be named @code{NO_REGS} and contain no registers.  Often the
2007union of two classes will be another class; however, this is not required.
2008
2009@findex GENERAL_REGS
2010One of the classes must be named @code{GENERAL_REGS}.  There is nothing
2011terribly special about the name, but the operand constraint letters
2012@samp{r} and @samp{g} specify this class.  If @code{GENERAL_REGS} is
2013the same as @code{ALL_REGS}, just define it as a macro which expands
2014to @code{ALL_REGS}.
2015
2016Order the classes so that if class @var{x} is contained in class @var{y}
2017then @var{x} has a lower class number than @var{y}.
2018
2019The way classes other than @code{GENERAL_REGS} are specified in operand
2020constraints is through machine-dependent operand constraint letters.
2021You can define such letters to correspond to various classes, then use
2022them in operand constraints.
2023
2024You must define the narrowest register classes for allocatable
2025registers, so that each class either has no subclasses, or that for
2026some mode, the move cost between registers within the class is
2027cheaper than moving a register in the class to or from memory
2028(@pxref{Costs}).
2029
2030You should define a class for the union of two classes whenever some
2031instruction allows both classes.  For example, if an instruction allows
2032either a floating point (coprocessor) register or a general register for a
2033certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2034which includes both of them.  Otherwise you will get suboptimal code,
2035or even internal compiler errors when reload cannot find a register in the
2036class computed via @code{reg_class_subunion}.
2037
2038You must also specify certain redundant information about the register
2039classes: for each class, which classes contain it and which ones are
2040contained in it; for each pair of classes, the largest class contained
2041in their union.
2042
2043When a value occupying several consecutive registers is expected in a
2044certain class, all the registers used must belong to that class.
2045Therefore, register classes cannot be used to enforce a requirement for
2046a register pair to start with an even-numbered register.  The way to
2047specify this requirement is with @code{TARGET_HARD_REGNO_MODE_OK}.
2048
2049Register classes used for input-operands of bitwise-and or shift
2050instructions have a special requirement: each such class must have, for
2051each fixed-point machine mode, a subclass whose registers can transfer that
2052mode to or from memory.  For example, on some machines, the operations for
2053single-byte values (@code{QImode}) are limited to certain registers.  When
2054this is so, each register class that is used in a bitwise-and or shift
2055instruction must have a subclass consisting of registers from which
2056single-byte values can be loaded or stored.  This is so that
2057@code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2058
2059@deftp {Data type} {enum reg_class}
2060An enumerated type that must be defined with all the register class names
2061as enumerated values.  @code{NO_REGS} must be first.  @code{ALL_REGS}
2062must be the last register class, followed by one more enumerated value,
2063@code{LIM_REG_CLASSES}, which is not a register class but rather
2064tells how many classes there are.
2065
2066Each register class has a number, which is the value of casting
2067the class name to type @code{int}.  The number serves as an index
2068in many of the tables described below.
2069@end deftp
2070
2071@defmac N_REG_CLASSES
2072The number of distinct register classes, defined as follows:
2073
2074@smallexample
2075#define N_REG_CLASSES (int) LIM_REG_CLASSES
2076@end smallexample
2077@end defmac
2078
2079@defmac REG_CLASS_NAMES
2080An initializer containing the names of the register classes as C string
2081constants.  These names are used in writing some of the debugging dumps.
2082@end defmac
2083
2084@defmac REG_CLASS_CONTENTS
2085An initializer containing the contents of the register classes, as integers
2086which are bit masks.  The @var{n}th integer specifies the contents of class
2087@var{n}.  The way the integer @var{mask} is interpreted is that
2088register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2089
2090When the machine has more than 32 registers, an integer does not suffice.
2091Then the integers are replaced by sub-initializers, braced groupings containing
2092several integers.  Each sub-initializer must be suitable as an initializer
2093for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2094In this situation, the first integer in each sub-initializer corresponds to
2095registers 0 through 31, the second integer to registers 32 through 63, and
2096so on.
2097@end defmac
2098
2099@defmac REGNO_REG_CLASS (@var{regno})
2100A C expression whose value is a register class containing hard register
2101@var{regno}.  In general there is more than one such class; choose a class
2102which is @dfn{minimal}, meaning that no smaller class also contains the
2103register.
2104@end defmac
2105
2106@defmac BASE_REG_CLASS
2107A macro whose definition is the name of the class to which a valid
2108base register must belong.  A base register is one used in an address
2109which is the register value plus a displacement.
2110@end defmac
2111
2112@defmac MODE_BASE_REG_CLASS (@var{mode})
2113This is a variation of the @code{BASE_REG_CLASS} macro which allows
2114the selection of a base register in a mode dependent manner.  If
2115@var{mode} is VOIDmode then it should return the same value as
2116@code{BASE_REG_CLASS}.
2117@end defmac
2118
2119@defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2120A C expression whose value is the register class to which a valid
2121base register must belong in order to be used in a base plus index
2122register address.  You should define this macro if base plus index
2123addresses have different requirements than other base register uses.
2124@end defmac
2125
2126@defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2127A C expression whose value is the register class to which a valid
2128base register for a memory reference in mode @var{mode} to address
2129space @var{address_space} must belong.  @var{outer_code} and @var{index_code}
2130define the context in which the base register occurs.  @var{outer_code} is
2131the code of the immediately enclosing expression (@code{MEM} for the top level
2132of an address, @code{ADDRESS} for something that occurs in an
2133@code{address_operand}).  @var{index_code} is the code of the corresponding
2134index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2135@end defmac
2136
2137@defmac INDEX_REG_CLASS
2138A macro whose definition is the name of the class to which a valid
2139index register must belong.  An index register is one used in an
2140address where its value is either multiplied by a scale factor or
2141added to another register (as well as added to a displacement).
2142@end defmac
2143
2144@defmac REGNO_OK_FOR_BASE_P (@var{num})
2145A C expression which is nonzero if register number @var{num} is
2146suitable for use as a base register in operand addresses.
2147@end defmac
2148
2149@defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2150A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2151that expression may examine the mode of the memory reference in
2152@var{mode}.  You should define this macro if the mode of the memory
2153reference affects whether a register may be used as a base register.  If
2154you define this macro, the compiler will use it instead of
2155@code{REGNO_OK_FOR_BASE_P}.  The mode may be @code{VOIDmode} for
2156addresses that appear outside a @code{MEM}, i.e., as an
2157@code{address_operand}.
2158@end defmac
2159
2160@defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2161A C expression which is nonzero if register number @var{num} is suitable for
2162use as a base register in base plus index operand addresses, accessing
2163memory in mode @var{mode}.  It may be either a suitable hard register or a
2164pseudo register that has been allocated such a hard register.  You should
2165define this macro if base plus index addresses have different requirements
2166than other base register uses.
2167
2168Use of this macro is deprecated; please use the more general
2169@code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2170@end defmac
2171
2172@defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code})
2173A C expression which is nonzero if register number @var{num} is
2174suitable for use as a base register in operand addresses, accessing
2175memory in mode @var{mode} in address space @var{address_space}.
2176This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except
2177that that expression may examine the context in which the register
2178appears in the memory reference.  @var{outer_code} is the code of the
2179immediately enclosing expression (@code{MEM} if at the top level of the
2180address, @code{ADDRESS} for something that occurs in an
2181@code{address_operand}).  @var{index_code} is the code of the
2182corresponding index expression if @var{outer_code} is @code{PLUS};
2183@code{SCRATCH} otherwise.  The mode may be @code{VOIDmode} for addresses
2184that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2185@end defmac
2186
2187@defmac REGNO_OK_FOR_INDEX_P (@var{num})
2188A C expression which is nonzero if register number @var{num} is
2189suitable for use as an index register in operand addresses.  It may be
2190either a suitable hard register or a pseudo register that has been
2191allocated such a hard register.
2192
2193The difference between an index register and a base register is that
2194the index register may be scaled.  If an address involves the sum of
2195two registers, neither one of them scaled, then either one may be
2196labeled the ``base'' and the other the ``index''; but whichever
2197labeling is used must fit the machine's constraints of which registers
2198may serve in each capacity.  The compiler will try both labelings,
2199looking for one that is valid, and will reload one or both registers
2200only if neither labeling works.
2201@end defmac
2202
2203@hook TARGET_PREFERRED_RENAME_CLASS
2204
2205@hook TARGET_PREFERRED_RELOAD_CLASS
2206
2207@defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2208A C expression that places additional restrictions on the register class
2209to use when it is necessary to copy value @var{x} into a register in class
2210@var{class}.  The value is a register class; perhaps @var{class}, or perhaps
2211another, smaller class.  On many machines, the following definition is
2212safe:
2213
2214@smallexample
2215#define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2216@end smallexample
2217
2218Sometimes returning a more restrictive class makes better code.  For
2219example, on the 68000, when @var{x} is an integer constant that is in range
2220for a @samp{moveq} instruction, the value of this macro is always
2221@code{DATA_REGS} as long as @var{class} includes the data registers.
2222Requiring a data register guarantees that a @samp{moveq} will be used.
2223
2224One case where @code{PREFERRED_RELOAD_CLASS} must not return
2225@var{class} is if @var{x} is a legitimate constant which cannot be
2226loaded into some register class.  By returning @code{NO_REGS} you can
2227force @var{x} into a memory location.  For example, rs6000 can load
2228immediate values into general-purpose registers, but does not have an
2229instruction for loading an immediate value into a floating-point
2230register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2231@var{x} is a floating-point constant.  If the constant cannot be loaded
2232into any kind of register, code generation will be better if
2233@code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2234of using @code{TARGET_PREFERRED_RELOAD_CLASS}.
2235
2236If an insn has pseudos in it after register allocation, reload will go
2237through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2238to find the best one.  Returning @code{NO_REGS}, in this case, makes
2239reload add a @code{!} in front of the constraint: the x86 back-end uses
2240this feature to discourage usage of 387 registers when math is done in
2241the SSE registers (and vice versa).
2242@end defmac
2243
2244@hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS
2245
2246@defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2247A C expression that places additional restrictions on the register class
2248to use when it is necessary to be able to hold a value of mode
2249@var{mode} in a reload register for which class @var{class} would
2250ordinarily be used.
2251
2252Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2253there are certain modes that simply cannot go in certain reload classes.
2254
2255The value is a register class; perhaps @var{class}, or perhaps another,
2256smaller class.
2257
2258Don't define this macro unless the target machine has limitations which
2259require the macro to do something nontrivial.
2260@end defmac
2261
2262@hook TARGET_SECONDARY_RELOAD
2263
2264@defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2265@defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2266@defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2267These macros are obsolete, new ports should use the target hook
2268@code{TARGET_SECONDARY_RELOAD} instead.
2269
2270These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2271target hook.  Older ports still define these macros to indicate to the
2272reload phase that it may
2273need to allocate at least one register for a reload in addition to the
2274register to contain the data.  Specifically, if copying @var{x} to a
2275register @var{class} in @var{mode} requires an intermediate register,
2276you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2277largest register class all of whose registers can be used as
2278intermediate registers or scratch registers.
2279
2280If copying a register @var{class} in @var{mode} to @var{x} requires an
2281intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2282was supposed to be defined to return the largest register
2283class required.  If the
2284requirements for input and output reloads were the same, the macro
2285@code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2286macros identically.
2287
2288The values returned by these macros are often @code{GENERAL_REGS}.
2289Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2290can be directly copied to or from a register of @var{class} in
2291@var{mode} without requiring a scratch register.  Do not define this
2292macro if it would always return @code{NO_REGS}.
2293
2294If a scratch register is required (either with or without an
2295intermediate register), you were supposed to define patterns for
2296@samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2297(@pxref{Standard Names}.  These patterns, which were normally
2298implemented with a @code{define_expand}, should be similar to the
2299@samp{mov@var{m}} patterns, except that operand 2 is the scratch
2300register.
2301
2302These patterns need constraints for the reload register and scratch
2303register that
2304contain a single register class.  If the original reload register (whose
2305class is @var{class}) can meet the constraint given in the pattern, the
2306value returned by these macros is used for the class of the scratch
2307register.  Otherwise, two additional reload registers are required.
2308Their classes are obtained from the constraints in the insn pattern.
2309
2310@var{x} might be a pseudo-register or a @code{subreg} of a
2311pseudo-register, which could either be in a hard register or in memory.
2312Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2313in memory and the hard register number if it is in a register.
2314
2315These macros should not be used in the case where a particular class of
2316registers can only be copied to memory and not to another class of
2317registers.  In that case, secondary reload registers are not needed and
2318would not be helpful.  Instead, a stack location must be used to perform
2319the copy and the @code{mov@var{m}} pattern should use memory as an
2320intermediate storage.  This case often occurs between floating-point and
2321general registers.
2322@end defmac
2323
2324@hook TARGET_SECONDARY_MEMORY_NEEDED
2325
2326@defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2327Normally when @code{TARGET_SECONDARY_MEMORY_NEEDED} is defined, the compiler
2328allocates a stack slot for a memory location needed for register copies.
2329If this macro is defined, the compiler instead uses the memory location
2330defined by this macro.
2331
2332Do not define this macro if you do not define
2333@code{TARGET_SECONDARY_MEMORY_NEEDED}.
2334@end defmac
2335
2336@hook TARGET_SECONDARY_MEMORY_NEEDED_MODE
2337
2338@hook TARGET_SELECT_EARLY_REMAT_MODES
2339
2340@hook TARGET_CLASS_LIKELY_SPILLED_P
2341
2342@hook TARGET_CLASS_MAX_NREGS
2343
2344@defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2345A C expression for the maximum number of consecutive registers
2346of class @var{class} needed to hold a value of mode @var{mode}.
2347
2348This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}.  In fact,
2349the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2350should be the maximum value of @code{TARGET_HARD_REGNO_NREGS (@var{regno},
2351@var{mode})} for all @var{regno} values in the class @var{class}.
2352
2353This macro helps control the handling of multiple-word values
2354in the reload pass.
2355@end defmac
2356
2357@hook TARGET_CAN_CHANGE_MODE_CLASS
2358
2359@hook TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS
2360
2361@hook TARGET_LRA_P
2362
2363@hook TARGET_REGISTER_PRIORITY
2364
2365@hook TARGET_REGISTER_USAGE_LEVELING_P
2366
2367@hook TARGET_DIFFERENT_ADDR_DISPLACEMENT_P
2368
2369@hook TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P
2370
2371@hook TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT
2372
2373@hook TARGET_SPILL_CLASS
2374
2375@hook TARGET_ADDITIONAL_ALLOCNO_CLASS_P
2376
2377@hook TARGET_CSTORE_MODE
2378
2379@hook TARGET_COMPUTE_PRESSURE_CLASSES
2380
2381@node Stack and Calling
2382@section Stack Layout and Calling Conventions
2383@cindex calling conventions
2384
2385@c prevent bad page break with this line
2386This describes the stack layout and calling conventions.
2387
2388@menu
2389* Frame Layout::
2390* Exception Handling::
2391* Stack Checking::
2392* Frame Registers::
2393* Elimination::
2394* Stack Arguments::
2395* Register Arguments::
2396* Scalar Return::
2397* Aggregate Return::
2398* Caller Saves::
2399* Function Entry::
2400* Profiling::
2401* Tail Calls::
2402* Shrink-wrapping separate components::
2403* Stack Smashing Protection::
2404* Miscellaneous Register Hooks::
2405@end menu
2406
2407@node Frame Layout
2408@subsection Basic Stack Layout
2409@cindex stack frame layout
2410@cindex frame layout
2411
2412@c prevent bad page break with this line
2413Here is the basic stack layout.
2414
2415@defmac STACK_GROWS_DOWNWARD
2416Define this macro to be true if pushing a word onto the stack moves the stack
2417pointer to a smaller address, and false otherwise.
2418@end defmac
2419
2420@defmac STACK_PUSH_CODE
2421This macro defines the operation used when something is pushed
2422on the stack.  In RTL, a push operation will be
2423@code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
2424
2425The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
2426and @code{POST_INC}.  Which of these is correct depends on
2427the stack direction and on whether the stack pointer points
2428to the last item on the stack or whether it points to the
2429space for the next item on the stack.
2430
2431The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
2432true, which is almost always right, and @code{PRE_INC} otherwise,
2433which is often wrong.
2434@end defmac
2435
2436@defmac FRAME_GROWS_DOWNWARD
2437Define this macro to nonzero value if the addresses of local variable slots
2438are at negative offsets from the frame pointer.
2439@end defmac
2440
2441@defmac ARGS_GROW_DOWNWARD
2442Define this macro if successive arguments to a function occupy decreasing
2443addresses on the stack.
2444@end defmac
2445
2446@hook TARGET_STARTING_FRAME_OFFSET
2447
2448@defmac STACK_ALIGNMENT_NEEDED
2449Define to zero to disable final alignment of the stack during reload.
2450The nonzero default for this macro is suitable for most ports.
2451
2452On ports where @code{TARGET_STARTING_FRAME_OFFSET} is nonzero or where there
2453is a register save block following the local block that doesn't require
2454alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
2455stack alignment and do it in the backend.
2456@end defmac
2457
2458@defmac STACK_POINTER_OFFSET
2459Offset from the stack pointer register to the first location at which
2460outgoing arguments are placed.  If not specified, the default value of
2461zero is used.  This is the proper value for most machines.
2462
2463If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2464the first location at which outgoing arguments are placed.
2465@end defmac
2466
2467@defmac FIRST_PARM_OFFSET (@var{fundecl})
2468Offset from the argument pointer register to the first argument's
2469address.  On some machines it may depend on the data type of the
2470function.
2471
2472If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
2473the first argument's address.
2474@end defmac
2475
2476@defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
2477Offset from the stack pointer register to an item dynamically allocated
2478on the stack, e.g., by @code{alloca}.
2479
2480The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
2481length of the outgoing arguments.  The default is correct for most
2482machines.  See @file{function.cc} for details.
2483@end defmac
2484
2485@defmac INITIAL_FRAME_ADDRESS_RTX
2486A C expression whose value is RTL representing the address of the initial
2487stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
2488@code{DYNAMIC_CHAIN_ADDRESS}.  If you don't define this macro, a reasonable
2489default value will be used.  Define this macro in order to make frame pointer
2490elimination work in the presence of @code{__builtin_frame_address (count)} and
2491@code{__builtin_return_address (count)} for @code{count} not equal to zero.
2492@end defmac
2493
2494@defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
2495A C expression whose value is RTL representing the address in a stack
2496frame where the pointer to the caller's frame is stored.  Assume that
2497@var{frameaddr} is an RTL expression for the address of the stack frame
2498itself.
2499
2500If you don't define this macro, the default is to return the value
2501of @var{frameaddr}---that is, the stack frame address is also the
2502address of the stack word that points to the previous frame.
2503@end defmac
2504
2505@defmac SETUP_FRAME_ADDRESSES
2506A C expression that produces the machine-specific code to
2507setup the stack so that arbitrary frames can be accessed.  For example,
2508on the SPARC, we must flush all of the register windows to the stack
2509before we can access arbitrary stack frames.  You will seldom need to
2510define this macro.  The default is to do nothing.
2511@end defmac
2512
2513@hook TARGET_BUILTIN_SETJMP_FRAME_VALUE
2514
2515@defmac FRAME_ADDR_RTX (@var{frameaddr})
2516A C expression whose value is RTL representing the value of the frame
2517address for the current frame.  @var{frameaddr} is the frame pointer
2518of the current frame.  This is used for __builtin_frame_address.
2519You need only define this macro if the frame address is not the same
2520as the frame pointer.  Most machines do not need to define it.
2521@end defmac
2522
2523@defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
2524A C expression whose value is RTL representing the value of the return
2525address for the frame @var{count} steps up from the current frame, after
2526the prologue.  @var{frameaddr} is the frame pointer of the @var{count}
2527frame, or the frame pointer of the @var{count} @minus{} 1 frame if
2528@code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero.
2529
2530The value of the expression must always be the correct address when
2531@var{count} is zero, but may be @code{NULL_RTX} if there is no way to
2532determine the return address of other frames.
2533@end defmac
2534
2535@defmac RETURN_ADDR_IN_PREVIOUS_FRAME
2536Define this macro to nonzero value if the return address of a particular
2537stack frame is accessed from the frame pointer of the previous stack
2538frame.  The zero default for this macro is suitable for most ports.
2539@end defmac
2540
2541@defmac INCOMING_RETURN_ADDR_RTX
2542A C expression whose value is RTL representing the location of the
2543incoming return address at the beginning of any function, before the
2544prologue.  This RTL is either a @code{REG}, indicating that the return
2545value is saved in @samp{REG}, or a @code{MEM} representing a location in
2546the stack.
2547
2548You only need to define this macro if you want to support call frame
2549debugging information like that provided by DWARF 2.
2550
2551If this RTL is a @code{REG}, you should also define
2552@code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
2553@end defmac
2554
2555@defmac DWARF_ALT_FRAME_RETURN_COLUMN
2556A C expression whose value is an integer giving a DWARF 2 column
2557number that may be used as an alternative return column.  The column
2558must not correspond to any gcc hard register (that is, it must not
2559be in the range of @code{DWARF_FRAME_REGNUM}).
2560
2561This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
2562general register, but an alternative column needs to be used for signal
2563frames.  Some targets have also used different frame return columns
2564over time.
2565@end defmac
2566
2567@defmac DWARF_ZERO_REG
2568A C expression whose value is an integer giving a DWARF 2 register
2569number that is considered to always have the value zero.  This should
2570only be defined if the target has an architected zero register, and
2571someone decided it was a good idea to use that register number to
2572terminate the stack backtrace.  New ports should avoid this.
2573@end defmac
2574
2575@hook TARGET_DWARF_HANDLE_FRAME_UNSPEC
2576
2577@hook TARGET_DWARF_POLY_INDETERMINATE_VALUE
2578
2579@defmac INCOMING_FRAME_SP_OFFSET
2580A C expression whose value is an integer giving the offset, in bytes,
2581from the value of the stack pointer register to the top of the stack
2582frame at the beginning of any function, before the prologue.  The top of
2583the frame is defined to be the value of the stack pointer in the
2584previous frame, just before the call instruction.
2585
2586You only need to define this macro if you want to support call frame
2587debugging information like that provided by DWARF 2.
2588@end defmac
2589
2590@defmac DEFAULT_INCOMING_FRAME_SP_OFFSET
2591Like @code{INCOMING_FRAME_SP_OFFSET}, but must be the same for all
2592functions of the same ABI, and when using GAS @code{.cfi_*} directives
2593must also agree with the default CFI GAS emits.  Define this macro
2594only if @code{INCOMING_FRAME_SP_OFFSET} can have different values
2595between different functions of the same ABI or when
2596@code{INCOMING_FRAME_SP_OFFSET} does not agree with GAS default CFI.
2597@end defmac
2598
2599@defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
2600A C expression whose value is an integer giving the offset, in bytes,
2601from the argument pointer to the canonical frame address (cfa).  The
2602final value should coincide with that calculated by
2603@code{INCOMING_FRAME_SP_OFFSET}.  Which is unfortunately not usable
2604during virtual register instantiation.
2605
2606The default value for this macro is
2607@code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size},
2608which is correct for most machines; in general, the arguments are found
2609immediately before the stack frame.  Note that this is not the case on
2610some targets that save registers into the caller's frame, such as SPARC
2611and rs6000, and so such targets need to define this macro.
2612
2613You only need to define this macro if the default is incorrect, and you
2614want to support call frame debugging information like that provided by
2615DWARF 2.
2616@end defmac
2617
2618@defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
2619If defined, a C expression whose value is an integer giving the offset
2620in bytes from the frame pointer to the canonical frame address (cfa).
2621The final value should coincide with that calculated by
2622@code{INCOMING_FRAME_SP_OFFSET}.
2623
2624Normally the CFA is calculated as an offset from the argument pointer,
2625via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
2626variable due to the ABI, this may not be possible.  If this macro is
2627defined, it implies that the virtual register instantiation should be
2628based on the frame pointer instead of the argument pointer.  Only one
2629of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
2630should be defined.
2631@end defmac
2632
2633@defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
2634If defined, a C expression whose value is an integer giving the offset
2635in bytes from the canonical frame address (cfa) to the frame base used
2636in DWARF 2 debug information.  The default is zero.  A different value
2637may reduce the size of debug information on some ports.
2638@end defmac
2639
2640@node Exception Handling
2641@subsection Exception Handling Support
2642@cindex exception handling
2643
2644@defmac EH_RETURN_DATA_REGNO (@var{N})
2645A C expression whose value is the @var{N}th register number used for
2646data by exception handlers, or @code{INVALID_REGNUM} if fewer than
2647@var{N} registers are usable.
2648
2649The exception handling library routines communicate with the exception
2650handlers via a set of agreed upon registers.  Ideally these registers
2651should be call-clobbered; it is possible to use call-saved registers,
2652but may negatively impact code size.  The target must support at least
26532 data registers, but should define 4 if there are enough free registers.
2654
2655You must define this macro if you want to support call frame exception
2656handling like that provided by DWARF 2.
2657@end defmac
2658
2659@defmac EH_RETURN_STACKADJ_RTX
2660A C expression whose value is RTL representing a location in which
2661to store a stack adjustment to be applied before function return.
2662This is used to unwind the stack to an exception handler's call frame.
2663It will be assigned zero on code paths that return normally.
2664
2665Typically this is a call-clobbered hard register that is otherwise
2666untouched by the epilogue, but could also be a stack slot.
2667
2668Do not define this macro if the stack pointer is saved and restored
2669by the regular prolog and epilog code in the call frame itself; in
2670this case, the exception handling library routines will update the
2671stack location to be restored in place.  Otherwise, you must define
2672this macro if you want to support call frame exception handling like
2673that provided by DWARF 2.
2674@end defmac
2675
2676@defmac EH_RETURN_HANDLER_RTX
2677A C expression whose value is RTL representing a location in which
2678to store the address of an exception handler to which we should
2679return.  It will not be assigned on code paths that return normally.
2680
2681Typically this is the location in the call frame at which the normal
2682return address is stored.  For targets that return by popping an
2683address off the stack, this might be a memory address just below
2684the @emph{target} call frame rather than inside the current call
2685frame.  If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
2686been assigned, so it may be used to calculate the location of the
2687target call frame.
2688
2689Some targets have more complex requirements than storing to an
2690address calculable during initial code generation.  In that case
2691the @code{eh_return} instruction pattern should be used instead.
2692
2693If you want to support call frame exception handling, you must
2694define either this macro or the @code{eh_return} instruction pattern.
2695@end defmac
2696
2697@defmac RETURN_ADDR_OFFSET
2698If defined, an integer-valued C expression for which rtl will be generated
2699to add it to the exception handler address before it is searched in the
2700exception handling tables, and to subtract it again from the address before
2701using it to return to the exception handler.
2702@end defmac
2703
2704@defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
2705This macro chooses the encoding of pointers embedded in the exception
2706handling sections.  If at all possible, this should be defined such
2707that the exception handling section will not require dynamic relocations,
2708and so may be read-only.
2709
2710@var{code} is 0 for data, 1 for code labels, 2 for function pointers.
2711@var{global} is true if the symbol may be affected by dynamic relocations.
2712The macro should return a combination of the @code{DW_EH_PE_*} defines
2713as found in @file{dwarf2.h}.
2714
2715If this macro is not defined, pointers will not be encoded but
2716represented directly.
2717@end defmac
2718
2719@defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
2720This macro allows the target to emit whatever special magic is required
2721to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
2722Generic code takes care of pc-relative and indirect encodings; this must
2723be defined if the target uses text-relative or data-relative encodings.
2724
2725This is a C statement that branches to @var{done} if the format was
2726handled.  @var{encoding} is the format chosen, @var{size} is the number
2727of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
2728to be emitted.
2729@end defmac
2730
2731@defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
2732This macro allows the target to add CPU and operating system specific
2733code to the call-frame unwinder for use when there is no unwind data
2734available.  The most common reason to implement this macro is to unwind
2735through signal frames.
2736
2737This macro is called from @code{uw_frame_state_for} in
2738@file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
2739@file{unwind-ia64.c}.  @var{context} is an @code{_Unwind_Context};
2740@var{fs} is an @code{_Unwind_FrameState}.  Examine @code{context->ra}
2741for the address of the code being executed and @code{context->cfa} for
2742the stack pointer value.  If the frame can be decoded, the register
2743save addresses should be updated in @var{fs} and the macro should
2744evaluate to @code{_URC_NO_REASON}.  If the frame cannot be decoded,
2745the macro should evaluate to @code{_URC_END_OF_STACK}.
2746
2747For proper signal handling in Java this macro is accompanied by
2748@code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
2749@end defmac
2750
2751@defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
2752This macro allows the target to add operating system specific code to the
2753call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
2754usually used for signal or interrupt frames.
2755
2756This macro is called from @code{uw_update_context} in libgcc's
2757@file{unwind-ia64.c}.  @var{context} is an @code{_Unwind_Context};
2758@var{fs} is an @code{_Unwind_FrameState}.  Examine @code{fs->unwabi}
2759for the abi and context in the @code{.unwabi} directive.  If the
2760@code{.unwabi} directive can be handled, the register save addresses should
2761be updated in @var{fs}.
2762@end defmac
2763
2764@defmac TARGET_USES_WEAK_UNWIND_INFO
2765A C expression that evaluates to true if the target requires unwind
2766info to be given comdat linkage.  Define it to be @code{1} if comdat
2767linkage is necessary.  The default is @code{0}.
2768@end defmac
2769
2770@node Stack Checking
2771@subsection Specifying How Stack Checking is Done
2772
2773GCC will check that stack references are within the boundaries of the
2774stack, if the option @option{-fstack-check} is specified, in one of
2775three ways:
2776
2777@enumerate
2778@item
2779If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
2780will assume that you have arranged for full stack checking to be done
2781at appropriate places in the configuration files.  GCC will not do
2782other special processing.
2783
2784@item
2785If @code{STACK_CHECK_BUILTIN} is zero and the value of the
2786@code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
2787that you have arranged for static stack checking (checking of the
2788static stack frame of functions) to be done at appropriate places
2789in the configuration files.  GCC will only emit code to do dynamic
2790stack checking (checking on dynamic stack allocations) using the third
2791approach below.
2792
2793@item
2794If neither of the above are true, GCC will generate code to periodically
2795``probe'' the stack pointer using the values of the macros defined below.
2796@end enumerate
2797
2798If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
2799GCC will change its allocation strategy for large objects if the option
2800@option{-fstack-check} is specified: they will always be allocated
2801dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
2802
2803@defmac STACK_CHECK_BUILTIN
2804A nonzero value if stack checking is done by the configuration files in a
2805machine-dependent manner.  You should define this macro if stack checking
2806is required by the ABI of your machine or if you would like to do stack
2807checking in some more efficient way than the generic approach.  The default
2808value of this macro is zero.
2809@end defmac
2810
2811@defmac STACK_CHECK_STATIC_BUILTIN
2812A nonzero value if static stack checking is done by the configuration files
2813in a machine-dependent manner.  You should define this macro if you would
2814like to do static stack checking in some more efficient way than the generic
2815approach.  The default value of this macro is zero.
2816@end defmac
2817
2818@defmac STACK_CHECK_PROBE_INTERVAL_EXP
2819An integer specifying the interval at which GCC must generate stack probe
2820instructions, defined as 2 raised to this integer.  You will normally
2821define this macro so that the interval be no larger than the size of
2822the ``guard pages'' at the end of a stack area.  The default value
2823of 12 (4096-byte interval) is suitable for most systems.
2824@end defmac
2825
2826@defmac STACK_CHECK_MOVING_SP
2827An integer which is nonzero if GCC should move the stack pointer page by page
2828when doing probes.  This can be necessary on systems where the stack pointer
2829contains the bottom address of the memory area accessible to the executing
2830thread at any point in time.  In this situation an alternate signal stack
2831is required in order to be able to recover from a stack overflow.  The
2832default value of this macro is zero.
2833@end defmac
2834
2835@defmac STACK_CHECK_PROTECT
2836The number of bytes of stack needed to recover from a stack overflow, for
2837languages where such a recovery is supported.  The default value of 4KB/8KB
2838with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and
28398KB/12KB with other exception handling mechanisms should be adequate for most
2840architectures and operating systems.
2841@end defmac
2842
2843The following macros are relevant only if neither STACK_CHECK_BUILTIN
2844nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
2845in the opposite case.
2846
2847@defmac STACK_CHECK_MAX_FRAME_SIZE
2848The maximum size of a stack frame, in bytes.  GCC will generate probe
2849instructions in non-leaf functions to ensure at least this many bytes of
2850stack are available.  If a stack frame is larger than this size, stack
2851checking will not be reliable and GCC will issue a warning.  The
2852default is chosen so that GCC only generates one instruction on most
2853systems.  You should normally not change the default value of this macro.
2854@end defmac
2855
2856@defmac STACK_CHECK_FIXED_FRAME_SIZE
2857GCC uses this value to generate the above warning message.  It
2858represents the amount of fixed frame used by a function, not including
2859space for any callee-saved registers, temporaries and user variables.
2860You need only specify an upper bound for this amount and will normally
2861use the default of four words.
2862@end defmac
2863
2864@defmac STACK_CHECK_MAX_VAR_SIZE
2865The maximum size, in bytes, of an object that GCC will place in the
2866fixed area of the stack frame when the user specifies
2867@option{-fstack-check}.
2868GCC computed the default from the values of the above macros and you will
2869normally not need to override that default.
2870@end defmac
2871
2872@hook TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE
2873
2874@need 2000
2875@node Frame Registers
2876@subsection Registers That Address the Stack Frame
2877
2878@c prevent bad page break with this line
2879This discusses registers that address the stack frame.
2880
2881@defmac STACK_POINTER_REGNUM
2882The register number of the stack pointer register, which must also be a
2883fixed register according to @code{FIXED_REGISTERS}.  On most machines,
2884the hardware determines which register this is.
2885@end defmac
2886
2887@defmac FRAME_POINTER_REGNUM
2888The register number of the frame pointer register, which is used to
2889access automatic variables in the stack frame.  On some machines, the
2890hardware determines which register this is.  On other machines, you can
2891choose any register you wish for this purpose.
2892@end defmac
2893
2894@defmac HARD_FRAME_POINTER_REGNUM
2895On some machines the offset between the frame pointer and starting
2896offset of the automatic variables is not known until after register
2897allocation has been done (for example, because the saved registers are
2898between these two locations).  On those machines, define
2899@code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
2900be used internally until the offset is known, and define
2901@code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
2902used for the frame pointer.
2903
2904You should define this macro only in the very rare circumstances when it
2905is not possible to calculate the offset between the frame pointer and
2906the automatic variables until after register allocation has been
2907completed.  When this macro is defined, you must also indicate in your
2908definition of @code{ELIMINABLE_REGS} how to eliminate
2909@code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
2910or @code{STACK_POINTER_REGNUM}.
2911
2912Do not define this macro if it would be the same as
2913@code{FRAME_POINTER_REGNUM}.
2914@end defmac
2915
2916@defmac ARG_POINTER_REGNUM
2917The register number of the arg pointer register, which is used to access
2918the function's argument list.  On some machines, this is the same as the
2919frame pointer register.  On some machines, the hardware determines which
2920register this is.  On other machines, you can choose any register you
2921wish for this purpose.  If this is not the same register as the frame
2922pointer register, then you must mark it as a fixed register according to
2923@code{FIXED_REGISTERS}, or arrange to be able to eliminate it
2924(@pxref{Elimination}).
2925@end defmac
2926
2927@defmac HARD_FRAME_POINTER_IS_FRAME_POINTER
2928Define this to a preprocessor constant that is nonzero if
2929@code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be
2930the same.  The default definition is @samp{(HARD_FRAME_POINTER_REGNUM
2931== FRAME_POINTER_REGNUM)}; you only need to define this macro if that
2932definition is not suitable for use in preprocessor conditionals.
2933@end defmac
2934
2935@defmac HARD_FRAME_POINTER_IS_ARG_POINTER
2936Define this to a preprocessor constant that is nonzero if
2937@code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the
2938same.  The default definition is @samp{(HARD_FRAME_POINTER_REGNUM ==
2939ARG_POINTER_REGNUM)}; you only need to define this macro if that
2940definition is not suitable for use in preprocessor conditionals.
2941@end defmac
2942
2943@defmac RETURN_ADDRESS_POINTER_REGNUM
2944The register number of the return address pointer register, which is used to
2945access the current function's return address from the stack.  On some
2946machines, the return address is not at a fixed offset from the frame
2947pointer or stack pointer or argument pointer.  This register can be defined
2948to point to the return address on the stack, and then be converted by
2949@code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
2950
2951Do not define this macro unless there is no other way to get the return
2952address from the stack.
2953@end defmac
2954
2955@defmac STATIC_CHAIN_REGNUM
2956@defmacx STATIC_CHAIN_INCOMING_REGNUM
2957Register numbers used for passing a function's static chain pointer.  If
2958register windows are used, the register number as seen by the called
2959function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
2960number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}.  If
2961these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
2962not be defined.
2963
2964The static chain register need not be a fixed register.
2965
2966If the static chain is passed in memory, these macros should not be
2967defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
2968@end defmac
2969
2970@hook TARGET_STATIC_CHAIN
2971
2972@defmac DWARF_FRAME_REGISTERS
2973This macro specifies the maximum number of hard registers that can be
2974saved in a call frame.  This is used to size data structures used in
2975DWARF2 exception handling.
2976
2977Prior to GCC 3.0, this macro was needed in order to establish a stable
2978exception handling ABI in the face of adding new hard registers for ISA
2979extensions.  In GCC 3.0 and later, the EH ABI is insulated from changes
2980in the number of hard registers.  Nevertheless, this macro can still be
2981used to reduce the runtime memory requirements of the exception handling
2982routines, which can be substantial if the ISA contains a lot of
2983registers that are not call-saved.
2984
2985If this macro is not defined, it defaults to
2986@code{FIRST_PSEUDO_REGISTER}.
2987@end defmac
2988
2989@defmac PRE_GCC3_DWARF_FRAME_REGISTERS
2990
2991This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
2992for backward compatibility in pre GCC 3.0 compiled code.
2993
2994If this macro is not defined, it defaults to
2995@code{DWARF_FRAME_REGISTERS}.
2996@end defmac
2997
2998@defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
2999
3000Define this macro if the target's representation for dwarf registers
3001is different than the internal representation for unwind column.
3002Given a dwarf register, this macro should return the internal unwind
3003column number to use instead.
3004@end defmac
3005
3006@defmac DWARF_FRAME_REGNUM (@var{regno})
3007
3008Define this macro if the target's representation for dwarf registers
3009used in .eh_frame or .debug_frame is different from that used in other
3010debug info sections.  Given a GCC hard register number, this macro
3011should return the .eh_frame register number.  The default is
3012@code{DBX_REGISTER_NUMBER (@var{regno})}.
3013
3014@end defmac
3015
3016@defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3017
3018Define this macro to map register numbers held in the call frame info
3019that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3020should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3021.eh_frame (@code{@var{for_eh}} is nonzero).  The default is to
3022return @code{@var{regno}}.
3023
3024@end defmac
3025
3026@defmac REG_VALUE_IN_UNWIND_CONTEXT
3027
3028Define this macro if the target stores register values as
3029@code{_Unwind_Word} type in unwind context.  It should be defined if
3030target register size is larger than the size of @code{void *}.  The
3031default is to store register values as @code{void *} type.
3032
3033@end defmac
3034
3035@defmac ASSUME_EXTENDED_UNWIND_CONTEXT
3036
3037Define this macro to be 1 if the target always uses extended unwind
3038context with version, args_size and by_value fields.  If it is undefined,
3039it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is
3040defined and 0 otherwise.
3041
3042@end defmac
3043
3044@defmac DWARF_LAZY_REGISTER_VALUE (@var{regno}, @var{value})
3045Define this macro if the target has pseudo DWARF registers whose
3046values need to be computed lazily on demand by the unwinder (such as when
3047referenced in a CFA expression).  The macro returns true if @var{regno}
3048is such a register and stores its value in @samp{*@var{value}} if so.
3049@end defmac
3050
3051@node Elimination
3052@subsection Eliminating Frame Pointer and Arg Pointer
3053
3054@c prevent bad page break with this line
3055This is about eliminating the frame pointer and arg pointer.
3056
3057@hook TARGET_FRAME_POINTER_REQUIRED
3058
3059@defmac ELIMINABLE_REGS
3060This macro specifies a table of register pairs used to eliminate
3061unneeded registers that point into the stack frame.
3062
3063The definition of this macro is a list of structure initializations, each
3064of which specifies an original and replacement register.
3065
3066On some machines, the position of the argument pointer is not known until
3067the compilation is completed.  In such a case, a separate hard register
3068must be used for the argument pointer.  This register can be eliminated by
3069replacing it with either the frame pointer or the argument pointer,
3070depending on whether or not the frame pointer has been eliminated.
3071
3072In this case, you might specify:
3073@smallexample
3074#define ELIMINABLE_REGS  \
3075@{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3076 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3077 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3078@end smallexample
3079
3080Note that the elimination of the argument pointer with the stack pointer is
3081specified first since that is the preferred elimination.
3082@end defmac
3083
3084@hook TARGET_CAN_ELIMINATE
3085
3086@defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3087This macro returns the initial difference between the specified pair
3088of registers.  The value would be computed from information
3089such as the result of @code{get_frame_size ()} and the tables of
3090registers @code{df_regs_ever_live_p} and @code{call_used_regs}.
3091@end defmac
3092
3093@hook TARGET_COMPUTE_FRAME_LAYOUT
3094
3095@node Stack Arguments
3096@subsection Passing Function Arguments on the Stack
3097@cindex arguments on stack
3098@cindex stack arguments
3099
3100The macros in this section control how arguments are passed
3101on the stack.  See the following section for other macros that
3102control passing certain arguments in registers.
3103
3104@hook TARGET_PROMOTE_PROTOTYPES
3105
3106@hook TARGET_PUSH_ARGUMENT
3107
3108@defmac PUSH_ARGS_REVERSED
3109A C expression.  If nonzero, function arguments will be evaluated from
3110last to first, rather than from first to last.  If this macro is not
3111defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3112and args grow in opposite directions, and 0 otherwise.
3113@end defmac
3114
3115@defmac PUSH_ROUNDING (@var{npushed})
3116A C expression that is the number of bytes actually pushed onto the
3117stack when an instruction attempts to push @var{npushed} bytes.
3118
3119On some machines, the definition
3120
3121@smallexample
3122#define PUSH_ROUNDING(BYTES) (BYTES)
3123@end smallexample
3124
3125@noindent
3126will suffice.  But on other machines, instructions that appear
3127to push one byte actually push two bytes in an attempt to maintain
3128alignment.  Then the definition should be
3129
3130@smallexample
3131#define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3132@end smallexample
3133
3134If the value of this macro has a type, it should be an unsigned type.
3135@end defmac
3136
3137@findex outgoing_args_size
3138@findex crtl->outgoing_args_size
3139@defmac ACCUMULATE_OUTGOING_ARGS
3140A C expression.  If nonzero, the maximum amount of space required for outgoing arguments
3141will be computed and placed into
3142@code{crtl->outgoing_args_size}.  No space will be pushed
3143onto the stack for each call; instead, the function prologue should
3144increase the stack frame size by this amount.
3145
3146Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3147is not proper.
3148@end defmac
3149
3150@defmac REG_PARM_STACK_SPACE (@var{fndecl})
3151Define this macro if functions should assume that stack space has been
3152allocated for arguments even when their values are passed in
3153registers.
3154
3155The value of this macro is the size, in bytes, of the area reserved for
3156arguments passed in registers for the function represented by @var{fndecl},
3157which can be zero if GCC is calling a library function.
3158The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3159of the function.
3160
3161This space can be allocated by the caller, or be a part of the
3162machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3163which.
3164@end defmac
3165@c above is overfull.  not sure what to do.  --mew 5feb93  did
3166@c something, not sure if it looks good.  --mew 10feb93
3167
3168@defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl})
3169Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments.
3170Define this macro if space guaranteed when compiling a function body
3171is different to space required when making a call, a situation that
3172can arise with K&R style function definitions.
3173@end defmac
3174
3175@defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3176Define this to a nonzero value if it is the responsibility of the
3177caller to allocate the area reserved for arguments passed in registers
3178when calling a function of @var{fntype}.  @var{fntype} may be NULL
3179if the function called is a library function.
3180
3181If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3182whether the space for these arguments counts in the value of
3183@code{crtl->outgoing_args_size}.
3184@end defmac
3185
3186@defmac STACK_PARMS_IN_REG_PARM_AREA
3187Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3188stack parameters don't skip the area specified by it.
3189@c i changed this, makes more sens and it should have taken care of the
3190@c overfull.. not as specific, tho.  --mew 5feb93
3191
3192Normally, when a parameter is not passed in registers, it is placed on the
3193stack beyond the @code{REG_PARM_STACK_SPACE} area.  Defining this macro
3194suppresses this behavior and causes the parameter to be passed on the
3195stack in its natural location.
3196@end defmac
3197
3198@hook TARGET_RETURN_POPS_ARGS
3199
3200@defmac CALL_POPS_ARGS (@var{cum})
3201A C expression that should indicate the number of bytes a call sequence
3202pops off the stack.  It is added to the value of @code{RETURN_POPS_ARGS}
3203when compiling a function call.
3204
3205@var{cum} is the variable in which all arguments to the called function
3206have been accumulated.
3207
3208On certain architectures, such as the SH5, a call trampoline is used
3209that pops certain registers off the stack, depending on the arguments
3210that have been passed to the function.  Since this is a property of the
3211call site, not of the called function, @code{RETURN_POPS_ARGS} is not
3212appropriate.
3213@end defmac
3214
3215@node Register Arguments
3216@subsection Passing Arguments in Registers
3217@cindex arguments in registers
3218@cindex registers arguments
3219
3220This section describes the macros which let you control how various
3221types of arguments are passed in registers or how they are arranged in
3222the stack.
3223
3224@hook TARGET_FUNCTION_ARG
3225
3226@hook TARGET_MUST_PASS_IN_STACK
3227
3228@hook TARGET_FUNCTION_INCOMING_ARG
3229
3230@hook TARGET_USE_PSEUDO_PIC_REG
3231
3232@hook TARGET_INIT_PIC_REG
3233
3234@hook TARGET_ARG_PARTIAL_BYTES
3235
3236@hook TARGET_PASS_BY_REFERENCE
3237
3238@hook TARGET_CALLEE_COPIES
3239
3240@defmac CUMULATIVE_ARGS
3241A C type for declaring a variable that is used as the first argument
3242of @code{TARGET_FUNCTION_ARG} and other related values.  For some
3243target machines, the type @code{int} suffices and can hold the number
3244of bytes of argument so far.
3245
3246There is no need to record in @code{CUMULATIVE_ARGS} anything about the
3247arguments that have been passed on the stack.  The compiler has other
3248variables to keep track of that.  For target machines on which all
3249arguments are passed on the stack, there is no need to store anything in
3250@code{CUMULATIVE_ARGS}; however, the data structure must exist and
3251should not be empty, so use @code{int}.
3252@end defmac
3253
3254@defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
3255If defined, this macro is called before generating any code for a
3256function, but after the @var{cfun} descriptor for the function has been
3257created.  The back end may use this macro to update @var{cfun} to
3258reflect an ABI other than that which would normally be used by default.
3259If the compiler is generating code for a compiler-generated function,
3260@var{fndecl} may be @code{NULL}.
3261@end defmac
3262
3263@defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
3264A C statement (sans semicolon) for initializing the variable
3265@var{cum} for the state at the beginning of the argument list.  The
3266variable has type @code{CUMULATIVE_ARGS}.  The value of @var{fntype}
3267is the tree node for the data type of the function which will receive
3268the args, or 0 if the args are to a compiler support library function.
3269For direct calls that are not libcalls, @var{fndecl} contain the
3270declaration node of the function.  @var{fndecl} is also set when
3271@code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
3272being compiled.  @var{n_named_args} is set to the number of named
3273arguments, including a structure return address if it is passed as a
3274parameter, when making a call.  When processing incoming arguments,
3275@var{n_named_args} is set to @minus{}1.
3276
3277When processing a call to a compiler support library function,
3278@var{libname} identifies which one.  It is a @code{symbol_ref} rtx which
3279contains the name of the function, as a string.  @var{libname} is 0 when
3280an ordinary C function call is being processed.  Thus, each time this
3281macro is called, either @var{libname} or @var{fntype} is nonzero, but
3282never both of them at once.
3283@end defmac
3284
3285@defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
3286Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
3287it gets a @code{MODE} argument instead of @var{fntype}, that would be
3288@code{NULL}.  @var{indirect} would always be zero, too.  If this macro
3289is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
32900)} is used instead.
3291@end defmac
3292
3293@defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
3294Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
3295finding the arguments for the function being compiled.  If this macro is
3296undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
3297
3298The value passed for @var{libname} is always 0, since library routines
3299with special calling conventions are never compiled with GCC@.  The
3300argument @var{libname} exists for symmetry with
3301@code{INIT_CUMULATIVE_ARGS}.
3302@c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
3303@c --mew 5feb93   i switched the order of the sentences.  --mew 10feb93
3304@end defmac
3305
3306@hook TARGET_FUNCTION_ARG_ADVANCE
3307
3308@hook TARGET_FUNCTION_ARG_OFFSET
3309
3310@hook TARGET_FUNCTION_ARG_PADDING
3311
3312@defmac PAD_VARARGS_DOWN
3313If defined, a C expression which determines whether the default
3314implementation of va_arg will attempt to pad down before reading the
3315next argument, if that argument is smaller than its aligned space as
3316controlled by @code{PARM_BOUNDARY}.  If this macro is not defined, all such
3317arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
3318@end defmac
3319
3320@defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
3321Specify padding for the last element of a block move between registers and
3322memory.  @var{first} is nonzero if this is the only element.  Defining this
3323macro allows better control of register function parameters on big-endian
3324machines, without using @code{PARALLEL} rtl.  In particular,
3325@code{MUST_PASS_IN_STACK} need not test padding and mode of types in
3326registers, as there is no longer a "wrong" part of a register;  For example,
3327a three byte aggregate may be passed in the high part of a register if so
3328required.
3329@end defmac
3330
3331@hook TARGET_FUNCTION_ARG_BOUNDARY
3332
3333@hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY
3334
3335@defmac FUNCTION_ARG_REGNO_P (@var{regno})
3336A C expression that is nonzero if @var{regno} is the number of a hard
3337register in which function arguments are sometimes passed.  This does
3338@emph{not} include implicit arguments such as the static chain and
3339the structure-value address.  On many machines, no registers can be
3340used for this purpose since all function arguments are pushed on the
3341stack.
3342@end defmac
3343
3344@hook TARGET_SPLIT_COMPLEX_ARG
3345
3346@hook TARGET_BUILD_BUILTIN_VA_LIST
3347
3348@hook TARGET_ENUM_VA_LIST_P
3349
3350@hook TARGET_FN_ABI_VA_LIST
3351
3352@hook TARGET_CANONICAL_VA_LIST_TYPE
3353
3354@hook TARGET_GIMPLIFY_VA_ARG_EXPR
3355
3356@hook TARGET_VALID_POINTER_MODE
3357
3358@hook TARGET_REF_MAY_ALIAS_ERRNO
3359
3360@hook TARGET_TRANSLATE_MODE_ATTRIBUTE
3361
3362@hook TARGET_SCALAR_MODE_SUPPORTED_P
3363
3364@hook TARGET_VECTOR_MODE_SUPPORTED_P
3365
3366@hook TARGET_COMPATIBLE_VECTOR_TYPES_P
3367
3368@hook TARGET_ARRAY_MODE
3369
3370@hook TARGET_ARRAY_MODE_SUPPORTED_P
3371
3372@hook TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P
3373
3374@hook TARGET_FLOATN_MODE
3375
3376@hook TARGET_FLOATN_BUILTIN_P
3377
3378@hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P
3379
3380@node Scalar Return
3381@subsection How Scalar Function Values Are Returned
3382@cindex return values in registers
3383@cindex values, returned by functions
3384@cindex scalars, returned as values
3385
3386This section discusses the macros that control returning scalars as
3387values---values that can fit in registers.
3388
3389@hook TARGET_FUNCTION_VALUE
3390
3391@defmac FUNCTION_VALUE (@var{valtype}, @var{func})
3392This macro has been deprecated.  Use @code{TARGET_FUNCTION_VALUE} for
3393a new target instead.
3394@end defmac
3395
3396@defmac LIBCALL_VALUE (@var{mode})
3397A C expression to create an RTX representing the place where a library
3398function returns a value of mode @var{mode}.
3399
3400Note that ``library function'' in this context means a compiler
3401support routine, used to perform arithmetic, whose name is known
3402specially by the compiler and was not mentioned in the C code being
3403compiled.
3404@end defmac
3405
3406@hook TARGET_LIBCALL_VALUE
3407
3408@defmac FUNCTION_VALUE_REGNO_P (@var{regno})
3409A C expression that is nonzero if @var{regno} is the number of a hard
3410register in which the values of called function may come back.
3411
3412A register whose use for returning values is limited to serving as the
3413second of a pair (for a value of type @code{double}, say) need not be
3414recognized by this macro.  So for most machines, this definition
3415suffices:
3416
3417@smallexample
3418#define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
3419@end smallexample
3420
3421If the machine has register windows, so that the caller and the called
3422function use different registers for the return value, this macro
3423should recognize only the caller's register numbers.
3424
3425This macro has been deprecated.  Use @code{TARGET_FUNCTION_VALUE_REGNO_P}
3426for a new target instead.
3427@end defmac
3428
3429@hook TARGET_FUNCTION_VALUE_REGNO_P
3430
3431@defmac APPLY_RESULT_SIZE
3432Define this macro if @samp{untyped_call} and @samp{untyped_return}
3433need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
3434saving and restoring an arbitrary return value.
3435@end defmac
3436
3437@hook TARGET_OMIT_STRUCT_RETURN_REG
3438
3439@hook TARGET_RETURN_IN_MSB
3440
3441@node Aggregate Return
3442@subsection How Large Values Are Returned
3443@cindex aggregates as return values
3444@cindex large return values
3445@cindex returning aggregate values
3446@cindex structure value address
3447
3448When a function value's mode is @code{BLKmode} (and in some other
3449cases), the value is not returned according to
3450@code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}).  Instead, the
3451caller passes the address of a block of memory in which the value
3452should be stored.  This address is called the @dfn{structure value
3453address}.
3454
3455This section describes how to control returning structure values in
3456memory.
3457
3458@hook TARGET_RETURN_IN_MEMORY
3459
3460@defmac DEFAULT_PCC_STRUCT_RETURN
3461Define this macro to be 1 if all structure and union return values must be
3462in memory.  Since this results in slower code, this should be defined
3463only if needed for compatibility with other compilers or with an ABI@.
3464If you define this macro to be 0, then the conventions used for structure
3465and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
3466target hook.
3467
3468If not defined, this defaults to the value 1.
3469@end defmac
3470
3471@hook TARGET_STRUCT_VALUE_RTX
3472
3473@defmac PCC_STATIC_STRUCT_RETURN
3474Define this macro if the usual system convention on the target machine
3475for returning structures and unions is for the called function to return
3476the address of a static variable containing the value.
3477
3478Do not define this if the usual system convention is for the caller to
3479pass an address to the subroutine.
3480
3481This macro has effect in @option{-fpcc-struct-return} mode, but it does
3482nothing when you use @option{-freg-struct-return} mode.
3483@end defmac
3484
3485@hook TARGET_GET_RAW_RESULT_MODE
3486
3487@hook TARGET_GET_RAW_ARG_MODE
3488
3489@hook TARGET_EMPTY_RECORD_P
3490
3491@hook TARGET_WARN_PARAMETER_PASSING_ABI
3492
3493@node Caller Saves
3494@subsection Caller-Saves Register Allocation
3495
3496If you enable it, GCC can save registers around function calls.  This
3497makes it possible to use call-clobbered registers to hold variables that
3498must live across calls.
3499
3500@defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
3501A C expression specifying which mode is required for saving @var{nregs}
3502of a pseudo-register in call-clobbered hard register @var{regno}.  If
3503@var{regno} is unsuitable for caller save, @code{VOIDmode} should be
3504returned.  For most machines this macro need not be defined since GCC
3505will select the smallest suitable mode.
3506@end defmac
3507
3508@node Function Entry
3509@subsection Function Entry and Exit
3510@cindex function entry and exit
3511@cindex prologue
3512@cindex epilogue
3513
3514This section describes the macros that output function entry
3515(@dfn{prologue}) and exit (@dfn{epilogue}) code.
3516
3517@hook TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY
3518
3519@hook TARGET_ASM_FUNCTION_PROLOGUE
3520
3521@hook TARGET_ASM_FUNCTION_END_PROLOGUE
3522
3523@hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE
3524
3525@hook TARGET_ASM_FUNCTION_EPILOGUE
3526
3527@itemize @bullet
3528@item
3529@findex pretend_args_size
3530@findex crtl->args.pretend_args_size
3531A region of @code{crtl->args.pretend_args_size} bytes of
3532uninitialized space just underneath the first argument arriving on the
3533stack.  (This may not be at the very start of the allocated stack region
3534if the calling sequence has pushed anything else since pushing the stack
3535arguments.  But usually, on such machines, nothing else has been pushed
3536yet, because the function prologue itself does all the pushing.)  This
3537region is used on machines where an argument may be passed partly in
3538registers and partly in memory, and, in some cases to support the
3539features in @code{<stdarg.h>}.
3540
3541@item
3542An area of memory used to save certain registers used by the function.
3543The size of this area, which may also include space for such things as
3544the return address and pointers to previous stack frames, is
3545machine-specific and usually depends on which registers have been used
3546in the function.  Machines with register windows often do not require
3547a save area.
3548
3549@item
3550A region of at least @var{size} bytes, possibly rounded up to an allocation
3551boundary, to contain the local variables of the function.  On some machines,
3552this region and the save area may occur in the opposite order, with the
3553save area closer to the top of the stack.
3554
3555@item
3556@cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
3557Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
3558@code{crtl->outgoing_args_size} bytes to be used for outgoing
3559argument lists of the function.  @xref{Stack Arguments}.
3560@end itemize
3561
3562@defmac EXIT_IGNORE_STACK
3563Define this macro as a C expression that is nonzero if the return
3564instruction or the function epilogue ignores the value of the stack
3565pointer; in other words, if it is safe to delete an instruction to
3566adjust the stack pointer before a return from the function.  The
3567default is 0.
3568
3569Note that this macro's value is relevant only for functions for which
3570frame pointers are maintained.  It is never safe to delete a final
3571stack adjustment in a function that has no frame pointer, and the
3572compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
3573@end defmac
3574
3575@defmac EPILOGUE_USES (@var{regno})
3576Define this macro as a C expression that is nonzero for registers that are
3577used by the epilogue or the @samp{return} pattern.  The stack and frame
3578pointer registers are already assumed to be used as needed.
3579@end defmac
3580
3581@defmac EH_USES (@var{regno})
3582Define this macro as a C expression that is nonzero for registers that are
3583used by the exception handling mechanism, and so should be considered live
3584on entry to an exception edge.
3585@end defmac
3586
3587@hook TARGET_ASM_OUTPUT_MI_THUNK
3588
3589@hook TARGET_ASM_CAN_OUTPUT_MI_THUNK
3590
3591@node Profiling
3592@subsection Generating Code for Profiling
3593@cindex profiling, code generation
3594
3595These macros will help you generate code for profiling.
3596
3597@defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
3598A C statement or compound statement to output to @var{file} some
3599assembler code to call the profiling subroutine @code{mcount}.
3600
3601@findex mcount
3602The details of how @code{mcount} expects to be called are determined by
3603your operating system environment, not by GCC@.  To figure them out,
3604compile a small program for profiling using the system's installed C
3605compiler and look at the assembler code that results.
3606
3607Older implementations of @code{mcount} expect the address of a counter
3608variable to be loaded into some register.  The name of this variable is
3609@samp{LP} followed by the number @var{labelno}, so you would generate
3610the name using @samp{LP%d} in a @code{fprintf}.
3611@end defmac
3612
3613@defmac PROFILE_HOOK
3614A C statement or compound statement to output to @var{file} some assembly
3615code to call the profiling subroutine @code{mcount} even the target does
3616not support profiling.
3617@end defmac
3618
3619@defmac NO_PROFILE_COUNTERS
3620Define this macro to be an expression with a nonzero value if the
3621@code{mcount} subroutine on your system does not need a counter variable
3622allocated for each function.  This is true for almost all modern
3623implementations.  If you define this macro, you must not use the
3624@var{labelno} argument to @code{FUNCTION_PROFILER}.
3625@end defmac
3626
3627@defmac PROFILE_BEFORE_PROLOGUE
3628Define this macro if the code for function profiling should come before
3629the function prologue.  Normally, the profiling code comes after.
3630@end defmac
3631
3632@hook TARGET_KEEP_LEAF_WHEN_PROFILED
3633
3634@node Tail Calls
3635@subsection Permitting tail calls
3636@cindex tail calls
3637
3638@hook TARGET_FUNCTION_OK_FOR_SIBCALL
3639
3640@hook TARGET_EXTRA_LIVE_ON_ENTRY
3641
3642@hook TARGET_SET_UP_BY_PROLOGUE
3643
3644@hook TARGET_WARN_FUNC_RETURN
3645
3646@node Shrink-wrapping separate components
3647@subsection Shrink-wrapping separate components
3648@cindex shrink-wrapping separate components
3649
3650The prologue may perform a variety of target dependent tasks such as
3651saving callee-saved registers, saving the return address, aligning the
3652stack, creating a stack frame, initializing the PIC register, setting
3653up the static chain, etc.
3654
3655On some targets some of these tasks may be independent of others and
3656thus may be shrink-wrapped separately.  These independent tasks are
3657referred to as components and are handled generically by the target
3658independent parts of GCC.
3659
3660Using the following hooks those prologue or epilogue components can be
3661shrink-wrapped separately, so that the initialization (and possibly
3662teardown) those components do is not done as frequently on execution
3663paths where this would unnecessary.
3664
3665What exactly those components are is up to the target code; the generic
3666code treats them abstractly, as a bit in an @code{sbitmap}.  These
3667@code{sbitmap}s are allocated by the @code{shrink_wrap.get_separate_components}
3668and @code{shrink_wrap.components_for_bb} hooks, and deallocated by the
3669generic code.
3670
3671@hook TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS
3672
3673@hook TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB
3674
3675@hook TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS
3676
3677@hook TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS
3678
3679@hook TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS
3680
3681@hook TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS
3682
3683@node Stack Smashing Protection
3684@subsection Stack smashing protection
3685@cindex stack smashing protection
3686
3687@hook TARGET_STACK_PROTECT_GUARD
3688
3689@hook TARGET_STACK_PROTECT_FAIL
3690
3691@hook TARGET_STACK_PROTECT_RUNTIME_ENABLED_P
3692
3693@hook TARGET_SUPPORTS_SPLIT_STACK
3694
3695@hook TARGET_GET_VALID_OPTION_VALUES
3696
3697@node Miscellaneous Register Hooks
3698@subsection Miscellaneous register hooks
3699@cindex miscellaneous register hooks
3700
3701@hook TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS
3702
3703@node Varargs
3704@section Implementing the Varargs Macros
3705@cindex varargs implementation
3706
3707GCC comes with an implementation of @code{<varargs.h>} and
3708@code{<stdarg.h>} that work without change on machines that pass arguments
3709on the stack.  Other machines require their own implementations of
3710varargs, and the two machine independent header files must have
3711conditionals to include it.
3712
3713ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
3714the calling convention for @code{va_start}.  The traditional
3715implementation takes just one argument, which is the variable in which
3716to store the argument pointer.  The ISO implementation of
3717@code{va_start} takes an additional second argument.  The user is
3718supposed to write the last named argument of the function here.
3719
3720However, @code{va_start} should not use this argument.  The way to find
3721the end of the named arguments is with the built-in functions described
3722below.
3723
3724@defmac __builtin_saveregs ()
3725Use this built-in function to save the argument registers in memory so
3726that the varargs mechanism can access them.  Both ISO and traditional
3727versions of @code{va_start} must use @code{__builtin_saveregs}, unless
3728you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
3729
3730On some machines, @code{__builtin_saveregs} is open-coded under the
3731control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}.  On
3732other machines, it calls a routine written in assembler language,
3733found in @file{libgcc2.c}.
3734
3735Code generated for the call to @code{__builtin_saveregs} appears at the
3736beginning of the function, as opposed to where the call to
3737@code{__builtin_saveregs} is written, regardless of what the code is.
3738This is because the registers must be saved before the function starts
3739to use them for its own purposes.
3740@c i rewrote the first sentence above to fix an overfull hbox. --mew
3741@c 10feb93
3742@end defmac
3743
3744@defmac __builtin_next_arg (@var{lastarg})
3745This builtin returns the address of the first anonymous stack
3746argument, as type @code{void *}.  If @code{ARGS_GROW_DOWNWARD}, it
3747returns the address of the location above the first anonymous stack
3748argument.  Use it in @code{va_start} to initialize the pointer for
3749fetching arguments from the stack.  Also use it in @code{va_start} to
3750verify that the second parameter @var{lastarg} is the last named argument
3751of the current function.
3752@end defmac
3753
3754@defmac __builtin_classify_type (@var{object})
3755Since each machine has its own conventions for which data types are
3756passed in which kind of register, your implementation of @code{va_arg}
3757has to embody these conventions.  The easiest way to categorize the
3758specified data type is to use @code{__builtin_classify_type} together
3759with @code{sizeof} and @code{__alignof__}.
3760
3761@code{__builtin_classify_type} ignores the value of @var{object},
3762considering only its data type.  It returns an integer describing what
3763kind of type that is---integer, floating, pointer, structure, and so on.
3764
3765The file @file{typeclass.h} defines an enumeration that you can use to
3766interpret the values of @code{__builtin_classify_type}.
3767@end defmac
3768
3769These machine description macros help implement varargs:
3770
3771@hook TARGET_EXPAND_BUILTIN_SAVEREGS
3772
3773@hook TARGET_SETUP_INCOMING_VARARGS
3774
3775@hook TARGET_STRICT_ARGUMENT_NAMING
3776
3777@hook TARGET_CALL_ARGS
3778
3779@hook TARGET_END_CALL_ARGS
3780
3781@hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED
3782
3783@node Trampolines
3784@section Support for Nested Functions
3785@cindex support for nested functions
3786@cindex trampolines for nested functions
3787@cindex descriptors for nested functions
3788@cindex nested functions, support for
3789
3790Taking the address of a nested function requires special compiler
3791handling to ensure that the static chain register is loaded when
3792the function is invoked via an indirect call.
3793
3794GCC has traditionally supported nested functions by creating an
3795executable @dfn{trampoline} at run time when the address of a nested
3796function is taken.  This is a small piece of code which normally
3797resides on the stack, in the stack frame of the containing function.
3798The trampoline loads the static chain register and then jumps to the
3799real address of the nested function.
3800
3801The use of trampolines requires an executable stack, which is a
3802security risk.  To avoid this problem, GCC also supports another
3803strategy: using descriptors for nested functions.  Under this model,
3804taking the address of a nested function results in a pointer to a
3805non-executable function descriptor object.  Initializing the static chain
3806from the descriptor is handled at indirect call sites.
3807
3808On some targets, including HPPA and IA-64, function descriptors may be
3809mandated by the ABI or be otherwise handled in a target-specific way
3810by the back end in its code generation strategy for indirect calls.
3811GCC also provides its own generic descriptor implementation to support the
3812@option{-fno-trampolines} option.  In this case runtime detection of
3813function descriptors at indirect call sites relies on descriptor
3814pointers being tagged with a bit that is never set in bare function
3815addresses.  Since GCC's generic function descriptors are
3816not ABI-compliant, this option is typically used only on a
3817per-language basis (notably by Ada) or when it can otherwise be
3818applied to the whole program.
3819
3820For languages other than Ada, the @code{-ftrampolines} and
3821@code{-fno-trampolines} options currently have no effect, and
3822trampolines are always generated on platforms that need them
3823for nested functions.
3824
3825Define the following hook if your backend either implements ABI-specified
3826descriptor support, or can use GCC's generic descriptor implementation
3827for nested functions.
3828
3829@hook TARGET_CUSTOM_FUNCTION_DESCRIPTORS
3830
3831The following macros tell GCC how to generate code to allocate and
3832initialize an executable trampoline.  You can also use this interface
3833if your back end needs to create ABI-specified non-executable descriptors; in
3834this case the "trampoline" created is the descriptor containing data only.
3835
3836The instructions in an executable trampoline must do two things: load
3837a constant address into the static chain register, and jump to the real
3838address of the nested function.  On CISC machines such as the m68k,
3839this requires two instructions, a move immediate and a jump.  Then the
3840two addresses exist in the trampoline as word-long immediate operands.
3841On RISC machines, it is often necessary to load each address into a
3842register in two parts.  Then pieces of each address form separate
3843immediate operands.
3844
3845The code generated to initialize the trampoline must store the variable
3846parts---the static chain value and the function address---into the
3847immediate operands of the instructions.  On a CISC machine, this is
3848simply a matter of copying each address to a memory reference at the
3849proper offset from the start of the trampoline.  On a RISC machine, it
3850may be necessary to take out pieces of the address and store them
3851separately.
3852
3853@hook TARGET_ASM_TRAMPOLINE_TEMPLATE
3854
3855@defmac TRAMPOLINE_SECTION
3856Return the section into which the trampoline template is to be placed
3857(@pxref{Sections}).  The default value is @code{readonly_data_section}.
3858@end defmac
3859
3860@defmac TRAMPOLINE_SIZE
3861A C expression for the size in bytes of the trampoline, as an integer.
3862@end defmac
3863
3864@defmac TRAMPOLINE_ALIGNMENT
3865Alignment required for trampolines, in bits.
3866
3867If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
3868is used for aligning trampolines.
3869@end defmac
3870
3871@hook TARGET_TRAMPOLINE_INIT
3872
3873@hook TARGET_EMIT_CALL_BUILTIN___CLEAR_CACHE
3874
3875@hook TARGET_TRAMPOLINE_ADJUST_ADDRESS
3876
3877Implementing trampolines is difficult on many machines because they have
3878separate instruction and data caches.  Writing into a stack location
3879fails to clear the memory in the instruction cache, so when the program
3880jumps to that location, it executes the old contents.
3881
3882Here are two possible solutions.  One is to clear the relevant parts of
3883the instruction cache whenever a trampoline is set up.  The other is to
3884make all trampolines identical, by having them jump to a standard
3885subroutine.  The former technique makes trampoline execution faster; the
3886latter makes initialization faster.
3887
3888To clear the instruction cache when a trampoline is initialized, define
3889the following macro.
3890
3891@defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
3892If defined, expands to a C expression clearing the @emph{instruction
3893cache} in the specified interval.  The definition of this macro would
3894typically be a series of @code{asm} statements.  Both @var{beg} and
3895@var{end} are pointer expressions.
3896@end defmac
3897
3898To use a standard subroutine, define the following macro.  In addition,
3899you must make sure that the instructions in a trampoline fill an entire
3900cache line with identical instructions, or else ensure that the
3901beginning of the trampoline code is always aligned at the same point in
3902its cache line.  Look in @file{m68k.h} as a guide.
3903
3904@defmac TRANSFER_FROM_TRAMPOLINE
3905Define this macro if trampolines need a special subroutine to do their
3906work.  The macro should expand to a series of @code{asm} statements
3907which will be compiled with GCC@.  They go in a library function named
3908@code{__transfer_from_trampoline}.
3909
3910If you need to avoid executing the ordinary prologue code of a compiled
3911C function when you jump to the subroutine, you can do so by placing a
3912special label of your own in the assembler code.  Use one @code{asm}
3913statement to generate an assembler label, and another to make the label
3914global.  Then trampolines can use that label to jump directly to your
3915special assembler code.
3916@end defmac
3917
3918@node Library Calls
3919@section Implicit Calls to Library Routines
3920@cindex library subroutine names
3921@cindex @file{libgcc.a}
3922
3923@c prevent bad page break with this line
3924Here is an explanation of implicit calls to library routines.
3925
3926@defmac DECLARE_LIBRARY_RENAMES
3927This macro, if defined, should expand to a piece of C code that will get
3928expanded when compiling functions for libgcc.a.  It can be used to
3929provide alternate names for GCC's internal library functions if there
3930are ABI-mandated names that the compiler should provide.
3931@end defmac
3932
3933@findex set_optab_libfunc
3934@findex init_one_libfunc
3935@hook TARGET_INIT_LIBFUNCS
3936
3937@hook TARGET_LIBFUNC_GNU_PREFIX
3938
3939@defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
3940This macro should return @code{true} if the library routine that
3941implements the floating point comparison operator @var{comparison} in
3942mode @var{mode} will return a boolean, and @var{false} if it will
3943return a tristate.
3944
3945GCC's own floating point libraries return tristates from the
3946comparison operators, so the default returns false always.  Most ports
3947don't need to define this macro.
3948@end defmac
3949
3950@defmac TARGET_LIB_INT_CMP_BIASED
3951This macro should evaluate to @code{true} if the integer comparison
3952functions (like @code{__cmpdi2}) return 0 to indicate that the first
3953operand is smaller than the second, 1 to indicate that they are equal,
3954and 2 to indicate that the first operand is greater than the second.
3955If this macro evaluates to @code{false} the comparison functions return
3956@minus{}1, 0, and 1 instead of 0, 1, and 2.  If the target uses the routines
3957in @file{libgcc.a}, you do not need to define this macro.
3958@end defmac
3959
3960@defmac TARGET_HAS_NO_HW_DIVIDE
3961This macro should be defined if the target has no hardware divide
3962instructions.  If this macro is defined, GCC will use an algorithm which
3963make use of simple logical and arithmetic operations for 64-bit
3964division.  If the macro is not defined, GCC will use an algorithm which
3965make use of a 64-bit by 32-bit divide primitive.
3966@end defmac
3967
3968@cindex @code{EDOM}, implicit usage
3969@findex matherr
3970@defmac TARGET_EDOM
3971The value of @code{EDOM} on the target machine, as a C integer constant
3972expression.  If you don't define this macro, GCC does not attempt to
3973deposit the value of @code{EDOM} into @code{errno} directly.  Look in
3974@file{/usr/include/errno.h} to find the value of @code{EDOM} on your
3975system.
3976
3977If you do not define @code{TARGET_EDOM}, then compiled code reports
3978domain errors by calling the library function and letting it report the
3979error.  If mathematical functions on your system use @code{matherr} when
3980there is an error, then you should leave @code{TARGET_EDOM} undefined so
3981that @code{matherr} is used normally.
3982@end defmac
3983
3984@cindex @code{errno}, implicit usage
3985@defmac GEN_ERRNO_RTX
3986Define this macro as a C expression to create an rtl expression that
3987refers to the global ``variable'' @code{errno}.  (On certain systems,
3988@code{errno} may not actually be a variable.)  If you don't define this
3989macro, a reasonable default is used.
3990@end defmac
3991
3992@hook TARGET_LIBC_HAS_FUNCTION
3993
3994@hook TARGET_LIBC_HAS_FAST_FUNCTION
3995
3996@defmac NEXT_OBJC_RUNTIME
3997Set this macro to 1 to use the "NeXT" Objective-C message sending conventions
3998by default.  This calling convention involves passing the object, the selector
3999and the method arguments all at once to the method-lookup library function.
4000This is the usual setting when targeting Darwin/Mac OS X systems, which have
4001the NeXT runtime installed.
4002
4003If the macro is set to 0, the "GNU" Objective-C message sending convention
4004will be used by default.  This convention passes just the object and the
4005selector to the method-lookup function, which returns a pointer to the method.
4006
4007In either case, it remains possible to select code-generation for the alternate
4008scheme, by means of compiler command line switches.
4009@end defmac
4010
4011@node Addressing Modes
4012@section Addressing Modes
4013@cindex addressing modes
4014
4015@c prevent bad page break with this line
4016This is about addressing modes.
4017
4018@defmac HAVE_PRE_INCREMENT
4019@defmacx HAVE_PRE_DECREMENT
4020@defmacx HAVE_POST_INCREMENT
4021@defmacx HAVE_POST_DECREMENT
4022A C expression that is nonzero if the machine supports pre-increment,
4023pre-decrement, post-increment, or post-decrement addressing respectively.
4024@end defmac
4025
4026@defmac HAVE_PRE_MODIFY_DISP
4027@defmacx HAVE_POST_MODIFY_DISP
4028A C expression that is nonzero if the machine supports pre- or
4029post-address side-effect generation involving constants other than
4030the size of the memory operand.
4031@end defmac
4032
4033@defmac HAVE_PRE_MODIFY_REG
4034@defmacx HAVE_POST_MODIFY_REG
4035A C expression that is nonzero if the machine supports pre- or
4036post-address side-effect generation involving a register displacement.
4037@end defmac
4038
4039@defmac CONSTANT_ADDRESS_P (@var{x})
4040A C expression that is 1 if the RTX @var{x} is a constant which
4041is a valid address.  On most machines the default definition of
4042@code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)}
4043is acceptable, but a few machines are more restrictive as to which
4044constant addresses are supported.
4045@end defmac
4046
4047@defmac CONSTANT_P (@var{x})
4048@code{CONSTANT_P}, which is defined by target-independent code,
4049accepts integer-values expressions whose values are not explicitly
4050known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
4051expressions and @code{const} arithmetic expressions, in addition to
4052@code{const_int} and @code{const_double} expressions.
4053@end defmac
4054
4055@defmac MAX_REGS_PER_ADDRESS
4056A number, the maximum number of registers that can appear in a valid
4057memory address.  Note that it is up to you to specify a value equal to
4058the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
4059accept.
4060@end defmac
4061
4062@hook TARGET_LEGITIMATE_ADDRESS_P
4063
4064@defmac TARGET_MEM_CONSTRAINT
4065A single character to be used instead of the default @code{'m'}
4066character for general memory addresses.  This defines the constraint
4067letter which matches the memory addresses accepted by
4068@code{TARGET_LEGITIMATE_ADDRESS_P}.  Define this macro if you want to
4069support new address formats in your back end without changing the
4070semantics of the @code{'m'} constraint.  This is necessary in order to
4071preserve functionality of inline assembly constructs using the
4072@code{'m'} constraint.
4073@end defmac
4074
4075@defmac FIND_BASE_TERM (@var{x})
4076A C expression to determine the base term of address @var{x},
4077or to provide a simplified version of @var{x} from which @file{alias.cc}
4078can easily find the base term.  This macro is used in only two places:
4079@code{find_base_value} and @code{find_base_term} in @file{alias.cc}.
4080
4081It is always safe for this macro to not be defined.  It exists so
4082that alias analysis can understand machine-dependent addresses.
4083
4084The typical use of this macro is to handle addresses containing
4085a label_ref or symbol_ref within an UNSPEC@.
4086@end defmac
4087
4088@hook TARGET_LEGITIMIZE_ADDRESS
4089
4090@defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
4091A C compound statement that attempts to replace @var{x}, which is an address
4092that needs reloading, with a valid memory address for an operand of mode
4093@var{mode}.  @var{win} will be a C statement label elsewhere in the code.
4094It is not necessary to define this macro, but it might be useful for
4095performance reasons.
4096
4097For example, on the i386, it is sometimes possible to use a single
4098reload register instead of two by reloading a sum of two pseudo
4099registers into a register.  On the other hand, for number of RISC
4100processors offsets are limited so that often an intermediate address
4101needs to be generated in order to address a stack slot.  By defining
4102@code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
4103generated for adjacent some stack slots can be made identical, and thus
4104be shared.
4105
4106@emph{Note}: This macro should be used with caution.  It is necessary
4107to know something of how reload works in order to effectively use this,
4108and it is quite easy to produce macros that build in too much knowledge
4109of reload internals.
4110
4111@emph{Note}: This macro must be able to reload an address created by a
4112previous invocation of this macro.  If it fails to handle such addresses
4113then the compiler may generate incorrect code or abort.
4114
4115@findex push_reload
4116The macro definition should use @code{push_reload} to indicate parts that
4117need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
4118suitable to be passed unaltered to @code{push_reload}.
4119
4120The code generated by this macro must not alter the substructure of
4121@var{x}.  If it transforms @var{x} into a more legitimate form, it
4122should assign @var{x} (which will always be a C variable) a new value.
4123This also applies to parts that you change indirectly by calling
4124@code{push_reload}.
4125
4126@findex strict_memory_address_p
4127The macro definition may use @code{strict_memory_address_p} to test if
4128the address has become legitimate.
4129
4130@findex copy_rtx
4131If you want to change only a part of @var{x}, one standard way of doing
4132this is to use @code{copy_rtx}.  Note, however, that it unshares only a
4133single level of rtl.  Thus, if the part to be changed is not at the
4134top level, you'll need to replace first the top level.
4135It is not necessary for this macro to come up with a legitimate
4136address;  but often a machine-dependent strategy can generate better code.
4137@end defmac
4138
4139@hook TARGET_MODE_DEPENDENT_ADDRESS_P
4140
4141@hook TARGET_LEGITIMATE_CONSTANT_P
4142
4143@hook TARGET_PRECOMPUTE_TLS_P
4144
4145@hook TARGET_DELEGITIMIZE_ADDRESS
4146
4147@hook TARGET_CONST_NOT_OK_FOR_DEBUG_P
4148
4149@hook TARGET_CANNOT_FORCE_CONST_MEM
4150
4151@hook TARGET_USE_BLOCKS_FOR_CONSTANT_P
4152
4153@hook TARGET_USE_BLOCKS_FOR_DECL_P
4154
4155@hook TARGET_BUILTIN_RECIPROCAL
4156
4157@hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD
4158
4159@hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST
4160
4161@hook TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT
4162
4163@hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE
4164
4165@hook TARGET_VECTORIZE_VEC_PERM_CONST
4166
4167@hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION
4168
4169@hook TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION
4170
4171@hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT
4172
4173@hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE
4174
4175@hook TARGET_VECTORIZE_SPLIT_REDUCTION
4176
4177@hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_MODES
4178
4179@hook TARGET_VECTORIZE_RELATED_MODE
4180
4181@hook TARGET_VECTORIZE_GET_MASK_MODE
4182
4183@hook TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE
4184
4185@hook TARGET_VECTORIZE_CREATE_COSTS
4186
4187@hook TARGET_VECTORIZE_BUILTIN_GATHER
4188
4189@hook TARGET_VECTORIZE_BUILTIN_SCATTER
4190
4191@hook TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN
4192
4193@hook TARGET_SIMD_CLONE_ADJUST
4194
4195@hook TARGET_SIMD_CLONE_USABLE
4196
4197@hook TARGET_SIMT_VF
4198
4199@hook TARGET_OMP_DEVICE_KIND_ARCH_ISA
4200
4201@hook TARGET_GOACC_VALIDATE_DIMS
4202
4203@hook TARGET_GOACC_DIM_LIMIT
4204
4205@hook TARGET_GOACC_FORK_JOIN
4206
4207@hook TARGET_GOACC_REDUCTION
4208
4209@hook TARGET_PREFERRED_ELSE_VALUE
4210
4211@hook TARGET_GOACC_ADJUST_PRIVATE_DECL
4212
4213@hook TARGET_GOACC_EXPAND_VAR_DECL
4214
4215@hook TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD
4216
4217@hook TARGET_GOACC_SHARED_MEM_LAYOUT
4218
4219@node Anchored Addresses
4220@section Anchored Addresses
4221@cindex anchored addresses
4222@cindex @option{-fsection-anchors}
4223
4224GCC usually addresses every static object as a separate entity.
4225For example, if we have:
4226
4227@smallexample
4228static int a, b, c;
4229int foo (void) @{ return a + b + c; @}
4230@end smallexample
4231
4232the code for @code{foo} will usually calculate three separate symbolic
4233addresses: those of @code{a}, @code{b} and @code{c}.  On some targets,
4234it would be better to calculate just one symbolic address and access
4235the three variables relative to it.  The equivalent pseudocode would
4236be something like:
4237
4238@smallexample
4239int foo (void)
4240@{
4241  register int *xr = &x;
4242  return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
4243@}
4244@end smallexample
4245
4246(which isn't valid C).  We refer to shared addresses like @code{x} as
4247``section anchors''.  Their use is controlled by @option{-fsection-anchors}.
4248
4249The hooks below describe the target properties that GCC needs to know
4250in order to make effective use of section anchors.  It won't use
4251section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
4252or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
4253
4254@hook TARGET_MIN_ANCHOR_OFFSET
4255
4256@hook TARGET_MAX_ANCHOR_OFFSET
4257
4258@hook TARGET_ASM_OUTPUT_ANCHOR
4259
4260@hook TARGET_USE_ANCHORS_FOR_SYMBOL_P
4261
4262@node Condition Code
4263@section Condition Code Status
4264@cindex condition code status
4265
4266Condition codes in GCC are represented as registers,
4267which provides better schedulability for
4268architectures that do have a condition code register, but on which
4269most instructions do not affect it.  The latter category includes
4270most RISC machines.
4271
4272Implicit clobbering would pose a strong restriction on the placement of
4273the definition and use of the condition code.  In the past the definition
4274and use were always adjacent.  However, recent changes to support trapping
4275arithmetic may result in the definition and user being in different blocks.
4276Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them.  Additionally,
4277the definition may be the source of exception handling edges.
4278
4279These restrictions can prevent important
4280optimizations on some machines.  For example, on the IBM RS/6000, there
4281is a delay for taken branches unless the condition code register is set
4282three instructions earlier than the conditional branch.  The instruction
4283scheduler cannot perform this optimization if it is not permitted to
4284separate the definition and use of the condition code register.
4285
4286If there is a specific
4287condition code register in the machine, use a hard register.  If the
4288condition code or comparison result can be placed in any general register,
4289or if there are multiple condition registers, use a pseudo register.
4290Registers used to store the condition code value will usually have a mode
4291that is in class @code{MODE_CC}.
4292
4293Alternatively, you can use @code{BImode} if the comparison operator is
4294specified already in the compare instruction.  In this case, you are not
4295interested in most macros in this section.
4296
4297@menu
4298* MODE_CC Condition Codes::  Modern representation of condition codes.
4299@end menu
4300
4301@node MODE_CC Condition Codes
4302@subsection Representation of condition codes using registers
4303@findex CCmode
4304@findex MODE_CC
4305
4306@defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
4307On many machines, the condition code may be produced by other instructions
4308than compares, for example the branch can use directly the condition
4309code set by a subtract instruction.  However, on some machines
4310when the condition code is set this way some bits (such as the overflow
4311bit) are not set in the same way as a test instruction, so that a different
4312branch instruction must be used for some conditional branches.  When
4313this happens, use the machine mode of the condition code register to
4314record different formats of the condition code register.  Modes can
4315also be used to record which compare instruction (e.g.@: a signed or an
4316unsigned comparison) produced the condition codes.
4317
4318If other modes than @code{CCmode} are required, add them to
4319@file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
4320a mode given an operand of a compare.  This is needed because the modes
4321have to be chosen not only during RTL generation but also, for example,
4322by instruction combination.  The result of @code{SELECT_CC_MODE} should
4323be consistent with the mode used in the patterns; for example to support
4324the case of the add on the SPARC discussed above, we have the pattern
4325
4326@smallexample
4327(define_insn ""
4328  [(set (reg:CCNZ 0)
4329        (compare:CCNZ
4330          (plus:SI (match_operand:SI 0 "register_operand" "%r")
4331                   (match_operand:SI 1 "arith_operand" "rI"))
4332          (const_int 0)))]
4333  ""
4334  "@dots{}")
4335@end smallexample
4336
4337@noindent
4338together with a @code{SELECT_CC_MODE} that returns @code{CCNZmode}
4339for comparisons whose argument is a @code{plus}:
4340
4341@smallexample
4342#define SELECT_CC_MODE(OP,X,Y) \
4343  (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT           \
4344   ? ((OP == LT || OP == LE || OP == GT || OP == GE)     \
4345      ? CCFPEmode : CCFPmode)                            \
4346   : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS     \
4347       || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \
4348      ? CCNZmode : CCmode))
4349@end smallexample
4350
4351Another reason to use modes is to retain information on which operands
4352were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
4353this section.
4354
4355You should define this macro if and only if you define extra CC modes
4356in @file{@var{machine}-modes.def}.
4357@end defmac
4358
4359@hook TARGET_CANONICALIZE_COMPARISON
4360
4361@defmac REVERSIBLE_CC_MODE (@var{mode})
4362A C expression whose value is one if it is always safe to reverse a
4363comparison whose mode is @var{mode}.  If @code{SELECT_CC_MODE}
4364can ever return @var{mode} for a floating-point inequality comparison,
4365then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
4366
4367You need not define this macro if it would always returns zero or if the
4368floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
4369For example, here is the definition used on the SPARC, where floating-point
4370inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}:
4371
4372@smallexample
4373#define REVERSIBLE_CC_MODE(MODE) \
4374   ((MODE) != CCFPEmode && (MODE) != CCFPmode)
4375@end smallexample
4376@end defmac
4377
4378@defmac REVERSE_CONDITION (@var{code}, @var{mode})
4379A C expression whose value is reversed condition code of the @var{code} for
4380comparison done in CC_MODE @var{mode}.  The macro is used only in case
4381@code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero.  Define this macro in case
4382machine has some non-standard way how to reverse certain conditionals.  For
4383instance in case all floating point conditions are non-trapping, compiler may
4384freely convert unordered compares to ordered ones.  Then definition may look
4385like:
4386
4387@smallexample
4388#define REVERSE_CONDITION(CODE, MODE) \
4389   ((MODE) != CCFPmode ? reverse_condition (CODE) \
4390    : reverse_condition_maybe_unordered (CODE))
4391@end smallexample
4392@end defmac
4393
4394@hook TARGET_FIXED_CONDITION_CODE_REGS
4395
4396@hook TARGET_CC_MODES_COMPATIBLE
4397
4398@hook TARGET_FLAGS_REGNUM
4399
4400@node Costs
4401@section Describing Relative Costs of Operations
4402@cindex costs of instructions
4403@cindex relative costs
4404@cindex speed of instructions
4405
4406These macros let you describe the relative speed of various operations
4407on the target machine.
4408
4409@defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
4410A C expression for the cost of moving data of mode @var{mode} from a
4411register in class @var{from} to one in class @var{to}.  The classes are
4412expressed using the enumeration values such as @code{GENERAL_REGS}.  A
4413value of 2 is the default; other values are interpreted relative to
4414that.
4415
4416It is not required that the cost always equal 2 when @var{from} is the
4417same as @var{to}; on some machines it is expensive to move between
4418registers if they are not general registers.
4419
4420If reload sees an insn consisting of a single @code{set} between two
4421hard registers, and if @code{REGISTER_MOVE_COST} applied to their
4422classes returns a value of 2, reload does not check to ensure that the
4423constraints of the insn are met.  Setting a cost of other than 2 will
4424allow reload to verify that the constraints are met.  You should do this
4425if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
4426
4427These macros are obsolete, new ports should use the target hook
4428@code{TARGET_REGISTER_MOVE_COST} instead.
4429@end defmac
4430
4431@hook TARGET_REGISTER_MOVE_COST
4432
4433@defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
4434A C expression for the cost of moving data of mode @var{mode} between a
4435register of class @var{class} and memory; @var{in} is zero if the value
4436is to be written to memory, nonzero if it is to be read in.  This cost
4437is relative to those in @code{REGISTER_MOVE_COST}.  If moving between
4438registers and memory is more expensive than between two registers, you
4439should define this macro to express the relative cost.
4440
4441If you do not define this macro, GCC uses a default cost of 4 plus
4442the cost of copying via a secondary reload register, if one is
4443needed.  If your machine requires a secondary reload register to copy
4444between memory and a register of @var{class} but the reload mechanism is
4445more complex than copying via an intermediate, define this macro to
4446reflect the actual cost of the move.
4447
4448GCC defines the function @code{memory_move_secondary_cost} if
4449secondary reloads are needed.  It computes the costs due to copying via
4450a secondary register.  If your machine copies from memory using a
4451secondary register in the conventional way but the default base value of
44524 is not correct for your machine, define this macro to add some other
4453value to the result of that function.  The arguments to that function
4454are the same as to this macro.
4455
4456These macros are obsolete, new ports should use the target hook
4457@code{TARGET_MEMORY_MOVE_COST} instead.
4458@end defmac
4459
4460@hook TARGET_MEMORY_MOVE_COST
4461
4462@defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
4463A C expression for the cost of a branch instruction.  A value of 1 is
4464the default; other values are interpreted relative to that. Parameter
4465@var{speed_p} is true when the branch in question should be optimized
4466for speed.  When it is false, @code{BRANCH_COST} should return a value
4467optimal for code size rather than performance.  @var{predictable_p} is
4468true for well-predicted branches. On many architectures the
4469@code{BRANCH_COST} can be reduced then.
4470@end defmac
4471
4472Here are additional macros which do not specify precise relative costs,
4473but only that certain actions are more expensive than GCC would
4474ordinarily expect.
4475
4476@defmac SLOW_BYTE_ACCESS
4477Define this macro as a C expression which is nonzero if accessing less
4478than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
4479faster than accessing a word of memory, i.e., if such access
4480require more than one instruction or if there is no difference in cost
4481between byte and (aligned) word loads.
4482
4483When this macro is not defined, the compiler will access a field by
4484finding the smallest containing object; when it is defined, a fullword
4485load will be used if alignment permits.  Unless bytes accesses are
4486faster than word accesses, using word accesses is preferable since it
4487may eliminate subsequent memory access if subsequent accesses occur to
4488other fields in the same word of the structure, but to different bytes.
4489@end defmac
4490
4491@hook TARGET_SLOW_UNALIGNED_ACCESS
4492
4493@defmac MOVE_RATIO (@var{speed})
4494The threshold of number of scalar memory-to-memory move insns, @emph{below}
4495which a sequence of insns should be generated instead of a
4496string move insn or a library call.  Increasing the value will always
4497make code faster, but eventually incurs high cost in increased code size.
4498
4499Note that on machines where the corresponding move insn is a
4500@code{define_expand} that emits a sequence of insns, this macro counts
4501the number of such sequences.
4502
4503The parameter @var{speed} is true if the code is currently being
4504optimized for speed rather than size.
4505
4506If you don't define this, a reasonable default is used.
4507@end defmac
4508
4509@hook TARGET_USE_BY_PIECES_INFRASTRUCTURE_P
4510
4511@hook TARGET_OVERLAP_OP_BY_PIECES_P
4512
4513@hook TARGET_COMPARE_BY_PIECES_BRANCH_RATIO
4514
4515@defmac MOVE_MAX_PIECES
4516A C expression used by @code{move_by_pieces} to determine the largest unit
4517a load or store used to copy memory is.  Defaults to @code{MOVE_MAX}.
4518@end defmac
4519
4520@defmac STORE_MAX_PIECES
4521A C expression used by @code{store_by_pieces} to determine the largest unit
4522a store used to memory is.  Defaults to @code{MOVE_MAX_PIECES}, or two times
4523the size of @code{HOST_WIDE_INT}, whichever is smaller.
4524@end defmac
4525
4526@defmac COMPARE_MAX_PIECES
4527A C expression used by @code{compare_by_pieces} to determine the largest unit
4528a load or store used to compare memory is.  Defaults to
4529@code{MOVE_MAX_PIECES}.
4530@end defmac
4531
4532@defmac CLEAR_RATIO (@var{speed})
4533The threshold of number of scalar move insns, @emph{below} which a sequence
4534of insns should be generated to clear memory instead of a string clear insn
4535or a library call.  Increasing the value will always make code faster, but
4536eventually incurs high cost in increased code size.
4537
4538The parameter @var{speed} is true if the code is currently being
4539optimized for speed rather than size.
4540
4541If you don't define this, a reasonable default is used.
4542@end defmac
4543
4544@defmac SET_RATIO (@var{speed})
4545The threshold of number of scalar move insns, @emph{below} which a sequence
4546of insns should be generated to set memory to a constant value, instead of
4547a block set insn or a library call.
4548Increasing the value will always make code faster, but
4549eventually incurs high cost in increased code size.
4550
4551The parameter @var{speed} is true if the code is currently being
4552optimized for speed rather than size.
4553
4554If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
4555@end defmac
4556
4557@defmac USE_LOAD_POST_INCREMENT (@var{mode})
4558A C expression used to determine whether a load postincrement is a good
4559thing to use for a given mode.  Defaults to the value of
4560@code{HAVE_POST_INCREMENT}.
4561@end defmac
4562
4563@defmac USE_LOAD_POST_DECREMENT (@var{mode})
4564A C expression used to determine whether a load postdecrement is a good
4565thing to use for a given mode.  Defaults to the value of
4566@code{HAVE_POST_DECREMENT}.
4567@end defmac
4568
4569@defmac USE_LOAD_PRE_INCREMENT (@var{mode})
4570A C expression used to determine whether a load preincrement is a good
4571thing to use for a given mode.  Defaults to the value of
4572@code{HAVE_PRE_INCREMENT}.
4573@end defmac
4574
4575@defmac USE_LOAD_PRE_DECREMENT (@var{mode})
4576A C expression used to determine whether a load predecrement is a good
4577thing to use for a given mode.  Defaults to the value of
4578@code{HAVE_PRE_DECREMENT}.
4579@end defmac
4580
4581@defmac USE_STORE_POST_INCREMENT (@var{mode})
4582A C expression used to determine whether a store postincrement is a good
4583thing to use for a given mode.  Defaults to the value of
4584@code{HAVE_POST_INCREMENT}.
4585@end defmac
4586
4587@defmac USE_STORE_POST_DECREMENT (@var{mode})
4588A C expression used to determine whether a store postdecrement is a good
4589thing to use for a given mode.  Defaults to the value of
4590@code{HAVE_POST_DECREMENT}.
4591@end defmac
4592
4593@defmac USE_STORE_PRE_INCREMENT (@var{mode})
4594This macro is used to determine whether a store preincrement is a good
4595thing to use for a given mode.  Defaults to the value of
4596@code{HAVE_PRE_INCREMENT}.
4597@end defmac
4598
4599@defmac USE_STORE_PRE_DECREMENT (@var{mode})
4600This macro is used to determine whether a store predecrement is a good
4601thing to use for a given mode.  Defaults to the value of
4602@code{HAVE_PRE_DECREMENT}.
4603@end defmac
4604
4605@defmac NO_FUNCTION_CSE
4606Define this macro to be true if it is as good or better to call a constant
4607function address than to call an address kept in a register.
4608@end defmac
4609
4610@defmac LOGICAL_OP_NON_SHORT_CIRCUIT
4611Define this macro if a non-short-circuit operation produced by
4612@samp{fold_range_test ()} is optimal.  This macro defaults to true if
4613@code{BRANCH_COST} is greater than or equal to the value 2.
4614@end defmac
4615
4616@hook TARGET_OPTAB_SUPPORTED_P
4617
4618@hook TARGET_RTX_COSTS
4619
4620@hook TARGET_ADDRESS_COST
4621
4622@hook TARGET_INSN_COST
4623
4624@hook TARGET_MAX_NOCE_IFCVT_SEQ_COST
4625
4626@hook TARGET_NOCE_CONVERSION_PROFITABLE_P
4627
4628@hook TARGET_NEW_ADDRESS_PROFITABLE_P
4629
4630@hook TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P
4631
4632@hook TARGET_ESTIMATED_POLY_VALUE
4633
4634@node Scheduling
4635@section Adjusting the Instruction Scheduler
4636
4637The instruction scheduler may need a fair amount of machine-specific
4638adjustment in order to produce good code.  GCC provides several target
4639hooks for this purpose.  It is usually enough to define just a few of
4640them: try the first ones in this list first.
4641
4642@hook TARGET_SCHED_ISSUE_RATE
4643
4644@hook TARGET_SCHED_VARIABLE_ISSUE
4645
4646@hook TARGET_SCHED_ADJUST_COST
4647
4648@hook TARGET_SCHED_ADJUST_PRIORITY
4649
4650@hook TARGET_SCHED_REORDER
4651
4652@hook TARGET_SCHED_REORDER2
4653
4654@hook TARGET_SCHED_MACRO_FUSION_P
4655
4656@hook TARGET_SCHED_MACRO_FUSION_PAIR_P
4657
4658@hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK
4659
4660@hook TARGET_SCHED_INIT
4661
4662@hook TARGET_SCHED_FINISH
4663
4664@hook TARGET_SCHED_INIT_GLOBAL
4665
4666@hook TARGET_SCHED_FINISH_GLOBAL
4667
4668@hook TARGET_SCHED_DFA_PRE_CYCLE_INSN
4669
4670@hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN
4671
4672@hook TARGET_SCHED_DFA_POST_CYCLE_INSN
4673
4674@hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN
4675
4676@hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE
4677
4678@hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE
4679
4680@hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD
4681
4682@hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD
4683
4684@hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN
4685
4686@hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE
4687
4688@hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK
4689
4690@hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END
4691
4692@hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT
4693
4694@hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI
4695
4696@hook TARGET_SCHED_DFA_NEW_CYCLE
4697
4698@hook TARGET_SCHED_IS_COSTLY_DEPENDENCE
4699
4700@hook TARGET_SCHED_H_I_D_EXTENDED
4701
4702@hook TARGET_SCHED_ALLOC_SCHED_CONTEXT
4703
4704@hook TARGET_SCHED_INIT_SCHED_CONTEXT
4705
4706@hook TARGET_SCHED_SET_SCHED_CONTEXT
4707
4708@hook TARGET_SCHED_CLEAR_SCHED_CONTEXT
4709
4710@hook TARGET_SCHED_FREE_SCHED_CONTEXT
4711
4712@hook TARGET_SCHED_SPECULATE_INSN
4713
4714@hook TARGET_SCHED_NEEDS_BLOCK_P
4715
4716@hook TARGET_SCHED_GEN_SPEC_CHECK
4717
4718@hook TARGET_SCHED_SET_SCHED_FLAGS
4719
4720@hook TARGET_SCHED_CAN_SPECULATE_INSN
4721
4722@hook TARGET_SCHED_SMS_RES_MII
4723
4724@hook TARGET_SCHED_DISPATCH
4725
4726@hook TARGET_SCHED_DISPATCH_DO
4727
4728@hook TARGET_SCHED_EXPOSED_PIPELINE
4729
4730@hook TARGET_SCHED_REASSOCIATION_WIDTH
4731
4732@hook TARGET_SCHED_FUSION_PRIORITY
4733
4734@hook TARGET_EXPAND_DIVMOD_LIBFUNC
4735
4736@node Sections
4737@section Dividing the Output into Sections (Texts, Data, @dots{})
4738@c the above section title is WAY too long.  maybe cut the part between
4739@c the (...)?  --mew 10feb93
4740
4741An object file is divided into sections containing different types of
4742data.  In the most common case, there are three sections: the @dfn{text
4743section}, which holds instructions and read-only data; the @dfn{data
4744section}, which holds initialized writable data; and the @dfn{bss
4745section}, which holds uninitialized data.  Some systems have other kinds
4746of sections.
4747
4748@file{varasm.cc} provides several well-known sections, such as
4749@code{text_section}, @code{data_section} and @code{bss_section}.
4750The normal way of controlling a @code{@var{foo}_section} variable
4751is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
4752as described below.  The macros are only read once, when @file{varasm.cc}
4753initializes itself, so their values must be run-time constants.
4754They may however depend on command-line flags.
4755
4756@emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
4757use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
4758to be string literals.
4759
4760Some assemblers require a different string to be written every time a
4761section is selected.  If your assembler falls into this category, you
4762should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
4763@code{get_unnamed_section} to set up the sections.
4764
4765You must always create a @code{text_section}, either by defining
4766@code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
4767in @code{TARGET_ASM_INIT_SECTIONS}.  The same is true of
4768@code{data_section} and @code{DATA_SECTION_ASM_OP}.  If you do not
4769create a distinct @code{readonly_data_section}, the default is to
4770reuse @code{text_section}.
4771
4772All the other @file{varasm.cc} sections are optional, and are null
4773if the target does not provide them.
4774
4775@defmac TEXT_SECTION_ASM_OP
4776A C expression whose value is a string, including spacing, containing the
4777assembler operation that should precede instructions and read-only data.
4778Normally @code{"\t.text"} is right.
4779@end defmac
4780
4781@defmac HOT_TEXT_SECTION_NAME
4782If defined, a C string constant for the name of the section containing most
4783frequently executed functions of the program.  If not defined, GCC will provide
4784a default definition if the target supports named sections.
4785@end defmac
4786
4787@defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
4788If defined, a C string constant for the name of the section containing unlikely
4789executed functions in the program.
4790@end defmac
4791
4792@defmac DATA_SECTION_ASM_OP
4793A C expression whose value is a string, including spacing, containing the
4794assembler operation to identify the following data as writable initialized
4795data.  Normally @code{"\t.data"} is right.
4796@end defmac
4797
4798@defmac SDATA_SECTION_ASM_OP
4799If defined, a C expression whose value is a string, including spacing,
4800containing the assembler operation to identify the following data as
4801initialized, writable small data.
4802@end defmac
4803
4804@defmac READONLY_DATA_SECTION_ASM_OP
4805A C expression whose value is a string, including spacing, containing the
4806assembler operation to identify the following data as read-only initialized
4807data.
4808@end defmac
4809
4810@defmac BSS_SECTION_ASM_OP
4811If defined, a C expression whose value is a string, including spacing,
4812containing the assembler operation to identify the following data as
4813uninitialized global data.  If not defined, and
4814@code{ASM_OUTPUT_ALIGNED_BSS} not defined,
4815uninitialized global data will be output in the data section if
4816@option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
4817used.
4818@end defmac
4819
4820@defmac SBSS_SECTION_ASM_OP
4821If defined, a C expression whose value is a string, including spacing,
4822containing the assembler operation to identify the following data as
4823uninitialized, writable small data.
4824@end defmac
4825
4826@defmac TLS_COMMON_ASM_OP
4827If defined, a C expression whose value is a string containing the
4828assembler operation to identify the following data as thread-local
4829common data.  The default is @code{".tls_common"}.
4830@end defmac
4831
4832@defmac TLS_SECTION_ASM_FLAG
4833If defined, a C expression whose value is a character constant
4834containing the flag used to mark a section as a TLS section.  The
4835default is @code{'T'}.
4836@end defmac
4837
4838@defmac INIT_SECTION_ASM_OP
4839If defined, a C expression whose value is a string, including spacing,
4840containing the assembler operation to identify the following data as
4841initialization code.  If not defined, GCC will assume such a section does
4842not exist.  This section has no corresponding @code{init_section}
4843variable; it is used entirely in runtime code.
4844@end defmac
4845
4846@defmac FINI_SECTION_ASM_OP
4847If defined, a C expression whose value is a string, including spacing,
4848containing the assembler operation to identify the following data as
4849finalization code.  If not defined, GCC will assume such a section does
4850not exist.  This section has no corresponding @code{fini_section}
4851variable; it is used entirely in runtime code.
4852@end defmac
4853
4854@defmac INIT_ARRAY_SECTION_ASM_OP
4855If defined, a C expression whose value is a string, including spacing,
4856containing the assembler operation to identify the following data as
4857part of the @code{.init_array} (or equivalent) section.  If not
4858defined, GCC will assume such a section does not exist.  Do not define
4859both this macro and @code{INIT_SECTION_ASM_OP}.
4860@end defmac
4861
4862@defmac FINI_ARRAY_SECTION_ASM_OP
4863If defined, a C expression whose value is a string, including spacing,
4864containing the assembler operation to identify the following data as
4865part of the @code{.fini_array} (or equivalent) section.  If not
4866defined, GCC will assume such a section does not exist.  Do not define
4867both this macro and @code{FINI_SECTION_ASM_OP}.
4868@end defmac
4869
4870@defmac MACH_DEP_SECTION_ASM_FLAG
4871If defined, a C expression whose value is a character constant
4872containing the flag used to mark a machine-dependent section.  This
4873corresponds to the @code{SECTION_MACH_DEP} section flag.
4874@end defmac
4875
4876@defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
4877If defined, an ASM statement that switches to a different section
4878via @var{section_op}, calls @var{function}, and switches back to
4879the text section.  This is used in @file{crtstuff.c} if
4880@code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
4881to initialization and finalization functions from the init and fini
4882sections.  By default, this macro uses a simple function call.  Some
4883ports need hand-crafted assembly code to avoid dependencies on
4884registers initialized in the function prologue or to ensure that
4885constant pools don't end up too far way in the text section.
4886@end defmac
4887
4888@defmac TARGET_LIBGCC_SDATA_SECTION
4889If defined, a string which names the section into which small
4890variables defined in crtstuff and libgcc should go.  This is useful
4891when the target has options for optimizing access to small data, and
4892you want the crtstuff and libgcc routines to be conservative in what
4893they expect of your application yet liberal in what your application
4894expects.  For example, for targets with a @code{.sdata} section (like
4895MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
4896require small data support from your application, but use this macro
4897to put small data into @code{.sdata} so that your application can
4898access these variables whether it uses small data or not.
4899@end defmac
4900
4901@defmac FORCE_CODE_SECTION_ALIGN
4902If defined, an ASM statement that aligns a code section to some
4903arbitrary boundary.  This is used to force all fragments of the
4904@code{.init} and @code{.fini} sections to have to same alignment
4905and thus prevent the linker from having to add any padding.
4906@end defmac
4907
4908@defmac JUMP_TABLES_IN_TEXT_SECTION
4909Define this macro to be an expression with a nonzero value if jump
4910tables (for @code{tablejump} insns) should be output in the text
4911section, along with the assembler instructions.  Otherwise, the
4912readonly data section is used.
4913
4914This macro is irrelevant if there is no separate readonly data section.
4915@end defmac
4916
4917@hook TARGET_ASM_INIT_SECTIONS
4918
4919@hook TARGET_ASM_RELOC_RW_MASK
4920
4921@hook TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC
4922
4923@hook TARGET_ASM_SELECT_SECTION
4924
4925@defmac USE_SELECT_SECTION_FOR_FUNCTIONS
4926Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
4927for @code{FUNCTION_DECL}s as well as for variables and constants.
4928
4929In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
4930function has been determined to be likely to be called, and nonzero if
4931it is unlikely to be called.
4932@end defmac
4933
4934@hook TARGET_ASM_UNIQUE_SECTION
4935
4936@hook TARGET_ASM_FUNCTION_RODATA_SECTION
4937
4938@hook TARGET_ASM_MERGEABLE_RODATA_PREFIX
4939
4940@hook TARGET_ASM_TM_CLONE_TABLE_SECTION
4941
4942@hook TARGET_ASM_SELECT_RTX_SECTION
4943
4944@hook TARGET_MANGLE_DECL_ASSEMBLER_NAME
4945
4946@hook TARGET_ENCODE_SECTION_INFO
4947
4948@hook TARGET_STRIP_NAME_ENCODING
4949
4950@hook TARGET_IN_SMALL_DATA_P
4951
4952@hook TARGET_HAVE_SRODATA_SECTION
4953
4954@hook TARGET_PROFILE_BEFORE_PROLOGUE
4955
4956@hook TARGET_BINDS_LOCAL_P
4957
4958@hook TARGET_HAVE_TLS
4959
4960
4961@node PIC
4962@section Position Independent Code
4963@cindex position independent code
4964@cindex PIC
4965
4966This section describes macros that help implement generation of position
4967independent code.  Simply defining these macros is not enough to
4968generate valid PIC; you must also add support to the hook
4969@code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
4970@code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}.  You
4971must modify the definition of @samp{movsi} to do something appropriate
4972when the source operand contains a symbolic address.  You may also
4973need to alter the handling of switch statements so that they use
4974relative addresses.
4975@c i rearranged the order of the macros above to try to force one of
4976@c them to the next line, to eliminate an overfull hbox. --mew 10feb93
4977
4978@defmac PIC_OFFSET_TABLE_REGNUM
4979The register number of the register used to address a table of static
4980data addresses in memory.  In some cases this register is defined by a
4981processor's ``application binary interface'' (ABI)@.  When this macro
4982is defined, RTL is generated for this register once, as with the stack
4983pointer and frame pointer registers.  If this macro is not defined, it
4984is up to the machine-dependent files to allocate such a register (if
4985necessary).  Note that this register must be fixed when in use (e.g.@:
4986when @code{flag_pic} is true).
4987@end defmac
4988
4989@defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
4990A C expression that is nonzero if the register defined by
4991@code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls.  If not defined,
4992the default is zero.  Do not define
4993this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
4994@end defmac
4995
4996@defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
4997A C expression that is nonzero if @var{x} is a legitimate immediate
4998operand on the target machine when generating position independent code.
4999You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
5000check this.  You can also assume @var{flag_pic} is true, so you need not
5001check it either.  You need not define this macro if all constants
5002(including @code{SYMBOL_REF}) can be immediate operands when generating
5003position independent code.
5004@end defmac
5005
5006@node Assembler Format
5007@section Defining the Output Assembler Language
5008
5009This section describes macros whose principal purpose is to describe how
5010to write instructions in assembler language---rather than what the
5011instructions do.
5012
5013@menu
5014* File Framework::       Structural information for the assembler file.
5015* Data Output::          Output of constants (numbers, strings, addresses).
5016* Uninitialized Data::   Output of uninitialized variables.
5017* Label Output::         Output and generation of labels.
5018* Initialization::       General principles of initialization
5019                         and termination routines.
5020* Macros for Initialization::
5021                         Specific macros that control the handling of
5022                         initialization and termination routines.
5023* Instruction Output::   Output of actual instructions.
5024* Dispatch Tables::      Output of jump tables.
5025* Exception Region Output:: Output of exception region code.
5026* Alignment Output::     Pseudo ops for alignment and skipping data.
5027@end menu
5028
5029@node File Framework
5030@subsection The Overall Framework of an Assembler File
5031@cindex assembler format
5032@cindex output of assembler code
5033
5034@c prevent bad page break with this line
5035This describes the overall framework of an assembly file.
5036
5037@findex default_file_start
5038@hook TARGET_ASM_FILE_START
5039
5040@hook TARGET_ASM_FILE_START_APP_OFF
5041
5042@hook TARGET_ASM_FILE_START_FILE_DIRECTIVE
5043
5044@hook TARGET_ASM_FILE_END
5045
5046@deftypefun void file_end_indicate_exec_stack ()
5047Some systems use a common convention, the @samp{.note.GNU-stack}
5048special section, to indicate whether or not an object file relies on
5049the stack being executable.  If your system uses this convention, you
5050should define @code{TARGET_ASM_FILE_END} to this function.  If you
5051need to do other things in that hook, have your hook function call
5052this function.
5053@end deftypefun
5054
5055@hook TARGET_ASM_LTO_START
5056
5057@hook TARGET_ASM_LTO_END
5058
5059@hook TARGET_ASM_CODE_END
5060
5061@defmac ASM_COMMENT_START
5062A C string constant describing how to begin a comment in the target
5063assembler language.  The compiler assumes that the comment will end at
5064the end of the line.
5065@end defmac
5066
5067@defmac ASM_APP_ON
5068A C string constant for text to be output before each @code{asm}
5069statement or group of consecutive ones.  Normally this is
5070@code{"#APP"}, which is a comment that has no effect on most
5071assemblers but tells the GNU assembler that it must check the lines
5072that follow for all valid assembler constructs.
5073@end defmac
5074
5075@defmac ASM_APP_OFF
5076A C string constant for text to be output after each @code{asm}
5077statement or group of consecutive ones.  Normally this is
5078@code{"#NO_APP"}, which tells the GNU assembler to resume making the
5079time-saving assumptions that are valid for ordinary compiler output.
5080@end defmac
5081
5082@defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
5083A C statement to output COFF information or DWARF debugging information
5084which indicates that filename @var{name} is the current source file to
5085the stdio stream @var{stream}.
5086
5087This macro need not be defined if the standard form of output
5088for the file format in use is appropriate.
5089@end defmac
5090
5091@hook TARGET_ASM_OUTPUT_SOURCE_FILENAME
5092
5093@hook TARGET_ASM_OUTPUT_IDENT
5094
5095@defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
5096A C statement to output the string @var{string} to the stdio stream
5097@var{stream}.  If you do not call the function @code{output_quoted_string}
5098in your config files, GCC will only call it to output filenames to
5099the assembler source.  So you can use it to canonicalize the format
5100of the filename using this macro.
5101@end defmac
5102
5103@hook TARGET_ASM_NAMED_SECTION
5104
5105@hook TARGET_ASM_ELF_FLAGS_NUMERIC
5106
5107@hook TARGET_ASM_FUNCTION_SECTION
5108
5109@hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS
5110
5111@hook TARGET_HAVE_NAMED_SECTIONS
5112This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
5113It must not be modified by command-line option processing.
5114@end deftypevr
5115
5116@anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
5117@hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
5118
5119@hook TARGET_SECTION_TYPE_FLAGS
5120
5121@hook TARGET_ASM_RECORD_GCC_SWITCHES
5122
5123@hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
5124
5125@need 2000
5126@node Data Output
5127@subsection Output of Data
5128
5129
5130@hook TARGET_ASM_BYTE_OP
5131
5132@hook TARGET_ASM_INTEGER
5133
5134@hook TARGET_ASM_DECL_END
5135
5136@hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA
5137
5138@defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
5139A C statement to output to the stdio stream @var{stream} an assembler
5140instruction to assemble a string constant containing the @var{len}
5141bytes at @var{ptr}.  @var{ptr} will be a C expression of type
5142@code{char *} and @var{len} a C expression of type @code{int}.
5143
5144If the assembler has a @code{.ascii} pseudo-op as found in the
5145Berkeley Unix assembler, do not define the macro
5146@code{ASM_OUTPUT_ASCII}.
5147@end defmac
5148
5149@defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
5150A C statement to output word @var{n} of a function descriptor for
5151@var{decl}.  This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
5152is defined, and is otherwise unused.
5153@end defmac
5154
5155@defmac CONSTANT_POOL_BEFORE_FUNCTION
5156You may define this macro as a C expression.  You should define the
5157expression to have a nonzero value if GCC should output the constant
5158pool for a function before the code for the function, or a zero value if
5159GCC should output the constant pool after the function.  If you do
5160not define this macro, the usual case, GCC will output the constant
5161pool before the function.
5162@end defmac
5163
5164@defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
5165A C statement to output assembler commands to define the start of the
5166constant pool for a function.  @var{funname} is a string giving
5167the name of the function.  Should the return type of the function
5168be required, it can be obtained via @var{fundecl}.  @var{size}
5169is the size, in bytes, of the constant pool that will be written
5170immediately after this call.
5171
5172If no constant-pool prefix is required, the usual case, this macro need
5173not be defined.
5174@end defmac
5175
5176@defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
5177A C statement (with or without semicolon) to output a constant in the
5178constant pool, if it needs special treatment.  (This macro need not do
5179anything for RTL expressions that can be output normally.)
5180
5181The argument @var{file} is the standard I/O stream to output the
5182assembler code on.  @var{x} is the RTL expression for the constant to
5183output, and @var{mode} is the machine mode (in case @var{x} is a
5184@samp{const_int}).  @var{align} is the required alignment for the value
5185@var{x}; you should output an assembler directive to force this much
5186alignment.
5187
5188The argument @var{labelno} is a number to use in an internal label for
5189the address of this pool entry.  The definition of this macro is
5190responsible for outputting the label definition at the proper place.
5191Here is how to do this:
5192
5193@smallexample
5194@code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
5195@end smallexample
5196
5197When you output a pool entry specially, you should end with a
5198@code{goto} to the label @var{jumpto}.  This will prevent the same pool
5199entry from being output a second time in the usual manner.
5200
5201You need not define this macro if it would do nothing.
5202@end defmac
5203
5204@defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
5205A C statement to output assembler commands to at the end of the constant
5206pool for a function.  @var{funname} is a string giving the name of the
5207function.  Should the return type of the function be required, you can
5208obtain it via @var{fundecl}.  @var{size} is the size, in bytes, of the
5209constant pool that GCC wrote immediately before this call.
5210
5211If no constant-pool epilogue is required, the usual case, you need not
5212define this macro.
5213@end defmac
5214
5215@defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
5216Define this macro as a C expression which is nonzero if @var{C} is
5217used as a logical line separator by the assembler.  @var{STR} points
5218to the position in the string where @var{C} was found; this can be used if
5219a line separator uses multiple characters.
5220
5221If you do not define this macro, the default is that only
5222the character @samp{;} is treated as a logical line separator.
5223@end defmac
5224
5225@hook TARGET_ASM_OPEN_PAREN
5226
5227These macros are provided by @file{real.h} for writing the definitions
5228of @code{ASM_OUTPUT_DOUBLE} and the like:
5229
5230@defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
5231@defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
5232@defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
5233@defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
5234@defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
5235@defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
5236These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
5237target's floating point representation, and store its bit pattern in
5238the variable @var{l}.  For @code{REAL_VALUE_TO_TARGET_SINGLE} and
5239@code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
5240simple @code{long int}.  For the others, it should be an array of
5241@code{long int}.  The number of elements in this array is determined
5242by the size of the desired target floating point data type: 32 bits of
5243it go in each @code{long int} array element.  Each array element holds
524432 bits of the result, even if @code{long int} is wider than 32 bits
5245on the host machine.
5246
5247The array element values are designed so that you can print them out
5248using @code{fprintf} in the order they should appear in the target
5249machine's memory.
5250@end defmac
5251
5252@node Uninitialized Data
5253@subsection Output of Uninitialized Variables
5254
5255Each of the macros in this section is used to do the whole job of
5256outputting a single uninitialized variable.
5257
5258@defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
5259A C statement (sans semicolon) to output to the stdio stream
5260@var{stream} the assembler definition of a common-label named
5261@var{name} whose size is @var{size} bytes.  The variable @var{rounded}
5262is the size rounded up to whatever alignment the caller wants.  It is
5263possible that @var{size} may be zero, for instance if a struct with no
5264other member than a zero-length array is defined.  In this case, the
5265backend must output a symbol definition that allocates at least one
5266byte, both so that the address of the resulting object does not compare
5267equal to any other, and because some object formats cannot even express
5268the concept of a zero-sized common symbol, as that is how they represent
5269an ordinary undefined external.
5270
5271Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5272output the name itself; before and after that, output the additional
5273assembler syntax for defining the name, and a newline.
5274
5275This macro controls how the assembler definitions of uninitialized
5276common global variables are output.
5277@end defmac
5278
5279@defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
5280Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
5281separate, explicit argument.  If you define this macro, it is used in
5282place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
5283handling the required alignment of the variable.  The alignment is specified
5284as the number of bits.
5285@end defmac
5286
5287@defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5288Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
5289variable to be output, if there is one, or @code{NULL_TREE} if there
5290is no corresponding variable.  If you define this macro, GCC will use it
5291in place of both @code{ASM_OUTPUT_COMMON} and
5292@code{ASM_OUTPUT_ALIGNED_COMMON}.  Define this macro when you need to see
5293the variable's decl in order to chose what to output.
5294@end defmac
5295
5296@defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5297A C statement (sans semicolon) to output to the stdio stream
5298@var{stream} the assembler definition of uninitialized global @var{decl} named
5299@var{name} whose size is @var{size} bytes.  The variable @var{alignment}
5300is the alignment specified as the number of bits.
5301
5302Try to use function @code{asm_output_aligned_bss} defined in file
5303@file{varasm.cc} when defining this macro.  If unable, use the expression
5304@code{assemble_name (@var{stream}, @var{name})} to output the name itself;
5305before and after that, output the additional assembler syntax for defining
5306the name, and a newline.
5307
5308There are two ways of handling global BSS@.  One is to define this macro.
5309The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
5310switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
5311You do not need to do both.
5312
5313Some languages do not have @code{common} data, and require a
5314non-common form of global BSS in order to handle uninitialized globals
5315efficiently.  C++ is one example of this.  However, if the target does
5316not support global BSS, the front end may choose to make globals
5317common in order to save space in the object file.
5318@end defmac
5319
5320@defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
5321A C statement (sans semicolon) to output to the stdio stream
5322@var{stream} the assembler definition of a local-common-label named
5323@var{name} whose size is @var{size} bytes.  The variable @var{rounded}
5324is the size rounded up to whatever alignment the caller wants.
5325
5326Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5327output the name itself; before and after that, output the additional
5328assembler syntax for defining the name, and a newline.
5329
5330This macro controls how the assembler definitions of uninitialized
5331static variables are output.
5332@end defmac
5333
5334@defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
5335Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
5336separate, explicit argument.  If you define this macro, it is used in
5337place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
5338handling the required alignment of the variable.  The alignment is specified
5339as the number of bits.
5340@end defmac
5341
5342@defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
5343Like @code{ASM_OUTPUT_ALIGNED_LOCAL} except that @var{decl} of the
5344variable to be output, if there is one, or @code{NULL_TREE} if there
5345is no corresponding variable.  If you define this macro, GCC will use it
5346in place of both @code{ASM_OUTPUT_LOCAL} and
5347@code{ASM_OUTPUT_ALIGNED_LOCAL}.  Define this macro when you need to see
5348the variable's decl in order to chose what to output.
5349@end defmac
5350
5351@node Label Output
5352@subsection Output and Generation of Labels
5353
5354@c prevent bad page break with this line
5355This is about outputting labels.
5356
5357@findex assemble_name
5358@defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
5359A C statement (sans semicolon) to output to the stdio stream
5360@var{stream} the assembler definition of a label named @var{name}.
5361Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5362output the name itself; before and after that, output the additional
5363assembler syntax for defining the name, and a newline.  A default
5364definition of this macro is provided which is correct for most systems.
5365@end defmac
5366
5367@defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl})
5368A C statement (sans semicolon) to output to the stdio stream
5369@var{stream} the assembler definition of a label named @var{name} of
5370a function.
5371Use the expression @code{assemble_name (@var{stream}, @var{name})} to
5372output the name itself; before and after that, output the additional
5373assembler syntax for defining the name, and a newline.  A default
5374definition of this macro is provided which is correct for most systems.
5375
5376If this macro is not defined, then the function name is defined in the
5377usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5378@end defmac
5379
5380@findex assemble_name_raw
5381@defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
5382Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
5383to refer to a compiler-generated label.  The default definition uses
5384@code{assemble_name_raw}, which is like @code{assemble_name} except
5385that it is more efficient.
5386@end defmac
5387
5388@defmac SIZE_ASM_OP
5389A C string containing the appropriate assembler directive to specify the
5390size of a symbol, without any arguments.  On systems that use ELF, the
5391default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
5392systems, the default is not to define this macro.
5393
5394Define this macro only if it is correct to use the default definitions
5395of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
5396for your system.  If you need your own custom definitions of those
5397macros, or if you do not need explicit symbol sizes at all, do not
5398define this macro.
5399@end defmac
5400
5401@defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
5402A C statement (sans semicolon) to output to the stdio stream
5403@var{stream} a directive telling the assembler that the size of the
5404symbol @var{name} is @var{size}.  @var{size} is a @code{HOST_WIDE_INT}.
5405If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5406provided.
5407@end defmac
5408
5409@defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
5410A C statement (sans semicolon) to output to the stdio stream
5411@var{stream} a directive telling the assembler to calculate the size of
5412the symbol @var{name} by subtracting its address from the current
5413address.
5414
5415If you define @code{SIZE_ASM_OP}, a default definition of this macro is
5416provided.  The default assumes that the assembler recognizes a special
5417@samp{.} symbol as referring to the current address, and can calculate
5418the difference between this and another symbol.  If your assembler does
5419not recognize @samp{.} or cannot do calculations with it, you will need
5420to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
5421@end defmac
5422
5423@defmac NO_DOLLAR_IN_LABEL
5424Define this macro if the assembler does not accept the character
5425@samp{$} in label names.  By default constructors and destructors in
5426G++ have @samp{$} in the identifiers.  If this macro is defined,
5427@samp{.} is used instead.
5428@end defmac
5429
5430@defmac NO_DOT_IN_LABEL
5431Define this macro if the assembler does not accept the character
5432@samp{.} in label names.  By default constructors and destructors in G++
5433have names that use @samp{.}.  If this macro is defined, these names
5434are rewritten to avoid @samp{.}.
5435@end defmac
5436
5437@defmac TYPE_ASM_OP
5438A C string containing the appropriate assembler directive to specify the
5439type of a symbol, without any arguments.  On systems that use ELF, the
5440default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
5441systems, the default is not to define this macro.
5442
5443Define this macro only if it is correct to use the default definition of
5444@code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system.  If you need your own
5445custom definition of this macro, or if you do not need explicit symbol
5446types at all, do not define this macro.
5447@end defmac
5448
5449@defmac TYPE_OPERAND_FMT
5450A C string which specifies (using @code{printf} syntax) the format of
5451the second operand to @code{TYPE_ASM_OP}.  On systems that use ELF, the
5452default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
5453the default is not to define this macro.
5454
5455Define this macro only if it is correct to use the default definition of
5456@code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system.  If you need your own
5457custom definition of this macro, or if you do not need explicit symbol
5458types at all, do not define this macro.
5459@end defmac
5460
5461@defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
5462A C statement (sans semicolon) to output to the stdio stream
5463@var{stream} a directive telling the assembler that the type of the
5464symbol @var{name} is @var{type}.  @var{type} is a C string; currently,
5465that string is always either @samp{"function"} or @samp{"object"}, but
5466you should not count on this.
5467
5468If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
5469definition of this macro is provided.
5470@end defmac
5471
5472@defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5473A C statement (sans semicolon) to output to the stdio stream
5474@var{stream} any text necessary for declaring the name @var{name} of a
5475function which is being defined.  This macro is responsible for
5476outputting the label definition (perhaps using
5477@code{ASM_OUTPUT_FUNCTION_LABEL}).  The argument @var{decl} is the
5478@code{FUNCTION_DECL} tree node representing the function.
5479
5480If this macro is not defined, then the function name is defined in the
5481usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}).
5482
5483You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5484of this macro.
5485@end defmac
5486
5487@defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5488A C statement (sans semicolon) to output to the stdio stream
5489@var{stream} any text necessary for declaring the size of a function
5490which is being defined.  The argument @var{name} is the name of the
5491function.  The argument @var{decl} is the @code{FUNCTION_DECL} tree node
5492representing the function.
5493
5494If this macro is not defined, then the function size is not defined.
5495
5496You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5497of this macro.
5498@end defmac
5499
5500@defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
5501A C statement (sans semicolon) to output to the stdio stream
5502@var{stream} any text necessary for declaring the name @var{name} of a
5503cold function partition which is being defined.  This macro is responsible
5504for outputting the label definition (perhaps using
5505@code{ASM_OUTPUT_FUNCTION_LABEL}).  The argument @var{decl} is the
5506@code{FUNCTION_DECL} tree node representing the function.
5507
5508If this macro is not defined, then the cold partition name is defined in the
5509usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5510
5511You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
5512of this macro.
5513@end defmac
5514
5515@defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
5516A C statement (sans semicolon) to output to the stdio stream
5517@var{stream} any text necessary for declaring the size of a cold function
5518partition which is being defined.  The argument @var{name} is the name of the
5519cold partition of the function.  The argument @var{decl} is the
5520@code{FUNCTION_DECL} tree node representing the function.
5521
5522If this macro is not defined, then the partition size is not defined.
5523
5524You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
5525of this macro.
5526@end defmac
5527
5528@defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
5529A C statement (sans semicolon) to output to the stdio stream
5530@var{stream} any text necessary for declaring the name @var{name} of an
5531initialized variable which is being defined.  This macro must output the
5532label definition (perhaps using @code{ASM_OUTPUT_LABEL}).  The argument
5533@var{decl} is the @code{VAR_DECL} tree node representing the variable.
5534
5535If this macro is not defined, then the variable name is defined in the
5536usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
5537
5538You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
5539@code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
5540@end defmac
5541
5542@hook TARGET_ASM_DECLARE_CONSTANT_NAME
5543
5544@defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
5545A C statement (sans semicolon) to output to the stdio stream
5546@var{stream} any text necessary for claiming a register @var{regno}
5547for a global variable @var{decl} with name @var{name}.
5548
5549If you don't define this macro, that is equivalent to defining it to do
5550nothing.
5551@end defmac
5552
5553@defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
5554A C statement (sans semicolon) to finish up declaring a variable name
5555once the compiler has processed its initializer fully and thus has had a
5556chance to determine the size of an array when controlled by an
5557initializer.  This is used on systems where it's necessary to declare
5558something about the size of the object.
5559
5560If you don't define this macro, that is equivalent to defining it to do
5561nothing.
5562
5563You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
5564@code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
5565@end defmac
5566
5567@hook TARGET_ASM_GLOBALIZE_LABEL
5568
5569@hook TARGET_ASM_GLOBALIZE_DECL_NAME
5570
5571@hook TARGET_ASM_ASSEMBLE_UNDEFINED_DECL
5572
5573@defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
5574A C statement (sans semicolon) to output to the stdio stream
5575@var{stream} some commands that will make the label @var{name} weak;
5576that is, available for reference from other files but only used if
5577no other definition is available.  Use the expression
5578@code{assemble_name (@var{stream}, @var{name})} to output the name
5579itself; before and after that, output the additional assembler syntax
5580for making that name weak, and a newline.
5581
5582If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
5583support weak symbols and you should not define the @code{SUPPORTS_WEAK}
5584macro.
5585@end defmac
5586
5587@defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
5588Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
5589@code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
5590or variable decl.  If @var{value} is not @code{NULL}, this C statement
5591should output to the stdio stream @var{stream} assembler code which
5592defines (equates) the weak symbol @var{name} to have the value
5593@var{value}.  If @var{value} is @code{NULL}, it should output commands
5594to make @var{name} weak.
5595@end defmac
5596
5597@defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
5598Outputs a directive that enables @var{name} to be used to refer to
5599symbol @var{value} with weak-symbol semantics.  @code{decl} is the
5600declaration of @code{name}.
5601@end defmac
5602
5603@defmac SUPPORTS_WEAK
5604A preprocessor constant expression which evaluates to true if the target
5605supports weak symbols.
5606
5607If you don't define this macro, @file{defaults.h} provides a default
5608definition.  If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
5609is defined, the default definition is @samp{1}; otherwise, it is @samp{0}.
5610@end defmac
5611
5612@defmac TARGET_SUPPORTS_WEAK
5613A C expression which evaluates to true if the target supports weak symbols.
5614
5615If you don't define this macro, @file{defaults.h} provides a default
5616definition.  The default definition is @samp{(SUPPORTS_WEAK)}.  Define
5617this macro if you want to control weak symbol support with a compiler
5618flag such as @option{-melf}.
5619@end defmac
5620
5621@defmac MAKE_DECL_ONE_ONLY (@var{decl})
5622A C statement (sans semicolon) to mark @var{decl} to be emitted as a
5623public symbol such that extra copies in multiple translation units will
5624be discarded by the linker.  Define this macro if your object file
5625format provides support for this concept, such as the @samp{COMDAT}
5626section flags in the Microsoft Windows PE/COFF format, and this support
5627requires changes to @var{decl}, such as putting it in a separate section.
5628@end defmac
5629
5630@defmac SUPPORTS_ONE_ONLY
5631A C expression which evaluates to true if the target supports one-only
5632semantics.
5633
5634If you don't define this macro, @file{varasm.cc} provides a default
5635definition.  If @code{MAKE_DECL_ONE_ONLY} is defined, the default
5636definition is @samp{1}; otherwise, it is @samp{0}.  Define this macro if
5637you want to control one-only symbol support with a compiler flag, or if
5638setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
5639be emitted as one-only.
5640@end defmac
5641
5642@hook TARGET_ASM_ASSEMBLE_VISIBILITY
5643
5644@defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
5645A C expression that evaluates to true if the target's linker expects
5646that weak symbols do not appear in a static archive's table of contents.
5647The default is @code{0}.
5648
5649Leaving weak symbols out of an archive's table of contents means that,
5650if a symbol will only have a definition in one translation unit and
5651will have undefined references from other translation units, that
5652symbol should not be weak.  Defining this macro to be nonzero will
5653thus have the effect that certain symbols that would normally be weak
5654(explicit template instantiations, and vtables for polymorphic classes
5655with noninline key methods) will instead be nonweak.
5656
5657The C++ ABI requires this macro to be zero.  Define this macro for
5658targets where full C++ ABI compliance is impossible and where linker
5659restrictions require weak symbols to be left out of a static archive's
5660table of contents.
5661@end defmac
5662
5663@defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
5664A C statement (sans semicolon) to output to the stdio stream
5665@var{stream} any text necessary for declaring the name of an external
5666symbol named @var{name} which is referenced in this compilation but
5667not defined.  The value of @var{decl} is the tree node for the
5668declaration.
5669
5670This macro need not be defined if it does not need to output anything.
5671The GNU assembler and most Unix assemblers don't require anything.
5672@end defmac
5673
5674@hook TARGET_ASM_EXTERNAL_LIBCALL
5675
5676@hook TARGET_ASM_MARK_DECL_PRESERVED
5677
5678@defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
5679A C statement (sans semicolon) to output to the stdio stream
5680@var{stream} a reference in assembler syntax to a label named
5681@var{name}.  This should add @samp{_} to the front of the name, if that
5682is customary on your operating system, as it is in most Berkeley Unix
5683systems.  This macro is used in @code{assemble_name}.
5684@end defmac
5685
5686@hook TARGET_MANGLE_ASSEMBLER_NAME
5687
5688@defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
5689A C statement (sans semicolon) to output a reference to
5690@code{SYMBOL_REF} @var{sym}.  If not defined, @code{assemble_name}
5691will be used to output the name of the symbol.  This macro may be used
5692to modify the way a symbol is referenced depending on information
5693encoded by @code{TARGET_ENCODE_SECTION_INFO}.
5694@end defmac
5695
5696@defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
5697A C statement (sans semicolon) to output a reference to @var{buf}, the
5698result of @code{ASM_GENERATE_INTERNAL_LABEL}.  If not defined,
5699@code{assemble_name} will be used to output the name of the symbol.
5700This macro is not used by @code{output_asm_label}, or the @code{%l}
5701specifier that calls it; the intention is that this macro should be set
5702when it is necessary to output a label differently when its address is
5703being taken.
5704@end defmac
5705
5706@hook TARGET_ASM_INTERNAL_LABEL
5707
5708@defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
5709A C statement to output to the stdio stream @var{stream} a debug info
5710label whose name is made from the string @var{prefix} and the number
5711@var{num}.  This is useful for VLIW targets, where debug info labels
5712may need to be treated differently than branch target labels.  On some
5713systems, branch target labels must be at the beginning of instruction
5714bundles, but debug info labels can occur in the middle of instruction
5715bundles.
5716
5717If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
5718used.
5719@end defmac
5720
5721@defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
5722A C statement to store into the string @var{string} a label whose name
5723is made from the string @var{prefix} and the number @var{num}.
5724
5725This string, when output subsequently by @code{assemble_name}, should
5726produce the output that @code{(*targetm.asm_out.internal_label)} would produce
5727with the same @var{prefix} and @var{num}.
5728
5729If the string begins with @samp{*}, then @code{assemble_name} will
5730output the rest of the string unchanged.  It is often convenient for
5731@code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way.  If the
5732string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
5733to output the string, and may change it.  (Of course,
5734@code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
5735you should know what it does on your machine.)
5736@end defmac
5737
5738@defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
5739A C expression to assign to @var{outvar} (which is a variable of type
5740@code{char *}) a newly allocated string made from the string
5741@var{name} and the number @var{number}, with some suitable punctuation
5742added.  Use @code{alloca} to get space for the string.
5743
5744The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
5745produce an assembler label for an internal static variable whose name is
5746@var{name}.  Therefore, the string must be such as to result in valid
5747assembler code.  The argument @var{number} is different each time this
5748macro is executed; it prevents conflicts between similarly-named
5749internal static variables in different scopes.
5750
5751Ideally this string should not be a valid C identifier, to prevent any
5752conflict with the user's own symbols.  Most assemblers allow periods
5753or percent signs in assembler symbols; putting at least one of these
5754between the name and the number will suffice.
5755
5756If this macro is not defined, a default definition will be provided
5757which is correct for most systems.
5758@end defmac
5759
5760@defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
5761A C statement to output to the stdio stream @var{stream} assembler code
5762which defines (equates) the symbol @var{name} to have the value @var{value}.
5763
5764@findex SET_ASM_OP
5765If @code{SET_ASM_OP} is defined, a default definition is provided which is
5766correct for most systems.
5767@end defmac
5768
5769@defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
5770A C statement to output to the stdio stream @var{stream} assembler code
5771which defines (equates) the symbol whose tree node is @var{decl_of_name}
5772to have the value of the tree node @var{decl_of_value}.  This macro will
5773be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
5774the tree nodes are available.
5775
5776@findex SET_ASM_OP
5777If @code{SET_ASM_OP} is defined, a default definition is provided which is
5778correct for most systems.
5779@end defmac
5780
5781@defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
5782A C statement that evaluates to true if the assembler code which defines
5783(equates) the symbol whose tree node is @var{decl_of_name} to have the value
5784of the tree node @var{decl_of_value} should be emitted near the end of the
5785current compilation unit.  The default is to not defer output of defines.
5786This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
5787@samp{ASM_OUTPUT_DEF_FROM_DECLS}.
5788@end defmac
5789
5790@defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
5791A C statement to output to the stdio stream @var{stream} assembler code
5792which defines (equates) the weak symbol @var{name} to have the value
5793@var{value}.  If @var{value} is @code{NULL}, it defines @var{name} as
5794an undefined weak symbol.
5795
5796Define this macro if the target only supports weak aliases; define
5797@code{ASM_OUTPUT_DEF} instead if possible.
5798@end defmac
5799
5800@defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
5801Define this macro to override the default assembler names used for
5802Objective-C methods.
5803
5804The default name is a unique method number followed by the name of the
5805class (e.g.@: @samp{_1_Foo}).  For methods in categories, the name of
5806the category is also included in the assembler name (e.g.@:
5807@samp{_1_Foo_Bar}).
5808
5809These names are safe on most systems, but make debugging difficult since
5810the method's selector is not present in the name.  Therefore, particular
5811systems define other ways of computing names.
5812
5813@var{buf} is an expression of type @code{char *} which gives you a
5814buffer in which to store the name; its length is as long as
5815@var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
581650 characters extra.
5817
5818The argument @var{is_inst} specifies whether the method is an instance
5819method or a class method; @var{class_name} is the name of the class;
5820@var{cat_name} is the name of the category (or @code{NULL} if the method is not
5821in a category); and @var{sel_name} is the name of the selector.
5822
5823On systems where the assembler can handle quoted names, you can use this
5824macro to provide more human-readable names.
5825@end defmac
5826
5827@node Initialization
5828@subsection How Initialization Functions Are Handled
5829@cindex initialization routines
5830@cindex termination routines
5831@cindex constructors, output of
5832@cindex destructors, output of
5833
5834The compiled code for certain languages includes @dfn{constructors}
5835(also called @dfn{initialization routines})---functions to initialize
5836data in the program when the program is started.  These functions need
5837to be called before the program is ``started''---that is to say, before
5838@code{main} is called.
5839
5840Compiling some languages generates @dfn{destructors} (also called
5841@dfn{termination routines}) that should be called when the program
5842terminates.
5843
5844To make the initialization and termination functions work, the compiler
5845must output something in the assembler code to cause those functions to
5846be called at the appropriate time.  When you port the compiler to a new
5847system, you need to specify how to do this.
5848
5849There are two major ways that GCC currently supports the execution of
5850initialization and termination functions.  Each way has two variants.
5851Much of the structure is common to all four variations.
5852
5853@findex __CTOR_LIST__
5854@findex __DTOR_LIST__
5855The linker must build two lists of these functions---a list of
5856initialization functions, called @code{__CTOR_LIST__}, and a list of
5857termination functions, called @code{__DTOR_LIST__}.
5858
5859Each list always begins with an ignored function pointer (which may hold
58600, @minus{}1, or a count of the function pointers after it, depending on
5861the environment).  This is followed by a series of zero or more function
5862pointers to constructors (or destructors), followed by a function
5863pointer containing zero.
5864
5865Depending on the operating system and its executable file format, either
5866@file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
5867time and exit time.  Constructors are called in reverse order of the
5868list; destructors in forward order.
5869
5870The best way to handle static constructors works only for object file
5871formats which provide arbitrarily-named sections.  A section is set
5872aside for a list of constructors, and another for a list of destructors.
5873Traditionally these are called @samp{.ctors} and @samp{.dtors}.  Each
5874object file that defines an initialization function also puts a word in
5875the constructor section to point to that function.  The linker
5876accumulates all these words into one contiguous @samp{.ctors} section.
5877Termination functions are handled similarly.
5878
5879This method will be chosen as the default by @file{target-def.h} if
5880@code{TARGET_ASM_NAMED_SECTION} is defined.  A target that does not
5881support arbitrary sections, but does support special designated
5882constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
5883and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
5884
5885When arbitrary sections are available, there are two variants, depending
5886upon how the code in @file{crtstuff.c} is called.  On systems that
5887support a @dfn{.init} section which is executed at program startup,
5888parts of @file{crtstuff.c} are compiled into that section.  The
5889program is linked by the @command{gcc} driver like this:
5890
5891@smallexample
5892ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
5893@end smallexample
5894
5895The prologue of a function (@code{__init}) appears in the @code{.init}
5896section of @file{crti.o}; the epilogue appears in @file{crtn.o}.  Likewise
5897for the function @code{__fini} in the @dfn{.fini} section.  Normally these
5898files are provided by the operating system or by the GNU C library, but
5899are provided by GCC for a few targets.
5900
5901The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
5902compiled from @file{crtstuff.c}.  They contain, among other things, code
5903fragments within the @code{.init} and @code{.fini} sections that branch
5904to routines in the @code{.text} section.  The linker will pull all parts
5905of a section together, which results in a complete @code{__init} function
5906that invokes the routines we need at startup.
5907
5908To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
5909macro properly.
5910
5911If no init section is available, when GCC compiles any function called
5912@code{main} (or more accurately, any function designated as a program
5913entry point by the language front end calling @code{expand_main_function}),
5914it inserts a procedure call to @code{__main} as the first executable code
5915after the function prologue.  The @code{__main} function is defined
5916in @file{libgcc2.c} and runs the global constructors.
5917
5918In file formats that don't support arbitrary sections, there are again
5919two variants.  In the simplest variant, the GNU linker (GNU @code{ld})
5920and an `a.out' format must be used.  In this case,
5921@code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
5922entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
5923and with the address of the void function containing the initialization
5924code as its value.  The GNU linker recognizes this as a request to add
5925the value to a @dfn{set}; the values are accumulated, and are eventually
5926placed in the executable as a vector in the format described above, with
5927a leading (ignored) count and a trailing zero element.
5928@code{TARGET_ASM_DESTRUCTOR} is handled similarly.  Since no init
5929section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
5930the compilation of @code{main} to call @code{__main} as above, starting
5931the initialization process.
5932
5933The last variant uses neither arbitrary sections nor the GNU linker.
5934This is preferable when you want to do dynamic linking and when using
5935file formats which the GNU linker does not support, such as `ECOFF'@.  In
5936this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
5937termination functions are recognized simply by their names.  This requires
5938an extra program in the linkage step, called @command{collect2}.  This program
5939pretends to be the linker, for use with GCC; it does its job by running
5940the ordinary linker, but also arranges to include the vectors of
5941initialization and termination functions.  These functions are called
5942via @code{__main} as described above.  In order to use this method,
5943@code{use_collect2} must be defined in the target in @file{config.gcc}.
5944
5945@ifinfo
5946The following section describes the specific macros that control and
5947customize the handling of initialization and termination functions.
5948@end ifinfo
5949
5950@node Macros for Initialization
5951@subsection Macros Controlling Initialization Routines
5952
5953Here are the macros that control how the compiler handles initialization
5954and termination functions:
5955
5956@defmac INIT_SECTION_ASM_OP
5957If defined, a C string constant, including spacing, for the assembler
5958operation to identify the following data as initialization code.  If not
5959defined, GCC will assume such a section does not exist.  When you are
5960using special sections for initialization and termination functions, this
5961macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
5962run the initialization functions.
5963@end defmac
5964
5965@defmac HAS_INIT_SECTION
5966If defined, @code{main} will not call @code{__main} as described above.
5967This macro should be defined for systems that control start-up code
5968on a symbol-by-symbol basis, such as OSF/1, and should not
5969be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
5970@end defmac
5971
5972@defmac LD_INIT_SWITCH
5973If defined, a C string constant for a switch that tells the linker that
5974the following symbol is an initialization routine.
5975@end defmac
5976
5977@defmac LD_FINI_SWITCH
5978If defined, a C string constant for a switch that tells the linker that
5979the following symbol is a finalization routine.
5980@end defmac
5981
5982@defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
5983If defined, a C statement that will write a function that can be
5984automatically called when a shared library is loaded.  The function
5985should call @var{func}, which takes no arguments.  If not defined, and
5986the object format requires an explicit initialization function, then a
5987function called @code{_GLOBAL__DI} will be generated.
5988
5989This function and the following one are used by collect2 when linking a
5990shared library that needs constructors or destructors, or has DWARF2
5991exception tables embedded in the code.
5992@end defmac
5993
5994@defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
5995If defined, a C statement that will write a function that can be
5996automatically called when a shared library is unloaded.  The function
5997should call @var{func}, which takes no arguments.  If not defined, and
5998the object format requires an explicit finalization function, then a
5999function called @code{_GLOBAL__DD} will be generated.
6000@end defmac
6001
6002@defmac INVOKE__main
6003If defined, @code{main} will call @code{__main} despite the presence of
6004@code{INIT_SECTION_ASM_OP}.  This macro should be defined for systems
6005where the init section is not actually run automatically, but is still
6006useful for collecting the lists of constructors and destructors.
6007@end defmac
6008
6009@defmac SUPPORTS_INIT_PRIORITY
6010If nonzero, the C++ @code{init_priority} attribute is supported and the
6011compiler should emit instructions to control the order of initialization
6012of objects.  If zero, the compiler will issue an error message upon
6013encountering an @code{init_priority} attribute.
6014@end defmac
6015
6016@hook TARGET_HAVE_CTORS_DTORS
6017
6018@hook TARGET_DTORS_FROM_CXA_ATEXIT
6019
6020@hook TARGET_ASM_CONSTRUCTOR
6021
6022@hook TARGET_ASM_DESTRUCTOR
6023
6024If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
6025generated for the generated object file will have static linkage.
6026
6027If your system uses @command{collect2} as the means of processing
6028constructors, then that program normally uses @command{nm} to scan
6029an object file for constructor functions to be called.
6030
6031On certain kinds of systems, you can define this macro to make
6032@command{collect2} work faster (and, in some cases, make it work at all):
6033
6034@defmac OBJECT_FORMAT_COFF
6035Define this macro if the system uses COFF (Common Object File Format)
6036object files, so that @command{collect2} can assume this format and scan
6037object files directly for dynamic constructor/destructor functions.
6038
6039This macro is effective only in a native compiler; @command{collect2} as
6040part of a cross compiler always uses @command{nm} for the target machine.
6041@end defmac
6042
6043@defmac REAL_NM_FILE_NAME
6044Define this macro as a C string constant containing the file name to use
6045to execute @command{nm}.  The default is to search the path normally for
6046@command{nm}.
6047@end defmac
6048
6049@defmac NM_FLAGS
6050@command{collect2} calls @command{nm} to scan object files for static
6051constructors and destructors and LTO info.  By default, @option{-n} is
6052passed.  Define @code{NM_FLAGS} to a C string constant if other options
6053are needed to get the same output format as GNU @command{nm -n}
6054produces.
6055@end defmac
6056
6057If your system supports shared libraries and has a program to list the
6058dynamic dependencies of a given library or executable, you can define
6059these macros to enable support for running initialization and
6060termination functions in shared libraries:
6061
6062@defmac LDD_SUFFIX
6063Define this macro to a C string constant containing the name of the program
6064which lists dynamic dependencies, like @command{ldd} under SunOS 4.
6065@end defmac
6066
6067@defmac PARSE_LDD_OUTPUT (@var{ptr})
6068Define this macro to be C code that extracts filenames from the output
6069of the program denoted by @code{LDD_SUFFIX}.  @var{ptr} is a variable
6070of type @code{char *} that points to the beginning of a line of output
6071from @code{LDD_SUFFIX}.  If the line lists a dynamic dependency, the
6072code must advance @var{ptr} to the beginning of the filename on that
6073line.  Otherwise, it must set @var{ptr} to @code{NULL}.
6074@end defmac
6075
6076@defmac SHLIB_SUFFIX
6077Define this macro to a C string constant containing the default shared
6078library extension of the target (e.g., @samp{".so"}).  @command{collect2}
6079strips version information after this suffix when generating global
6080constructor and destructor names.  This define is only needed on targets
6081that use @command{collect2} to process constructors and destructors.
6082@end defmac
6083
6084@node Instruction Output
6085@subsection Output of Assembler Instructions
6086
6087@c prevent bad page break with this line
6088This describes assembler instruction output.
6089
6090@defmac REGISTER_NAMES
6091A C initializer containing the assembler's names for the machine
6092registers, each one as a C string constant.  This is what translates
6093register numbers in the compiler into assembler language.
6094@end defmac
6095
6096@defmac ADDITIONAL_REGISTER_NAMES
6097If defined, a C initializer for an array of structures containing a name
6098and a register number.  This macro defines additional names for hard
6099registers, thus allowing the @code{asm} option in declarations to refer
6100to registers using alternate names.
6101@end defmac
6102
6103@defmac OVERLAPPING_REGISTER_NAMES
6104If defined, a C initializer for an array of structures containing a
6105name, a register number and a count of the number of consecutive
6106machine registers the name overlaps.  This macro defines additional
6107names for hard registers, thus allowing the @code{asm} option in
6108declarations to refer to registers using alternate names.  Unlike
6109@code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the
6110register name implies multiple underlying registers.
6111
6112This macro should be used when it is important that a clobber in an
6113@code{asm} statement clobbers all the underlying values implied by the
6114register name.  For example, on ARM, clobbering the double-precision
6115VFP register ``d0'' implies clobbering both single-precision registers
6116``s0'' and ``s1''.
6117@end defmac
6118
6119@defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
6120Define this macro if you are using an unusual assembler that
6121requires different names for the machine instructions.
6122
6123The definition is a C statement or statements which output an
6124assembler instruction opcode to the stdio stream @var{stream}.  The
6125macro-operand @var{ptr} is a variable of type @code{char *} which
6126points to the opcode name in its ``internal'' form---the form that is
6127written in the machine description.  The definition should output the
6128opcode name to @var{stream}, performing any translation you desire, and
6129increment the variable @var{ptr} to point at the end of the opcode
6130so that it will not be output twice.
6131
6132In fact, your macro definition may process less than the entire opcode
6133name, or more than the opcode name; but if you want to process text
6134that includes @samp{%}-sequences to substitute operands, you must take
6135care of the substitution yourself.  Just be sure to increment
6136@var{ptr} over whatever text should not be output normally.
6137
6138@findex recog_data.operand
6139If you need to look at the operand values, they can be found as the
6140elements of @code{recog_data.operand}.
6141
6142If the macro definition does nothing, the instruction is output
6143in the usual way.
6144@end defmac
6145
6146@defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
6147If defined, a C statement to be executed just prior to the output of
6148assembler code for @var{insn}, to modify the extracted operands so
6149they will be output differently.
6150
6151Here the argument @var{opvec} is the vector containing the operands
6152extracted from @var{insn}, and @var{noperands} is the number of
6153elements of the vector which contain meaningful data for this insn.
6154The contents of this vector are what will be used to convert the insn
6155template into assembler code, so you can change the assembler output
6156by changing the contents of the vector.
6157
6158This macro is useful when various assembler syntaxes share a single
6159file of instruction patterns; by defining this macro differently, you
6160can cause a large class of instructions to be output differently (such
6161as with rearranged operands).  Naturally, variations in assembler
6162syntax affecting individual insn patterns ought to be handled by
6163writing conditional output routines in those patterns.
6164
6165If this macro is not defined, it is equivalent to a null statement.
6166@end defmac
6167
6168@hook TARGET_ASM_FINAL_POSTSCAN_INSN
6169
6170@defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
6171A C compound statement to output to stdio stream @var{stream} the
6172assembler syntax for an instruction operand @var{x}.  @var{x} is an
6173RTL expression.
6174
6175@var{code} is a value that can be used to specify one of several ways
6176of printing the operand.  It is used when identical operands must be
6177printed differently depending on the context.  @var{code} comes from
6178the @samp{%} specification that was used to request printing of the
6179operand.  If the specification was just @samp{%@var{digit}} then
6180@var{code} is 0; if the specification was @samp{%@var{ltr}
6181@var{digit}} then @var{code} is the ASCII code for @var{ltr}.
6182
6183@findex reg_names
6184If @var{x} is a register, this macro should print the register's name.
6185The names can be found in an array @code{reg_names} whose type is
6186@code{char *[]}.  @code{reg_names} is initialized from
6187@code{REGISTER_NAMES}.
6188
6189When the machine description has a specification @samp{%@var{punct}}
6190(a @samp{%} followed by a punctuation character), this macro is called
6191with a null pointer for @var{x} and the punctuation character for
6192@var{code}.
6193@end defmac
6194
6195@defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
6196A C expression which evaluates to true if @var{code} is a valid
6197punctuation character for use in the @code{PRINT_OPERAND} macro.  If
6198@code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
6199punctuation characters (except for the standard one, @samp{%}) are used
6200in this way.
6201@end defmac
6202
6203@defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
6204A C compound statement to output to stdio stream @var{stream} the
6205assembler syntax for an instruction operand that is a memory reference
6206whose address is @var{x}.  @var{x} is an RTL expression.
6207
6208@cindex @code{TARGET_ENCODE_SECTION_INFO} usage
6209On some machines, the syntax for a symbolic address depends on the
6210section that the address refers to.  On these machines, define the hook
6211@code{TARGET_ENCODE_SECTION_INFO} to store the information into the
6212@code{symbol_ref}, and then check for it here.  @xref{Assembler
6213Format}.
6214@end defmac
6215
6216@findex dbr_sequence_length
6217@defmac DBR_OUTPUT_SEQEND (@var{file})
6218A C statement, to be executed after all slot-filler instructions have
6219been output.  If necessary, call @code{dbr_sequence_length} to
6220determine the number of slots filled in a sequence (zero if not
6221currently outputting a sequence), to decide how many no-ops to output,
6222or whatever.
6223
6224Don't define this macro if it has nothing to do, but it is helpful in
6225reading assembly output if the extent of the delay sequence is made
6226explicit (e.g.@: with white space).
6227@end defmac
6228
6229@findex final_sequence
6230Note that output routines for instructions with delay slots must be
6231prepared to deal with not being output as part of a sequence
6232(i.e.@: when the scheduling pass is not run, or when no slot fillers could be
6233found.)  The variable @code{final_sequence} is null when not
6234processing a sequence, otherwise it contains the @code{sequence} rtx
6235being output.
6236
6237@findex asm_fprintf
6238@defmac REGISTER_PREFIX
6239@defmacx LOCAL_LABEL_PREFIX
6240@defmacx USER_LABEL_PREFIX
6241@defmacx IMMEDIATE_PREFIX
6242If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
6243@samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
6244@file{final.cc}).  These are useful when a single @file{md} file must
6245support multiple assembler formats.  In that case, the various @file{tm.h}
6246files can define these macros differently.
6247@end defmac
6248
6249@defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
6250If defined this macro should expand to a series of @code{case}
6251statements which will be parsed inside the @code{switch} statement of
6252the @code{asm_fprintf} function.  This allows targets to define extra
6253printf formats which may useful when generating their assembler
6254statements.  Note that uppercase letters are reserved for future
6255generic extensions to asm_fprintf, and so are not available to target
6256specific code.  The output file is given by the parameter @var{file}.
6257The varargs input pointer is @var{argptr} and the rest of the format
6258string, starting the character after the one that is being switched
6259upon, is pointed to by @var{format}.
6260@end defmac
6261
6262@defmac ASSEMBLER_DIALECT
6263If your target supports multiple dialects of assembler language (such as
6264different opcodes), define this macro as a C expression that gives the
6265numeric index of the assembler language dialect to use, with zero as the
6266first variant.
6267
6268If this macro is defined, you may use constructs of the form
6269@smallexample
6270@samp{@{option0|option1|option2@dots{}@}}
6271@end smallexample
6272@noindent
6273in the output templates of patterns (@pxref{Output Template}) or in the
6274first argument of @code{asm_fprintf}.  This construct outputs
6275@samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
6276@code{ASSEMBLER_DIALECT} is zero, one, two, etc.  Any special characters
6277within these strings retain their usual meaning.  If there are fewer
6278alternatives within the braces than the value of
6279@code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed
6280to print curly braces or @samp{|} character in assembler output directly,
6281@samp{%@{}, @samp{%@}} and @samp{%|} can be used.
6282
6283If you do not define this macro, the characters @samp{@{}, @samp{|} and
6284@samp{@}} do not have any special meaning when used in templates or
6285operands to @code{asm_fprintf}.
6286
6287Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
6288@code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
6289the variations in assembler language syntax with that mechanism.  Define
6290@code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
6291if the syntax variant are larger and involve such things as different
6292opcodes or operand order.
6293@end defmac
6294
6295@defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
6296A C expression to output to @var{stream} some assembler code
6297which will push hard register number @var{regno} onto the stack.
6298The code need not be optimal, since this macro is used only when
6299profiling.
6300@end defmac
6301
6302@defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
6303A C expression to output to @var{stream} some assembler code
6304which will pop hard register number @var{regno} off of the stack.
6305The code need not be optimal, since this macro is used only when
6306profiling.
6307@end defmac
6308
6309@node Dispatch Tables
6310@subsection Output of Dispatch Tables
6311
6312@c prevent bad page break with this line
6313This concerns dispatch tables.
6314
6315@cindex dispatch table
6316@defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
6317A C statement to output to the stdio stream @var{stream} an assembler
6318pseudo-instruction to generate a difference between two labels.
6319@var{value} and @var{rel} are the numbers of two internal labels.  The
6320definitions of these labels are output using
6321@code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
6322way here.  For example,
6323
6324@smallexample
6325fprintf (@var{stream}, "\t.word L%d-L%d\n",
6326         @var{value}, @var{rel})
6327@end smallexample
6328
6329You must provide this macro on machines where the addresses in a
6330dispatch table are relative to the table's own address.  If defined, GCC
6331will also use this macro on all machines when producing PIC@.
6332@var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
6333mode and flags can be read.
6334@end defmac
6335
6336@defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
6337This macro should be provided on machines where the addresses
6338in a dispatch table are absolute.
6339
6340The definition should be a C statement to output to the stdio stream
6341@var{stream} an assembler pseudo-instruction to generate a reference to
6342a label.  @var{value} is the number of an internal label whose
6343definition is output using @code{(*targetm.asm_out.internal_label)}.
6344For example,
6345
6346@smallexample
6347fprintf (@var{stream}, "\t.word L%d\n", @var{value})
6348@end smallexample
6349@end defmac
6350
6351@defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
6352Define this if the label before a jump-table needs to be output
6353specially.  The first three arguments are the same as for
6354@code{(*targetm.asm_out.internal_label)}; the fourth argument is the
6355jump-table which follows (a @code{jump_table_data} containing an
6356@code{addr_vec} or @code{addr_diff_vec}).
6357
6358This feature is used on system V to output a @code{swbeg} statement
6359for the table.
6360
6361If this macro is not defined, these labels are output with
6362@code{(*targetm.asm_out.internal_label)}.
6363@end defmac
6364
6365@defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
6366Define this if something special must be output at the end of a
6367jump-table.  The definition should be a C statement to be executed
6368after the assembler code for the table is written.  It should write
6369the appropriate code to stdio stream @var{stream}.  The argument
6370@var{table} is the jump-table insn, and @var{num} is the label-number
6371of the preceding label.
6372
6373If this macro is not defined, nothing special is output at the end of
6374the jump-table.
6375@end defmac
6376
6377@hook TARGET_ASM_POST_CFI_STARTPROC
6378
6379@hook TARGET_ASM_EMIT_UNWIND_LABEL
6380
6381@hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL
6382
6383@hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY
6384
6385@hook TARGET_ASM_UNWIND_EMIT
6386
6387@hook TARGET_ASM_MAKE_EH_SYMBOL_INDIRECT
6388
6389@hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN
6390
6391@hook TARGET_ASM_SHOULD_RESTORE_CFA_STATE
6392
6393@node Exception Region Output
6394@subsection Assembler Commands for Exception Regions
6395
6396@c prevent bad page break with this line
6397
6398This describes commands marking the start and the end of an exception
6399region.
6400
6401@defmac EH_FRAME_SECTION_NAME
6402If defined, a C string constant for the name of the section containing
6403exception handling frame unwind information.  If not defined, GCC will
6404provide a default definition if the target supports named sections.
6405@file{crtstuff.c} uses this macro to switch to the appropriate section.
6406
6407You should define this symbol if your target supports DWARF 2 frame
6408unwind information and the default definition does not work.
6409@end defmac
6410
6411@defmac EH_FRAME_THROUGH_COLLECT2
6412If defined, DWARF 2 frame unwind information will identified by
6413specially named labels.  The collect2 process will locate these
6414labels and generate code to register the frames.
6415
6416This might be necessary, for instance, if the system linker will not
6417place the eh_frames in-between the sentinals from @file{crtstuff.c},
6418or if the system linker does garbage collection and sections cannot
6419be marked as not to be collected.
6420@end defmac
6421
6422@defmac EH_TABLES_CAN_BE_READ_ONLY
6423Define this macro to 1 if your target is such that no frame unwind
6424information encoding used with non-PIC code will ever require a
6425runtime relocation, but the linker may not support merging read-only
6426and read-write sections into a single read-write section.
6427@end defmac
6428
6429@defmac MASK_RETURN_ADDR
6430An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
6431that it does not contain any extraneous set bits in it.
6432@end defmac
6433
6434@defmac DWARF2_UNWIND_INFO
6435Define this macro to 0 if your target supports DWARF 2 frame unwind
6436information, but it does not yet work with exception handling.
6437Otherwise, if your target supports this information (if it defines
6438@code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}),
6439GCC will provide a default definition of 1.
6440@end defmac
6441
6442@hook TARGET_EXCEPT_UNWIND_INFO
6443This hook defines the mechanism that will be used for exception handling
6444by the target.  If the target has ABI specified unwind tables, the hook
6445should return @code{UI_TARGET}.  If the target is to use the
6446@code{setjmp}/@code{longjmp}-based exception handling scheme, the hook
6447should return @code{UI_SJLJ}.  If the target supports DWARF 2 frame unwind
6448information, the hook should return @code{UI_DWARF2}.
6449
6450A target may, if exceptions are disabled, choose to return @code{UI_NONE}.
6451This may end up simplifying other parts of target-specific code.  The
6452default implementation of this hook never returns @code{UI_NONE}.
6453
6454Note that the value returned by this hook should be constant.  It should
6455not depend on anything except the command-line switches described by
6456@var{opts}.  In particular, the
6457setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor
6458macros and builtin functions related to exception handling are set up
6459depending on this setting.
6460
6461The default implementation of the hook first honors the
6462@option{--enable-sjlj-exceptions} configure option, then
6463@code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}.  If
6464@code{DWARF2_UNWIND_INFO} depends on command-line options, the target
6465must define this hook so that @var{opts} is used correctly.
6466@end deftypefn
6467
6468@hook TARGET_UNWIND_TABLES_DEFAULT
6469This variable should be set to @code{true} if the target ABI requires unwinding
6470tables even when exceptions are not used.  It must not be modified by
6471command-line option processing.
6472@end deftypevr
6473
6474@defmac DONT_USE_BUILTIN_SETJMP
6475Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
6476should use the @code{setjmp}/@code{longjmp} functions from the C library
6477instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
6478@end defmac
6479
6480@defmac JMP_BUF_SIZE
6481This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also
6482defined.  Define this macro if the default size of @code{jmp_buf} buffer
6483for the @code{setjmp}/@code{longjmp}-based exception handling mechanism
6484is not large enough, or if it is much too large.
6485The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}.
6486@end defmac
6487
6488@defmac DWARF_CIE_DATA_ALIGNMENT
6489This macro need only be defined if the target might save registers in the
6490function prologue at an offset to the stack pointer that is not aligned to
6491@code{UNITS_PER_WORD}.  The definition should be the negative minimum
6492alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive
6493minimum alignment otherwise.  @xref{DWARF}.  Only applicable if
6494the target supports DWARF 2 frame unwind information.
6495@end defmac
6496
6497@hook TARGET_TERMINATE_DW2_EH_FRAME_INFO
6498
6499@hook TARGET_DWARF_REGISTER_SPAN
6500
6501@hook TARGET_DWARF_FRAME_REG_MODE
6502
6503@hook TARGET_INIT_DWARF_REG_SIZES_EXTRA
6504
6505@hook TARGET_ASM_TTYPE
6506
6507@hook TARGET_ARM_EABI_UNWINDER
6508
6509@node Alignment Output
6510@subsection Assembler Commands for Alignment
6511
6512@c prevent bad page break with this line
6513This describes commands for alignment.
6514
6515@defmac JUMP_ALIGN (@var{label})
6516The alignment (log base 2) to put in front of @var{label}, which is
6517a common destination of jumps and has no fallthru incoming edge.
6518
6519This macro need not be defined if you don't want any special alignment
6520to be done at such a time.  Most machine descriptions do not currently
6521define the macro.
6522
6523Unless it's necessary to inspect the @var{label} parameter, it is better
6524to set the variable @var{align_jumps} in the target's
6525@code{TARGET_OPTION_OVERRIDE}.  Otherwise, you should try to honor the user's
6526selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
6527@end defmac
6528
6529@defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
6530The alignment (log base 2) to put in front of @var{label}, which follows
6531a @code{BARRIER}.
6532
6533This macro need not be defined if you don't want any special alignment
6534to be done at such a time.  Most machine descriptions do not currently
6535define the macro.
6536@end defmac
6537
6538@defmac LOOP_ALIGN (@var{label})
6539The alignment (log base 2) to put in front of @var{label} that heads
6540a frequently executed basic block (usually the header of a loop).
6541
6542This macro need not be defined if you don't want any special alignment
6543to be done at such a time.  Most machine descriptions do not currently
6544define the macro.
6545
6546Unless it's necessary to inspect the @var{label} parameter, it is better
6547to set the variable @code{align_loops} in the target's
6548@code{TARGET_OPTION_OVERRIDE}.  Otherwise, you should try to honor the user's
6549selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
6550@end defmac
6551
6552@defmac LABEL_ALIGN (@var{label})
6553The alignment (log base 2) to put in front of @var{label}.
6554If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
6555the maximum of the specified values is used.
6556
6557Unless it's necessary to inspect the @var{label} parameter, it is better
6558to set the variable @code{align_labels} in the target's
6559@code{TARGET_OPTION_OVERRIDE}.  Otherwise, you should try to honor the user's
6560selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
6561@end defmac
6562
6563@defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
6564A C statement to output to the stdio stream @var{stream} an assembler
6565instruction to advance the location counter by @var{nbytes} bytes.
6566Those bytes should be zero when loaded.  @var{nbytes} will be a C
6567expression of type @code{unsigned HOST_WIDE_INT}.
6568@end defmac
6569
6570@defmac ASM_NO_SKIP_IN_TEXT
6571Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
6572text section because it fails to put zeros in the bytes that are skipped.
6573This is true on many Unix systems, where the pseudo--op to skip bytes
6574produces no-op instructions rather than zeros when used in the text
6575section.
6576@end defmac
6577
6578@defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
6579A C statement to output to the stdio stream @var{stream} an assembler
6580command to advance the location counter to a multiple of 2 to the
6581@var{power} bytes.  @var{power} will be a C expression of type @code{int}.
6582@end defmac
6583
6584@defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
6585Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
6586for padding, if necessary.
6587@end defmac
6588
6589@defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
6590A C statement to output to the stdio stream @var{stream} an assembler
6591command to advance the location counter to a multiple of 2 to the
6592@var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
6593satisfy the alignment request.  @var{power} and @var{max_skip} will be
6594a C expression of type @code{int}.
6595@end defmac
6596
6597@need 3000
6598@node Debugging Info
6599@section Controlling Debugging Information Format
6600
6601@c prevent bad page break with this line
6602This describes how to specify debugging information.
6603
6604@menu
6605* All Debuggers::      Macros that affect all debugging formats uniformly.
6606* DBX Options::        Macros enabling specific options in DBX format.
6607* DBX Hooks::          Hook macros for varying DBX format.
6608* File Names and DBX:: Macros controlling output of file names in DBX format.
6609* DWARF::              Macros for DWARF format.
6610* VMS Debug::          Macros for VMS debug format.
6611* CTF Debug::          Macros for CTF debug format.
6612* BTF Debug::          Macros for BTF debug format.
6613@end menu
6614
6615@node All Debuggers
6616@subsection Macros Affecting All Debugging Formats
6617
6618@c prevent bad page break with this line
6619These macros affect all debugging formats.
6620
6621@defmac DBX_REGISTER_NUMBER (@var{regno})
6622A C expression that returns the DBX register number for the compiler
6623register number @var{regno}.  In the default macro provided, the value
6624of this expression will be @var{regno} itself.  But sometimes there are
6625some registers that the compiler knows about and DBX does not, or vice
6626versa.  In such cases, some register may need to have one number in the
6627compiler and another for DBX@.
6628
6629If two registers have consecutive numbers inside GCC, and they can be
6630used as a pair to hold a multiword value, then they @emph{must} have
6631consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
6632Otherwise, debuggers will be unable to access such a pair, because they
6633expect register pairs to be consecutive in their own numbering scheme.
6634
6635If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
6636does not preserve register pairs, then what you must do instead is
6637redefine the actual register numbering scheme.
6638@end defmac
6639
6640@defmac DEBUGGER_AUTO_OFFSET (@var{x})
6641A C expression that returns the integer offset value for an automatic
6642variable having address @var{x} (an RTL expression).  The default
6643computation assumes that @var{x} is based on the frame-pointer and
6644gives the offset from the frame-pointer.  This is required for targets
6645that produce debugging output for DBX and allow the frame-pointer to be
6646eliminated when the @option{-g} option is used.
6647@end defmac
6648
6649@defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
6650A C expression that returns the integer offset value for an argument
6651having address @var{x} (an RTL expression).  The nominal offset is
6652@var{offset}.
6653@end defmac
6654
6655@defmac PREFERRED_DEBUGGING_TYPE
6656A C expression that returns the type of debugging output GCC should
6657produce when the user specifies just @option{-g}.  Define
6658this if you have arranged for GCC to support more than one format of
6659debugging output.  Currently, the allowable values are @code{DBX_DEBUG},
6660@code{DWARF2_DEBUG}, @code{XCOFF_DEBUG}, @code{VMS_DEBUG},
6661and @code{VMS_AND_DWARF2_DEBUG}.
6662
6663When the user specifies @option{-ggdb}, GCC normally also uses the
6664value of this macro to select the debugging output format, but with two
6665exceptions.  If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
6666value @code{DWARF2_DEBUG}.  Otherwise, if @code{DBX_DEBUGGING_INFO} is
6667defined, GCC uses @code{DBX_DEBUG}.
6668
6669The value of this macro only affects the default debugging output; the
6670user can always get a specific type of output by using @option{-gstabs},
6671@option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
6672@end defmac
6673
6674@node DBX Options
6675@subsection Specific Options for DBX Output
6676
6677@c prevent bad page break with this line
6678These are specific options for DBX output.
6679
6680@defmac DBX_DEBUGGING_INFO
6681Define this macro if GCC should produce debugging output for DBX
6682in response to the @option{-g} option.
6683@end defmac
6684
6685@defmac XCOFF_DEBUGGING_INFO
6686Define this macro if GCC should produce XCOFF format debugging output
6687in response to the @option{-g} option.  This is a variant of DBX format.
6688@end defmac
6689
6690@defmac DEFAULT_GDB_EXTENSIONS
6691Define this macro to control whether GCC should by default generate
6692GDB's extended version of DBX debugging information (assuming DBX-format
6693debugging information is enabled at all).  If you don't define the
6694macro, the default is 1: always generate the extended information
6695if there is any occasion to.
6696@end defmac
6697
6698@defmac DEBUG_SYMS_TEXT
6699Define this macro if all @code{.stabs} commands should be output while
6700in the text section.
6701@end defmac
6702
6703@defmac ASM_STABS_OP
6704A C string constant, including spacing, naming the assembler pseudo op to
6705use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
6706If you don't define this macro, @code{"\t.stabs\t"} is used.  This macro
6707applies only to DBX debugging information format.
6708@end defmac
6709
6710@defmac ASM_STABD_OP
6711A C string constant, including spacing, naming the assembler pseudo op to
6712use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
6713value is the current location.  If you don't define this macro,
6714@code{"\t.stabd\t"} is used.  This macro applies only to DBX debugging
6715information format.
6716@end defmac
6717
6718@defmac ASM_STABN_OP
6719A C string constant, including spacing, naming the assembler pseudo op to
6720use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
6721name.  If you don't define this macro, @code{"\t.stabn\t"} is used.  This
6722macro applies only to DBX debugging information format.
6723@end defmac
6724
6725@defmac DBX_NO_XREFS
6726Define this macro if DBX on your system does not support the construct
6727@samp{xs@var{tagname}}.  On some systems, this construct is used to
6728describe a forward reference to a structure named @var{tagname}.
6729On other systems, this construct is not supported at all.
6730@end defmac
6731
6732@defmac DBX_CONTIN_LENGTH
6733A symbol name in DBX-format debugging information is normally
6734continued (split into two separate @code{.stabs} directives) when it
6735exceeds a certain length (by default, 80 characters).  On some
6736operating systems, DBX requires this splitting; on others, splitting
6737must not be done.  You can inhibit splitting by defining this macro
6738with the value zero.  You can override the default splitting-length by
6739defining this macro as an expression for the length you desire.
6740@end defmac
6741
6742@defmac DBX_CONTIN_CHAR
6743Normally continuation is indicated by adding a @samp{\} character to
6744the end of a @code{.stabs} string when a continuation follows.  To use
6745a different character instead, define this macro as a character
6746constant for the character you want to use.  Do not define this macro
6747if backslash is correct for your system.
6748@end defmac
6749
6750@defmac DBX_STATIC_STAB_DATA_SECTION
6751Define this macro if it is necessary to go to the data section before
6752outputting the @samp{.stabs} pseudo-op for a non-global static
6753variable.
6754@end defmac
6755
6756@defmac DBX_TYPE_DECL_STABS_CODE
6757The value to use in the ``code'' field of the @code{.stabs} directive
6758for a typedef.  The default is @code{N_LSYM}.
6759@end defmac
6760
6761@defmac DBX_STATIC_CONST_VAR_CODE
6762The value to use in the ``code'' field of the @code{.stabs} directive
6763for a static variable located in the text section.  DBX format does not
6764provide any ``right'' way to do this.  The default is @code{N_FUN}.
6765@end defmac
6766
6767@defmac DBX_REGPARM_STABS_CODE
6768The value to use in the ``code'' field of the @code{.stabs} directive
6769for a parameter passed in registers.  DBX format does not provide any
6770``right'' way to do this.  The default is @code{N_RSYM}.
6771@end defmac
6772
6773@defmac DBX_REGPARM_STABS_LETTER
6774The letter to use in DBX symbol data to identify a symbol as a parameter
6775passed in registers.  DBX format does not customarily provide any way to
6776do this.  The default is @code{'P'}.
6777@end defmac
6778
6779@defmac DBX_FUNCTION_FIRST
6780Define this macro if the DBX information for a function and its
6781arguments should precede the assembler code for the function.  Normally,
6782in DBX format, the debugging information entirely follows the assembler
6783code.
6784@end defmac
6785
6786@defmac DBX_BLOCKS_FUNCTION_RELATIVE
6787Define this macro, with value 1, if the value of a symbol describing
6788the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
6789relative to the start of the enclosing function.  Normally, GCC uses
6790an absolute address.
6791@end defmac
6792
6793@defmac DBX_LINES_FUNCTION_RELATIVE
6794Define this macro, with value 1, if the value of a symbol indicating
6795the current line number (@code{N_SLINE}) should be relative to the
6796start of the enclosing function.  Normally, GCC uses an absolute address.
6797@end defmac
6798
6799@defmac DBX_USE_BINCL
6800Define this macro if GCC should generate @code{N_BINCL} and
6801@code{N_EINCL} stabs for included header files, as on Sun systems.  This
6802macro also directs GCC to output a type number as a pair of a file
6803number and a type number within the file.  Normally, GCC does not
6804generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
6805number for a type number.
6806@end defmac
6807
6808@node DBX Hooks
6809@subsection Open-Ended Hooks for DBX Format
6810
6811@c prevent bad page break with this line
6812These are hooks for DBX format.
6813
6814@defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
6815A C statement to output DBX debugging information before code for line
6816number @var{line} of the current source file to the stdio stream
6817@var{stream}.  @var{counter} is the number of time the macro was
6818invoked, including the current invocation; it is intended to generate
6819unique labels in the assembly output.
6820
6821This macro should not be defined if the default output is correct, or
6822if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
6823@end defmac
6824
6825@defmac NO_DBX_FUNCTION_END
6826Some stabs encapsulation formats (in particular ECOFF), cannot handle the
6827@code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
6828On those machines, define this macro to turn this feature off without
6829disturbing the rest of the gdb extensions.
6830@end defmac
6831
6832@defmac NO_DBX_BNSYM_ENSYM
6833Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
6834extension construct.  On those machines, define this macro to turn this
6835feature off without disturbing the rest of the gdb extensions.
6836@end defmac
6837
6838@node File Names and DBX
6839@subsection File Names in DBX Format
6840
6841@c prevent bad page break with this line
6842This describes file names in DBX format.
6843
6844@defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
6845A C statement to output DBX debugging information to the stdio stream
6846@var{stream}, which indicates that file @var{name} is the main source
6847file---the file specified as the input file for compilation.
6848This macro is called only once, at the beginning of compilation.
6849
6850This macro need not be defined if the standard form of output
6851for DBX debugging information is appropriate.
6852
6853It may be necessary to refer to a label equal to the beginning of the
6854text section.  You can use @samp{assemble_name (stream, ltext_label_name)}
6855to do so.  If you do this, you must also set the variable
6856@var{used_ltext_label_name} to @code{true}.
6857@end defmac
6858
6859@defmac NO_DBX_MAIN_SOURCE_DIRECTORY
6860Define this macro, with value 1, if GCC should not emit an indication
6861of the current directory for compilation and current source language at
6862the beginning of the file.
6863@end defmac
6864
6865@defmac NO_DBX_GCC_MARKER
6866Define this macro, with value 1, if GCC should not emit an indication
6867that this object file was compiled by GCC@.  The default is to emit
6868an @code{N_OPT} stab at the beginning of every source file, with
6869@samp{gcc2_compiled.} for the string and value 0.
6870@end defmac
6871
6872@defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
6873A C statement to output DBX debugging information at the end of
6874compilation of the main source file @var{name}.  Output should be
6875written to the stdio stream @var{stream}.
6876
6877If you don't define this macro, nothing special is output at the end
6878of compilation, which is correct for most machines.
6879@end defmac
6880
6881@defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
6882Define this macro @emph{instead of} defining
6883@code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
6884the end of compilation is an @code{N_SO} stab with an empty string,
6885whose value is the highest absolute text address in the file.
6886@end defmac
6887
6888@need 2000
6889@node DWARF
6890@subsection Macros for DWARF Output
6891
6892@c prevent bad page break with this line
6893Here are macros for DWARF output.
6894
6895@defmac DWARF2_DEBUGGING_INFO
6896Define this macro if GCC should produce dwarf version 2 format
6897debugging output in response to the @option{-g} option.
6898
6899@hook TARGET_DWARF_CALLING_CONVENTION
6900
6901To support optional call frame debugging information, you must also
6902define @code{INCOMING_RETURN_ADDR_RTX} and either set
6903@code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
6904prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
6905as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
6906@end defmac
6907
6908@defmac DWARF2_FRAME_INFO
6909Define this macro to a nonzero value if GCC should always output
6910Dwarf 2 frame information.  If @code{TARGET_EXCEPT_UNWIND_INFO}
6911(@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and
6912exceptions are enabled, GCC will output this information not matter
6913how you define @code{DWARF2_FRAME_INFO}.
6914@end defmac
6915
6916@hook TARGET_DEBUG_UNWIND_INFO
6917
6918@defmac DWARF2_ASM_LINE_DEBUG_INFO
6919Define this macro to be a nonzero value if the assembler can generate Dwarf 2
6920line debug info sections.  This will result in much more compact line number
6921tables, and hence is desirable if it works.
6922@end defmac
6923
6924@defmac DWARF2_ASM_VIEW_DEBUG_INFO
6925Define this macro to be a nonzero value if the assembler supports view
6926assignment and verification in @code{.loc}.  If it does not, but the
6927user enables location views, the compiler may have to fallback to
6928internal line number tables.
6929@end defmac
6930
6931@hook TARGET_RESET_LOCATION_VIEW
6932
6933@hook TARGET_WANT_DEBUG_PUB_SECTIONS
6934
6935@hook TARGET_DELAY_SCHED2
6936
6937@hook TARGET_DELAY_VARTRACK
6938
6939@hook TARGET_NO_REGISTER_ALLOCATION
6940
6941@defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
6942A C statement to issue assembly directives that create a difference
6943@var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
6944@end defmac
6945
6946@defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
6947A C statement to issue assembly directives that create a difference
6948between the two given labels in system defined units, e.g.@: instruction
6949slots on IA64 VMS, using an integer of the given size.
6950@end defmac
6951
6952@defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{offset}, @var{section})
6953A C statement to issue assembly directives that create a
6954section-relative reference to the given @var{label} plus @var{offset}, using
6955an integer of the given @var{size}.  The label is known to be defined in the
6956given @var{section}.
6957@end defmac
6958
6959@defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
6960A C statement to issue assembly directives that create a self-relative
6961reference to the given @var{label}, using an integer of the given @var{size}.
6962@end defmac
6963
6964@defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label})
6965A C statement to issue assembly directives that create a reference to the
6966given @var{label} relative to the dbase, using an integer of the given @var{size}.
6967@end defmac
6968
6969@defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
6970A C statement to issue assembly directives that create a reference to
6971the DWARF table identifier @var{label} from the current section.  This
6972is used on some systems to avoid garbage collecting a DWARF table which
6973is referenced by a function.
6974@end defmac
6975
6976@hook TARGET_ASM_OUTPUT_DWARF_DTPREL
6977
6978@need 2000
6979@node VMS Debug
6980@subsection Macros for VMS Debug Format
6981
6982@c prevent bad page break with this line
6983Here are macros for VMS debug format.
6984
6985@defmac VMS_DEBUGGING_INFO
6986Define this macro if GCC should produce debugging output for VMS
6987in response to the @option{-g} option.  The default behavior for VMS
6988is to generate minimal debug info for a traceback in the absence of
6989@option{-g} unless explicitly overridden with @option{-g0}.  This
6990behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and
6991@code{TARGET_OPTION_OVERRIDE}.
6992@end defmac
6993
6994@need 2000
6995@node CTF Debug
6996@subsection Macros for CTF Debug Format
6997
6998@c prevent bad page break with this line
6999Here are macros for CTF debug format.
7000
7001@defmac CTF_DEBUGGING_INFO
7002Define this macro if GCC should produce debugging output in CTF debug
7003format in response to the @option{-gctf} option.
7004@end defmac
7005
7006@need 2000
7007@node BTF Debug
7008@subsection Macros for BTF Debug Format
7009
7010@c prevent bad page break with this line
7011Here are macros for BTF debug format.
7012
7013@defmac BTF_DEBUGGING_INFO
7014Define this macro if GCC should produce debugging output in BTF debug
7015format in response to the @option{-gbtf} option.
7016@end defmac
7017
7018@node Floating Point
7019@section Cross Compilation and Floating Point
7020@cindex cross compilation and floating point
7021@cindex floating point and cross compilation
7022
7023While all modern machines use twos-complement representation for integers,
7024there are a variety of representations for floating point numbers.  This
7025means that in a cross-compiler the representation of floating point numbers
7026in the compiled program may be different from that used in the machine
7027doing the compilation.
7028
7029Because different representation systems may offer different amounts of
7030range and precision, all floating point constants must be represented in
7031the target machine's format.  Therefore, the cross compiler cannot
7032safely use the host machine's floating point arithmetic; it must emulate
7033the target's arithmetic.  To ensure consistency, GCC always uses
7034emulation to work with floating point values, even when the host and
7035target floating point formats are identical.
7036
7037The following macros are provided by @file{real.h} for the compiler to
7038use.  All parts of the compiler which generate or optimize
7039floating-point calculations must use these macros.  They may evaluate
7040their operands more than once, so operands must not have side effects.
7041
7042@defmac REAL_VALUE_TYPE
7043The C data type to be used to hold a floating point value in the target
7044machine's format.  Typically this is a @code{struct} containing an
7045array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
7046quantity.
7047@end defmac
7048
7049@deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
7050Truncates @var{x} to a signed integer, rounding toward zero.
7051@end deftypefn
7052
7053@deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
7054Truncates @var{x} to an unsigned integer, rounding toward zero.  If
7055@var{x} is negative, returns zero.
7056@end deftypefn
7057
7058@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode})
7059Converts @var{string} into a floating point number in the target machine's
7060representation for mode @var{mode}.  This routine can handle both
7061decimal and hexadecimal floating point constants, using the syntax
7062defined by the C language for both.
7063@end deftypefn
7064
7065@deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
7066Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
7067@end deftypefn
7068
7069@deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
7070Determines whether @var{x} represents infinity (positive or negative).
7071@end deftypefn
7072
7073@deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
7074Determines whether @var{x} represents a ``NaN'' (not-a-number).
7075@end deftypefn
7076
7077@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
7078Returns the negative of the floating point value @var{x}.
7079@end deftypefn
7080
7081@deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
7082Returns the absolute value of @var{x}.
7083@end deftypefn
7084
7085@node Mode Switching
7086@section Mode Switching Instructions
7087@cindex mode switching
7088The following macros control mode switching optimizations:
7089
7090@defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
7091Define this macro if the port needs extra instructions inserted for mode
7092switching in an optimizing compilation.
7093
7094For an example, the SH4 can perform both single and double precision
7095floating point operations, but to perform a single precision operation,
7096the FPSCR PR bit has to be cleared, while for a double precision
7097operation, this bit has to be set.  Changing the PR bit requires a general
7098purpose register as a scratch register, hence these FPSCR sets have to
7099be inserted before reload, i.e.@: you cannot put this into instruction emitting
7100or @code{TARGET_MACHINE_DEPENDENT_REORG}.
7101
7102You can have multiple entities that are mode-switched, and select at run time
7103which entities actually need it.  @code{OPTIMIZE_MODE_SWITCHING} should
7104return nonzero for any @var{entity} that needs mode-switching.
7105If you define this macro, you also have to define
7106@code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED},
7107@code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}.
7108@code{TARGET_MODE_AFTER}, @code{TARGET_MODE_ENTRY}, and @code{TARGET_MODE_EXIT}
7109are optional.
7110@end defmac
7111
7112@defmac NUM_MODES_FOR_MODE_SWITCHING
7113If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
7114initializer for an array of integers.  Each initializer element
7115N refers to an entity that needs mode switching, and specifies the number
7116of different modes that might need to be set for this entity.
7117The position of the initializer in the initializer---starting counting at
7118zero---determines the integer that is used to refer to the mode-switched
7119entity in question.
7120In macros that take mode arguments / yield a mode result, modes are
7121represented as numbers 0 @dots{} N @minus{} 1.  N is used to specify that no mode
7122switch is needed / supplied.
7123@end defmac
7124
7125@hook TARGET_MODE_EMIT
7126
7127@hook TARGET_MODE_NEEDED
7128
7129@hook TARGET_MODE_AFTER
7130
7131@hook TARGET_MODE_ENTRY
7132
7133@hook TARGET_MODE_EXIT
7134
7135@hook TARGET_MODE_PRIORITY
7136
7137@node Target Attributes
7138@section Defining target-specific uses of @code{__attribute__}
7139@cindex target attributes
7140@cindex machine attributes
7141@cindex attributes, target-specific
7142
7143Target-specific attributes may be defined for functions, data and types.
7144These are described using the following target hooks; they also need to
7145be documented in @file{extend.texi}.
7146
7147@hook TARGET_ATTRIBUTE_TABLE
7148
7149@hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P
7150
7151@hook TARGET_COMP_TYPE_ATTRIBUTES
7152
7153@hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES
7154
7155@hook TARGET_MERGE_TYPE_ATTRIBUTES
7156
7157@hook TARGET_MERGE_DECL_ATTRIBUTES
7158
7159@hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P
7160
7161@defmac TARGET_DECLSPEC
7162Define this macro to a nonzero value if you want to treat
7163@code{__declspec(X)} as equivalent to @code{__attribute((X))}.  By
7164default, this behavior is enabled only for targets that define
7165@code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}.  The current implementation
7166of @code{__declspec} is via a built-in macro, but you should not rely
7167on this implementation detail.
7168@end defmac
7169
7170@hook TARGET_INSERT_ATTRIBUTES
7171
7172@hook TARGET_HANDLE_GENERIC_ATTRIBUTE
7173
7174@hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P
7175
7176@hook TARGET_OPTION_VALID_ATTRIBUTE_P
7177
7178@hook TARGET_OPTION_SAVE
7179
7180@hook TARGET_OPTION_RESTORE
7181
7182@hook TARGET_OPTION_POST_STREAM_IN
7183
7184@hook TARGET_OPTION_PRINT
7185
7186@hook TARGET_OPTION_PRAGMA_PARSE
7187
7188@hook TARGET_OPTION_OVERRIDE
7189
7190@hook TARGET_OPTION_FUNCTION_VERSIONS
7191
7192@hook TARGET_CAN_INLINE_P
7193
7194@hook TARGET_UPDATE_IPA_FN_TARGET_INFO
7195
7196@hook TARGET_NEED_IPA_FN_TARGET_INFO
7197
7198@hook TARGET_RELAYOUT_FUNCTION
7199
7200@node Emulated TLS
7201@section Emulating TLS
7202@cindex Emulated TLS
7203
7204For targets whose psABI does not provide Thread Local Storage via
7205specific relocations and instruction sequences, an emulation layer is
7206used.  A set of target hooks allows this emulation layer to be
7207configured for the requirements of a particular target.  For instance
7208the psABI may in fact specify TLS support in terms of an emulation
7209layer.
7210
7211The emulation layer works by creating a control object for every TLS
7212object.  To access the TLS object, a lookup function is provided
7213which, when given the address of the control object, will return the
7214address of the current thread's instance of the TLS object.
7215
7216@hook TARGET_EMUTLS_GET_ADDRESS
7217
7218@hook TARGET_EMUTLS_REGISTER_COMMON
7219
7220@hook TARGET_EMUTLS_VAR_SECTION
7221
7222@hook TARGET_EMUTLS_TMPL_SECTION
7223
7224@hook TARGET_EMUTLS_VAR_PREFIX
7225
7226@hook TARGET_EMUTLS_TMPL_PREFIX
7227
7228@hook TARGET_EMUTLS_VAR_FIELDS
7229
7230@hook TARGET_EMUTLS_VAR_INIT
7231
7232@hook TARGET_EMUTLS_VAR_ALIGN_FIXED
7233
7234@hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
7235
7236@node MIPS Coprocessors
7237@section Defining coprocessor specifics for MIPS targets.
7238@cindex MIPS coprocessor-definition macros
7239
7240The MIPS specification allows MIPS implementations to have as many as 4
7241coprocessors, each with as many as 32 private registers.  GCC supports
7242accessing these registers and transferring values between the registers
7243and memory using asm-ized variables.  For example:
7244
7245@smallexample
7246  register unsigned int cp0count asm ("c0r1");
7247  unsigned int d;
7248
7249  d = cp0count + 3;
7250@end smallexample
7251
7252(``c0r1'' is the default name of register 1 in coprocessor 0; alternate
7253names may be added as described below, or the default names may be
7254overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
7255
7256Coprocessor registers are assumed to be epilogue-used; sets to them will
7257be preserved even if it does not appear that the register is used again
7258later in the function.
7259
7260Another note: according to the MIPS spec, coprocessor 1 (if present) is
7261the FPU@.  One accesses COP1 registers through standard mips
7262floating-point support; they are not included in this mechanism.
7263
7264@node PCH Target
7265@section Parameters for Precompiled Header Validity Checking
7266@cindex parameters, precompiled headers
7267
7268@hook TARGET_GET_PCH_VALIDITY
7269
7270@hook TARGET_PCH_VALID_P
7271
7272@hook TARGET_CHECK_PCH_TARGET_FLAGS
7273
7274@hook TARGET_PREPARE_PCH_SAVE
7275
7276@node C++ ABI
7277@section C++ ABI parameters
7278@cindex parameters, c++ abi
7279
7280@hook TARGET_CXX_GUARD_TYPE
7281
7282@hook TARGET_CXX_GUARD_MASK_BIT
7283
7284@hook TARGET_CXX_GET_COOKIE_SIZE
7285
7286@hook TARGET_CXX_COOKIE_HAS_SIZE
7287
7288@hook TARGET_CXX_IMPORT_EXPORT_CLASS
7289
7290@hook TARGET_CXX_CDTOR_RETURNS_THIS
7291
7292@hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE
7293
7294@hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY
7295
7296@hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT
7297
7298@hook TARGET_CXX_LIBRARY_RTTI_COMDAT
7299
7300@hook TARGET_CXX_USE_AEABI_ATEXIT
7301
7302@hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT
7303
7304@hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION
7305
7306@hook TARGET_CXX_DECL_MANGLING_CONTEXT
7307
7308@node D Language and ABI
7309@section D ABI parameters
7310@cindex parameters, d abi
7311
7312@hook TARGET_D_CPU_VERSIONS
7313
7314@hook TARGET_D_OS_VERSIONS
7315
7316@hook TARGET_D_REGISTER_CPU_TARGET_INFO
7317
7318@hook TARGET_D_REGISTER_OS_TARGET_INFO
7319
7320@hook TARGET_D_MINFO_SECTION
7321
7322@hook TARGET_D_MINFO_START_NAME
7323
7324@hook TARGET_D_MINFO_END_NAME
7325
7326@hook TARGET_D_HAS_STDCALL_CONVENTION
7327
7328@hook TARGET_D_TEMPLATES_ALWAYS_COMDAT
7329
7330@node Named Address Spaces
7331@section Adding support for named address spaces
7332@cindex named address spaces
7333
7334The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275
7335standards committee, @cite{Programming Languages - C - Extensions to
7336support embedded processors}, specifies a syntax for embedded
7337processors to specify alternate address spaces.  You can configure a
7338GCC port to support section 5.1 of the draft report to add support for
7339address spaces other than the default address space.  These address
7340spaces are new keywords that are similar to the @code{volatile} and
7341@code{const} type attributes.
7342
7343Pointers to named address spaces can have a different size than
7344pointers to the generic address space.
7345
7346For example, the SPU port uses the @code{__ea} address space to refer
7347to memory in the host processor, rather than memory local to the SPU
7348processor.  Access to memory in the @code{__ea} address space involves
7349issuing DMA operations to move data between the host processor and the
7350local processor memory address space.  Pointers in the @code{__ea}
7351address space are either 32 bits or 64 bits based on the
7352@option{-mea32} or @option{-mea64} switches (native SPU pointers are
7353always 32 bits).
7354
7355Internally, address spaces are represented as a small integer in the
7356range 0 to 15 with address space 0 being reserved for the generic
7357address space.
7358
7359To register a named address space qualifier keyword with the C front end,
7360the target may call the @code{c_register_addr_space} routine.  For example,
7361the SPU port uses the following to declare @code{__ea} as the keyword for
7362named address space #1:
7363@smallexample
7364#define ADDR_SPACE_EA 1
7365c_register_addr_space ("__ea", ADDR_SPACE_EA);
7366@end smallexample
7367
7368@hook TARGET_ADDR_SPACE_POINTER_MODE
7369
7370@hook TARGET_ADDR_SPACE_ADDRESS_MODE
7371
7372@hook TARGET_ADDR_SPACE_VALID_POINTER_MODE
7373
7374@hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P
7375
7376@hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS
7377
7378@hook TARGET_ADDR_SPACE_SUBSET_P
7379
7380@hook TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID
7381
7382@hook TARGET_ADDR_SPACE_CONVERT
7383
7384@hook TARGET_ADDR_SPACE_DEBUG
7385
7386@hook TARGET_ADDR_SPACE_DIAGNOSE_USAGE
7387
7388@node Misc
7389@section Miscellaneous Parameters
7390@cindex parameters, miscellaneous
7391
7392@c prevent bad page break with this line
7393Here are several miscellaneous parameters.
7394
7395@defmac HAS_LONG_COND_BRANCH
7396Define this boolean macro to indicate whether or not your architecture
7397has conditional branches that can span all of memory.  It is used in
7398conjunction with an optimization that partitions hot and cold basic
7399blocks into separate sections of the executable.  If this macro is
7400set to false, gcc will convert any conditional branches that attempt
7401to cross between sections into unconditional branches or indirect jumps.
7402@end defmac
7403
7404@defmac HAS_LONG_UNCOND_BRANCH
7405Define this boolean macro to indicate whether or not your architecture
7406has unconditional branches that can span all of memory.  It is used in
7407conjunction with an optimization that partitions hot and cold basic
7408blocks into separate sections of the executable.  If this macro is
7409set to false, gcc will convert any unconditional branches that attempt
7410to cross between sections into indirect jumps.
7411@end defmac
7412
7413@defmac CASE_VECTOR_MODE
7414An alias for a machine mode name.  This is the machine mode that
7415elements of a jump-table should have.
7416@end defmac
7417
7418@defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
7419Optional: return the preferred mode for an @code{addr_diff_vec}
7420when the minimum and maximum offset are known.  If you define this,
7421it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
7422To make this work, you also have to define @code{INSN_ALIGN} and
7423make the alignment for @code{addr_diff_vec} explicit.
7424The @var{body} argument is provided so that the offset_unsigned and scale
7425flags can be updated.
7426@end defmac
7427
7428@defmac CASE_VECTOR_PC_RELATIVE
7429Define this macro to be a C expression to indicate when jump-tables
7430should contain relative addresses.  You need not define this macro if
7431jump-tables never contain relative addresses, or jump-tables should
7432contain relative addresses only when @option{-fPIC} or @option{-fPIC}
7433is in effect.
7434@end defmac
7435
7436@hook TARGET_CASE_VALUES_THRESHOLD
7437
7438@defmac WORD_REGISTER_OPERATIONS
7439Define this macro to 1 if operations between registers with integral mode
7440smaller than a word are always performed on the entire register.  To be
7441more explicit, if you start with a pair of @code{word_mode} registers with
7442known values and you do a subword, for example @code{QImode}, addition on
7443the low part of the registers, then the compiler may consider that the
7444result has a known value in @code{word_mode} too if the macro is defined
7445to 1.  Most RISC machines have this property and most CISC machines do not.
7446@end defmac
7447
7448@hook TARGET_MIN_ARITHMETIC_PRECISION
7449
7450@defmac LOAD_EXTEND_OP (@var{mem_mode})
7451Define this macro to be a C expression indicating when insns that read
7452memory in @var{mem_mode}, an integral mode narrower than a word, set the
7453bits outside of @var{mem_mode} to be either the sign-extension or the
7454zero-extension of the data read.  Return @code{SIGN_EXTEND} for values
7455of @var{mem_mode} for which the
7456insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
7457@code{UNKNOWN} for other modes.
7458
7459This macro is not called with @var{mem_mode} non-integral or with a width
7460greater than or equal to @code{BITS_PER_WORD}, so you may return any
7461value in this case.  Do not define this macro if it would always return
7462@code{UNKNOWN}.  On machines where this macro is defined, you will normally
7463define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
7464
7465You may return a non-@code{UNKNOWN} value even if for some hard registers
7466the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
7467of these hard registers @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false
7468when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
7469integral mode larger than this but not larger than @code{word_mode}.
7470
7471You must return @code{UNKNOWN} if for some hard registers that allow this
7472mode, @code{TARGET_CAN_CHANGE_MODE_CLASS} says that they cannot change to
7473@code{word_mode}, but that they can change to another integral mode that
7474is larger then @var{mem_mode} but still smaller than @code{word_mode}.
7475@end defmac
7476
7477@defmac SHORT_IMMEDIATES_SIGN_EXTEND
7478Define this macro to 1 if loading short immediate values into registers sign
7479extends.
7480@end defmac
7481
7482@hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL
7483
7484@defmac MOVE_MAX
7485The maximum number of bytes that a single instruction can move quickly
7486between memory and registers or between two memory locations.
7487@end defmac
7488
7489@defmac MAX_MOVE_MAX
7490The maximum number of bytes that a single instruction can move quickly
7491between memory and registers or between two memory locations.  If this
7492is undefined, the default is @code{MOVE_MAX}.  Otherwise, it is the
7493constant value that is the largest value that @code{MOVE_MAX} can have
7494at run-time.
7495@end defmac
7496
7497@defmac SHIFT_COUNT_TRUNCATED
7498A C expression that is nonzero if on this machine the number of bits
7499actually used for the count of a shift operation is equal to the number
7500of bits needed to represent the size of the object being shifted.  When
7501this macro is nonzero, the compiler will assume that it is safe to omit
7502a sign-extend, zero-extend, and certain bitwise `and' instructions that
7503truncates the count of a shift operation.  On machines that have
7504instructions that act on bit-fields at variable positions, which may
7505include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
7506also enables deletion of truncations of the values that serve as
7507arguments to bit-field instructions.
7508
7509If both types of instructions truncate the count (for shifts) and
7510position (for bit-field operations), or if no variable-position bit-field
7511instructions exist, you should define this macro.
7512
7513However, on some machines, such as the 80386 and the 680x0, truncation
7514only applies to shift operations and not the (real or pretended)
7515bit-field operations.  Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
7516such machines.  Instead, add patterns to the @file{md} file that include
7517the implied truncation of the shift instructions.
7518
7519You need not define this macro if it would always have the value of zero.
7520@end defmac
7521
7522@anchor{TARGET_SHIFT_TRUNCATION_MASK}
7523@hook TARGET_SHIFT_TRUNCATION_MASK
7524
7525@hook TARGET_TRULY_NOOP_TRUNCATION
7526
7527@hook TARGET_MODE_REP_EXTENDED
7528
7529@hook TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P
7530
7531@defmac STORE_FLAG_VALUE
7532A C expression describing the value returned by a comparison operator
7533with an integral mode and stored by a store-flag instruction
7534(@samp{cstore@var{mode}4}) when the condition is true.  This description must
7535apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the
7536comparison operators whose results have a @code{MODE_INT} mode.
7537
7538A value of 1 or @minus{}1 means that the instruction implementing the
7539comparison operator returns exactly 1 or @minus{}1 when the comparison is true
7540and 0 when the comparison is false.  Otherwise, the value indicates
7541which bits of the result are guaranteed to be 1 when the comparison is
7542true.  This value is interpreted in the mode of the comparison
7543operation, which is given by the mode of the first operand in the
7544@samp{cstore@var{mode}4} pattern.  Either the low bit or the sign bit of
7545@code{STORE_FLAG_VALUE} be on.  Presently, only those bits are used by
7546the compiler.
7547
7548If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
7549generate code that depends only on the specified bits.  It can also
7550replace comparison operators with equivalent operations if they cause
7551the required bits to be set, even if the remaining bits are undefined.
7552For example, on a machine whose comparison operators return an
7553@code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
7554@samp{0x80000000}, saying that just the sign bit is relevant, the
7555expression
7556
7557@smallexample
7558(ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
7559@end smallexample
7560
7561@noindent
7562can be converted to
7563
7564@smallexample
7565(ashift:SI @var{x} (const_int @var{n}))
7566@end smallexample
7567
7568@noindent
7569where @var{n} is the appropriate shift count to move the bit being
7570tested into the sign bit.
7571
7572There is no way to describe a machine that always sets the low-order bit
7573for a true value, but does not guarantee the value of any other bits,
7574but we do not know of any machine that has such an instruction.  If you
7575are trying to port GCC to such a machine, include an instruction to
7576perform a logical-and of the result with 1 in the pattern for the
7577comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
7578
7579Often, a machine will have multiple instructions that obtain a value
7580from a comparison (or the condition codes).  Here are rules to guide the
7581choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
7582to be used:
7583
7584@itemize @bullet
7585@item
7586Use the shortest sequence that yields a valid definition for
7587@code{STORE_FLAG_VALUE}.  It is more efficient for the compiler to
7588``normalize'' the value (convert it to, e.g., 1 or 0) than for the
7589comparison operators to do so because there may be opportunities to
7590combine the normalization with other operations.
7591
7592@item
7593For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
7594slightly preferred on machines with expensive jumps and 1 preferred on
7595other machines.
7596
7597@item
7598As a second choice, choose a value of @samp{0x80000001} if instructions
7599exist that set both the sign and low-order bits but do not define the
7600others.
7601
7602@item
7603Otherwise, use a value of @samp{0x80000000}.
7604@end itemize
7605
7606Many machines can produce both the value chosen for
7607@code{STORE_FLAG_VALUE} and its negation in the same number of
7608instructions.  On those machines, you should also define a pattern for
7609those cases, e.g., one matching
7610
7611@smallexample
7612(set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
7613@end smallexample
7614
7615Some machines can also perform @code{and} or @code{plus} operations on
7616condition code values with less instructions than the corresponding
7617@samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}.  On those
7618machines, define the appropriate patterns.  Use the names @code{incscc}
7619and @code{decscc}, respectively, for the patterns which perform
7620@code{plus} or @code{minus} operations on condition code values.  See
7621@file{rs6000.md} for some examples.  The GNU Superoptimizer can be used to
7622find such instruction sequences on other machines.
7623
7624If this macro is not defined, the default value, 1, is used.  You need
7625not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
7626instructions, or if the value generated by these instructions is 1.
7627@end defmac
7628
7629@defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
7630A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
7631returned when comparison operators with floating-point results are true.
7632Define this macro on machines that have comparison operations that return
7633floating-point values.  If there are no such operations, do not define
7634this macro.
7635@end defmac
7636
7637@defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
7638A C expression that gives an rtx representing the nonzero true element
7639for vector comparisons.  The returned rtx should be valid for the inner
7640mode of @var{mode} which is guaranteed to be a vector mode.  Define
7641this macro on machines that have vector comparison operations that
7642return a vector result.  If there are no such operations, do not define
7643this macro.  Typically, this macro is defined as @code{const1_rtx} or
7644@code{constm1_rtx}.  This macro may return @code{NULL_RTX} to prevent
7645the compiler optimizing such vector comparison operations for the
7646given mode.
7647@end defmac
7648
7649@defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7650@defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
7651A C expression that indicates whether the architecture defines a value
7652for @code{clz} or @code{ctz} with a zero operand.
7653A result of @code{0} indicates the value is undefined.
7654If the value is defined for only the RTL expression, the macro should
7655evaluate to @code{1}; if the value applies also to the corresponding optab
7656entry (which is normally the case if it expands directly into
7657the corresponding RTL), then the macro should evaluate to @code{2}.
7658In the cases where the value is defined, @var{value} should be set to
7659this value.
7660
7661If this macro is not defined, the value of @code{clz} or
7662@code{ctz} at zero is assumed to be undefined.
7663
7664This macro must be defined if the target's expansion for @code{ffs}
7665relies on a particular value to get correct results.  Otherwise it
7666is not necessary, though it may be used to optimize some corner cases, and
7667to provide a default expansion for the @code{ffs} optab.
7668
7669Note that regardless of this macro the ``definedness'' of @code{clz}
7670and @code{ctz} at zero do @emph{not} extend to the builtin functions
7671visible to the user.  Thus one may be free to adjust the value at will
7672to match the target expansion of these operations without fear of
7673breaking the API@.
7674@end defmac
7675
7676@defmac Pmode
7677An alias for the machine mode for pointers.  On most machines, define
7678this to be the integer mode corresponding to the width of a hardware
7679pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
7680On some machines you must define this to be one of the partial integer
7681modes, such as @code{PSImode}.
7682
7683The width of @code{Pmode} must be at least as large as the value of
7684@code{POINTER_SIZE}.  If it is not equal, you must define the macro
7685@code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
7686to @code{Pmode}.
7687@end defmac
7688
7689@defmac FUNCTION_MODE
7690An alias for the machine mode used for memory references to functions
7691being called, in @code{call} RTL expressions.  On most CISC machines,
7692where an instruction can begin at any byte address, this should be
7693@code{QImode}.  On most RISC machines, where all instructions have fixed
7694size and alignment, this should be a mode with the same size and alignment
7695as the machine instruction words - typically @code{SImode} or @code{HImode}.
7696@end defmac
7697
7698@defmac STDC_0_IN_SYSTEM_HEADERS
7699In normal operation, the preprocessor expands @code{__STDC__} to the
7700constant 1, to signify that GCC conforms to ISO Standard C@.  On some
7701hosts, like Solaris, the system compiler uses a different convention,
7702where @code{__STDC__} is normally 0, but is 1 if the user specifies
7703strict conformance to the C Standard.
7704
7705Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
7706convention when processing system header files, but when processing user
7707files @code{__STDC__} will always expand to 1.
7708@end defmac
7709
7710@hook TARGET_C_PREINCLUDE
7711
7712@hook TARGET_CXX_IMPLICIT_EXTERN_C
7713
7714@defmac SYSTEM_IMPLICIT_EXTERN_C
7715Define this macro if the system header files do not support C++@.
7716This macro handles system header files by pretending that system
7717header files are enclosed in @samp{extern "C" @{@dots{}@}}.
7718@end defmac
7719
7720@findex #pragma
7721@findex pragma
7722@defmac REGISTER_TARGET_PRAGMAS ()
7723Define this macro if you want to implement any target-specific pragmas.
7724If defined, it is a C expression which makes a series of calls to
7725@code{c_register_pragma} or @code{c_register_pragma_with_expansion}
7726for each pragma.  The macro may also do any
7727setup required for the pragmas.
7728
7729The primary reason to define this macro is to provide compatibility with
7730other compilers for the same target.  In general, we discourage
7731definition of target-specific pragmas for GCC@.
7732
7733If the pragma can be implemented by attributes then you should consider
7734defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
7735
7736Preprocessor macros that appear on pragma lines are not expanded.  All
7737@samp{#pragma} directives that do not match any registered pragma are
7738silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
7739@end defmac
7740
7741@deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7742@deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
7743
7744Each call to @code{c_register_pragma} or
7745@code{c_register_pragma_with_expansion} establishes one pragma.  The
7746@var{callback} routine will be called when the preprocessor encounters a
7747pragma of the form
7748
7749@smallexample
7750#pragma [@var{space}] @var{name} @dots{}
7751@end smallexample
7752
7753@var{space} is the case-sensitive namespace of the pragma, or
7754@code{NULL} to put the pragma in the global namespace.  The callback
7755routine receives @var{pfile} as its first argument, which can be passed
7756on to cpplib's functions if necessary.  You can lex tokens after the
7757@var{name} by calling @code{pragma_lex}.  Tokens that are not read by the
7758callback will be silently ignored.  The end of the line is indicated by
7759a token of type @code{CPP_EOF}.  Macro expansion occurs on the
7760arguments of pragmas registered with
7761@code{c_register_pragma_with_expansion} but not on the arguments of
7762pragmas registered with @code{c_register_pragma}.
7763
7764Note that the use of @code{pragma_lex} is specific to the C and C++
7765compilers.  It will not work in the Java or Fortran compilers, or any
7766other language compilers for that matter.  Thus if @code{pragma_lex} is going
7767to be called from target-specific code, it must only be done so when
7768building the C and C++ compilers.  This can be done by defining the
7769variables @code{c_target_objs} and @code{cxx_target_objs} in the
7770target entry in the @file{config.gcc} file.  These variables should name
7771the target-specific, language-specific object file which contains the
7772code that uses @code{pragma_lex}.  Note it will also be necessary to add a
7773rule to the makefile fragment pointed to by @code{tmake_file} that shows
7774how to build this object file.
7775@end deftypefun
7776
7777@defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
7778Define this macro if macros should be expanded in the
7779arguments of @samp{#pragma pack}.
7780@end defmac
7781
7782@defmac TARGET_DEFAULT_PACK_STRUCT
7783If your target requires a structure packing default other than 0 (meaning
7784the machine default), define this macro to the necessary value (in bytes).
7785This must be a value that would also be valid to use with
7786@samp{#pragma pack()} (that is, a small power of two).
7787@end defmac
7788
7789@defmac DOLLARS_IN_IDENTIFIERS
7790Define this macro to control use of the character @samp{$} in
7791identifier names for the C family of languages.  0 means @samp{$} is
7792not allowed by default; 1 means it is allowed.  1 is the default;
7793there is no need to define this macro in that case.
7794@end defmac
7795
7796@defmac INSN_SETS_ARE_DELAYED (@var{insn})
7797Define this macro as a C expression that is nonzero if it is safe for the
7798delay slot scheduler to place instructions in the delay slot of @var{insn},
7799even if they appear to use a resource set or clobbered in @var{insn}.
7800@var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
7801every @code{call_insn} has this behavior.  On machines where some @code{insn}
7802or @code{jump_insn} is really a function call and hence has this behavior,
7803you should define this macro.
7804
7805You need not define this macro if it would always return zero.
7806@end defmac
7807
7808@defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
7809Define this macro as a C expression that is nonzero if it is safe for the
7810delay slot scheduler to place instructions in the delay slot of @var{insn},
7811even if they appear to set or clobber a resource referenced in @var{insn}.
7812@var{insn} is always a @code{jump_insn} or an @code{insn}.  On machines where
7813some @code{insn} or @code{jump_insn} is really a function call and its operands
7814are registers whose use is actually in the subroutine it calls, you should
7815define this macro.  Doing so allows the delay slot scheduler to move
7816instructions which copy arguments into the argument registers into the delay
7817slot of @var{insn}.
7818
7819You need not define this macro if it would always return zero.
7820@end defmac
7821
7822@defmac MULTIPLE_SYMBOL_SPACES
7823Define this macro as a C expression that is nonzero if, in some cases,
7824global symbols from one translation unit may not be bound to undefined
7825symbols in another translation unit without user intervention.  For
7826instance, under Microsoft Windows symbols must be explicitly imported
7827from shared libraries (DLLs).
7828
7829You need not define this macro if it would always evaluate to zero.
7830@end defmac
7831
7832@hook TARGET_MD_ASM_ADJUST
7833
7834@defmac MATH_LIBRARY
7835Define this macro as a C string constant for the linker argument to link
7836in the system math library, minus the initial @samp{"-l"}, or
7837@samp{""} if the target does not have a
7838separate math library.
7839
7840You need only define this macro if the default of @samp{"m"} is wrong.
7841@end defmac
7842
7843@defmac LIBRARY_PATH_ENV
7844Define this macro as a C string constant for the environment variable that
7845specifies where the linker should look for libraries.
7846
7847You need only define this macro if the default of @samp{"LIBRARY_PATH"}
7848is wrong.
7849@end defmac
7850
7851@defmac TARGET_POSIX_IO
7852Define this macro if the target supports the following POSIX@ file
7853functions, access, mkdir and  file locking with fcntl / F_SETLKW@.
7854Defining @code{TARGET_POSIX_IO} will enable the test coverage code
7855to use file locking when exiting a program, which avoids race conditions
7856if the program has forked. It will also create directories at run-time
7857for cross-profiling.
7858@end defmac
7859
7860@defmac MAX_CONDITIONAL_EXECUTE
7861
7862A C expression for the maximum number of instructions to execute via
7863conditional execution instructions instead of a branch.  A value of
7864@code{BRANCH_COST}+1 is the default.
7865@end defmac
7866
7867@defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
7868Used if the target needs to perform machine-dependent modifications on the
7869conditionals used for turning basic blocks into conditionally executed code.
7870@var{ce_info} points to a data structure, @code{struct ce_if_block}, which
7871contains information about the currently processed blocks.  @var{true_expr}
7872and @var{false_expr} are the tests that are used for converting the
7873then-block and the else-block, respectively.  Set either @var{true_expr} or
7874@var{false_expr} to a null pointer if the tests cannot be converted.
7875@end defmac
7876
7877@defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
7878Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
7879if-statements into conditions combined by @code{and} and @code{or} operations.
7880@var{bb} contains the basic block that contains the test that is currently
7881being processed and about to be turned into a condition.
7882@end defmac
7883
7884@defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
7885A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
7886be converted to conditional execution format.  @var{ce_info} points to
7887a data structure, @code{struct ce_if_block}, which contains information
7888about the currently processed blocks.
7889@end defmac
7890
7891@defmac IFCVT_MODIFY_FINAL (@var{ce_info})
7892A C expression to perform any final machine dependent modifications in
7893converting code to conditional execution.  The involved basic blocks
7894can be found in the @code{struct ce_if_block} structure that is pointed
7895to by @var{ce_info}.
7896@end defmac
7897
7898@defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
7899A C expression to cancel any machine dependent modifications in
7900converting code to conditional execution.  The involved basic blocks
7901can be found in the @code{struct ce_if_block} structure that is pointed
7902to by @var{ce_info}.
7903@end defmac
7904
7905@defmac IFCVT_MACHDEP_INIT (@var{ce_info})
7906A C expression to initialize any machine specific data for if-conversion
7907of the if-block in the @code{struct ce_if_block} structure that is pointed
7908to by @var{ce_info}.
7909@end defmac
7910
7911@hook TARGET_MACHINE_DEPENDENT_REORG
7912
7913@hook TARGET_INIT_BUILTINS
7914
7915@hook TARGET_BUILTIN_DECL
7916
7917@hook TARGET_EXPAND_BUILTIN
7918
7919@hook TARGET_RESOLVE_OVERLOADED_BUILTIN
7920
7921@hook TARGET_CHECK_BUILTIN_CALL
7922
7923@hook TARGET_FOLD_BUILTIN
7924
7925@hook TARGET_GIMPLE_FOLD_BUILTIN
7926
7927@hook TARGET_COMPARE_VERSION_PRIORITY
7928
7929@hook TARGET_GET_FUNCTION_VERSIONS_DISPATCHER
7930
7931@hook TARGET_GENERATE_VERSION_DISPATCHER_BODY
7932
7933@hook TARGET_PREDICT_DOLOOP_P
7934
7935@hook TARGET_HAVE_COUNT_REG_DECR_P
7936
7937@hook TARGET_DOLOOP_COST_FOR_GENERIC
7938
7939@hook TARGET_DOLOOP_COST_FOR_ADDRESS
7940
7941@hook TARGET_CAN_USE_DOLOOP_P
7942
7943@hook TARGET_INVALID_WITHIN_DOLOOP
7944
7945@hook TARGET_PREFERRED_DOLOOP_MODE
7946
7947@hook TARGET_LEGITIMATE_COMBINED_INSN
7948
7949@hook TARGET_CAN_FOLLOW_JUMP
7950
7951@hook TARGET_COMMUTATIVE_P
7952
7953@hook TARGET_ALLOCATE_INITIAL_VALUE
7954
7955@hook TARGET_UNSPEC_MAY_TRAP_P
7956
7957@hook TARGET_SET_CURRENT_FUNCTION
7958
7959@defmac TARGET_OBJECT_SUFFIX
7960Define this macro to be a C string representing the suffix for object
7961files on your target machine.  If you do not define this macro, GCC will
7962use @samp{.o} as the suffix for object files.
7963@end defmac
7964
7965@defmac TARGET_EXECUTABLE_SUFFIX
7966Define this macro to be a C string representing the suffix to be
7967automatically added to executable files on your target machine.  If you
7968do not define this macro, GCC will use the null string as the suffix for
7969executable files.
7970@end defmac
7971
7972@defmac COLLECT_EXPORT_LIST
7973If defined, @code{collect2} will scan the individual object files
7974specified on its command line and create an export list for the linker.
7975Define this macro for systems like AIX, where the linker discards
7976object files that are not referenced from @code{main} and uses export
7977lists.
7978@end defmac
7979
7980@hook TARGET_CANNOT_MODIFY_JUMPS_P
7981
7982@hook TARGET_HAVE_CONDITIONAL_EXECUTION
7983
7984@hook TARGET_GEN_CCMP_FIRST
7985
7986@hook TARGET_GEN_CCMP_NEXT
7987
7988@hook TARGET_GEN_MEMSET_SCRATCH_RTX
7989
7990@hook TARGET_LOOP_UNROLL_ADJUST
7991
7992@defmac POWI_MAX_MULTS
7993If defined, this macro is interpreted as a signed integer C expression
7994that specifies the maximum number of floating point multiplications
7995that should be emitted when expanding exponentiation by an integer
7996constant inline.  When this value is defined, exponentiation requiring
7997more than this number of multiplications is implemented by calling the
7998system library's @code{pow}, @code{powf} or @code{powl} routines.
7999The default value places no upper bound on the multiplication count.
8000@end defmac
8001
8002@deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8003This target hook should register any extra include files for the
8004target.  The parameter @var{stdinc} indicates if normal include files
8005are present.  The parameter @var{sysroot} is the system root directory.
8006The parameter @var{iprefix} is the prefix for the gcc directory.
8007@end deftypefn
8008
8009@deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
8010This target hook should register any extra include files for the
8011target before any standard headers.  The parameter @var{stdinc}
8012indicates if normal include files are present.  The parameter
8013@var{sysroot} is the system root directory.  The parameter
8014@var{iprefix} is the prefix for the gcc directory.
8015@end deftypefn
8016
8017@deftypefn Macro void TARGET_OPTF (char *@var{path})
8018This target hook should register special include paths for the target.
8019The parameter @var{path} is the include to register.  On Darwin
8020systems, this is used for Framework includes, which have semantics
8021that are different from @option{-I}.
8022@end deftypefn
8023
8024@defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
8025This target macro returns @code{true} if it is safe to use a local alias
8026for a virtual function @var{fndecl} when constructing thunks,
8027@code{false} otherwise.  By default, the macro returns @code{true} for all
8028functions, if a target supports aliases (i.e.@: defines
8029@code{ASM_OUTPUT_DEF}), @code{false} otherwise,
8030@end defmac
8031
8032@defmac TARGET_FORMAT_TYPES
8033If defined, this macro is the name of a global variable containing
8034target-specific format checking information for the @option{-Wformat}
8035option.  The default is to have no target-specific format checks.
8036@end defmac
8037
8038@defmac TARGET_N_FORMAT_TYPES
8039If defined, this macro is the number of entries in
8040@code{TARGET_FORMAT_TYPES}.
8041@end defmac
8042
8043@defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
8044If defined, this macro is the name of a global variable containing
8045target-specific format overrides for the @option{-Wformat} option. The
8046default is to have no target-specific format overrides. If defined,
8047@code{TARGET_FORMAT_TYPES} must be defined, too.
8048@end defmac
8049
8050@defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
8051If defined, this macro specifies the number of entries in
8052@code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
8053@end defmac
8054
8055@defmac TARGET_OVERRIDES_FORMAT_INIT
8056If defined, this macro specifies the optional initialization
8057routine for target specific customizations of the system printf
8058and scanf formatter settings.
8059@end defmac
8060
8061@hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN
8062
8063@hook TARGET_INVALID_CONVERSION
8064
8065@hook TARGET_INVALID_UNARY_OP
8066
8067@hook TARGET_INVALID_BINARY_OP
8068
8069@hook TARGET_PROMOTED_TYPE
8070
8071@hook TARGET_CONVERT_TO_TYPE
8072
8073@hook TARGET_VERIFY_TYPE_CONTEXT
8074
8075@defmac OBJC_JBLEN
8076This macro determines the size of the objective C jump buffer for the
8077NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
8078@end defmac
8079
8080@defmac LIBGCC2_UNWIND_ATTRIBUTE
8081Define this macro if any target-specific attributes need to be attached
8082to the functions in @file{libgcc} that provide low-level support for
8083call stack unwinding.  It is used in declarations in @file{unwind-generic.h}
8084and the associated definitions of those functions.
8085@end defmac
8086
8087@hook TARGET_UPDATE_STACK_BOUNDARY
8088
8089@hook TARGET_GET_DRAP_RTX
8090
8091@hook TARGET_ZERO_CALL_USED_REGS
8092
8093@hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS
8094
8095@hook TARGET_CONST_ANCHOR
8096
8097@hook TARGET_ASAN_SHADOW_OFFSET
8098
8099@hook TARGET_MEMMODEL_CHECK
8100
8101@hook TARGET_ATOMIC_TEST_AND_SET_TRUEVAL
8102
8103@hook TARGET_HAS_IFUNC_P
8104
8105@hook TARGET_IFUNC_REF_LOCAL_OK
8106
8107@hook TARGET_ATOMIC_ALIGN_FOR_MODE
8108
8109@hook TARGET_ATOMIC_ASSIGN_EXPAND_FENV
8110
8111@hook TARGET_RECORD_OFFLOAD_SYMBOL
8112
8113@hook TARGET_OFFLOAD_OPTIONS
8114
8115@defmac TARGET_SUPPORTS_WIDE_INT
8116
8117On older ports, large integers are stored in @code{CONST_DOUBLE} rtl
8118objects.  Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero
8119to indicate that large integers are stored in
8120@code{CONST_WIDE_INT} rtl objects.  The @code{CONST_WIDE_INT} allows
8121very large integer constants to be represented.  @code{CONST_DOUBLE}
8122is limited to twice the size of the host's @code{HOST_WIDE_INT}
8123representation.
8124
8125Converting a port mostly requires looking for the places where
8126@code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that
8127code with code that accesses @code{CONST_WIDE_INT}s.  @samp{"grep -i
8128const_double"} at the port level gets you to 95% of the changes that
8129need to be made.  There are a few places that require a deeper look.
8130
8131@itemize @bullet
8132@item
8133There is no equivalent to @code{hval} and @code{lval} for
8134@code{CONST_WIDE_INT}s.  This would be difficult to express in the md
8135language since there are a variable number of elements.
8136
8137Most ports only check that @code{hval} is either 0 or -1 to see if the
8138value is small.  As mentioned above, this will no longer be necessary
8139since small constants are always @code{CONST_INT}.  Of course there
8140are still a few exceptions, the alpha's constraint used by the zap
8141instruction certainly requires careful examination by C code.
8142However, all the current code does is pass the hval and lval to C
8143code, so evolving the c code to look at the @code{CONST_WIDE_INT} is
8144not really a large change.
8145
8146@item
8147Because there is no standard template that ports use to materialize
8148constants, there is likely to be some futzing that is unique to each
8149port in this code.
8150
8151@item
8152The rtx costs may have to be adjusted to properly account for larger
8153constants that are represented as @code{CONST_WIDE_INT}.
8154@end itemize
8155
8156All and all it does not take long to convert ports that the
8157maintainer is familiar with.
8158
8159@end defmac
8160
8161@hook TARGET_HAVE_SPECULATION_SAFE_VALUE
8162
8163@hook TARGET_SPECULATION_SAFE_VALUE
8164
8165@hook TARGET_RUN_TARGET_SELFTESTS
8166
8167@hook TARGET_MEMTAG_CAN_TAG_ADDRESSES
8168
8169@hook TARGET_MEMTAG_TAG_SIZE
8170
8171@hook TARGET_MEMTAG_GRANULE_SIZE
8172
8173@hook TARGET_MEMTAG_INSERT_RANDOM_TAG
8174
8175@hook TARGET_MEMTAG_ADD_TAG
8176
8177@hook TARGET_MEMTAG_SET_TAG
8178
8179@hook TARGET_MEMTAG_EXTRACT_TAG
8180
8181@hook TARGET_MEMTAG_UNTAGGED_POINTER
8182
8183@hook TARGET_GCOV_TYPE_SIZE
8184
8185@hook TARGET_HAVE_SHADOW_CALL_STACK
8186