]>
Commit | Line | Data |
---|---|---|
a945c346 | 1 | @c Copyright (C) 1988-2024 Free Software Foundation, Inc. |
d77de738 ML |
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 | ||
12 | In addition to the file @file{@var{machine}.md}, a machine description | |
13 | includes a C header file conventionally given the name | |
14 | @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}. | |
15 | The header file defines numerous macros that convey the information | |
16 | about the target machine that does not fit into the scheme of the | |
17 | @file{.md} file. The file @file{tm.h} should be a link to | |
18 | @file{@var{machine}.h}. The header file @file{config.h} includes | |
19 | @file{tm.h} and most compiler source files include @file{config.h}. The | |
20 | source file defines a variable @code{targetm}, which is a structure | |
21 | containing pointers to functions and data relating to the target | |
22 | machine. @file{@var{machine}.c} should also contain their definitions, | |
23 | if they are not defined elsewhere in GCC, and other functions called | |
24 | through the macros defined in the @file{.h} file. | |
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. | |
a7ea90e4 | 56 | * Rust Language and ABI:: Controlling Rust ABI changes. |
d77de738 ML |
57 | * Named Address Spaces:: Adding support for named address spaces |
58 | * Misc:: Everything else. | |
59 | @end menu | |
60 | ||
61 | @node Target Structure | |
62 | @section The Global @code{targetm} Variable | |
63 | @cindex target hooks | |
64 | @cindex target functions | |
65 | ||
66 | @deftypevar {struct gcc_target} targetm | |
67 | The target @file{.c} file must define the global @code{targetm} variable | |
68 | which contains pointers to functions and data relating to the target | |
69 | machine. The variable is declared in @file{target.h}; | |
70 | @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is | |
71 | used to initialize the variable, and macros for the default initializers | |
72 | for elements of the structure. The @file{.c} file should override those | |
73 | macros for which the default definition is inappropriate. For example: | |
74 | @smallexample | |
75 | #include "target.h" | |
76 | #include "target-def.h" | |
77 | ||
78 | /* @r{Initialize the GCC target structure.} */ | |
79 | ||
80 | #undef TARGET_COMP_TYPE_ATTRIBUTES | |
81 | #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes | |
82 | ||
83 | struct gcc_target targetm = TARGET_INITIALIZER; | |
84 | @end smallexample | |
85 | @end deftypevar | |
86 | ||
87 | Where a macro should be defined in the @file{.c} file in this manner to | |
88 | form part of the @code{targetm} structure, it is documented below as a | |
89 | ``Target Hook'' with a prototype. Many macros will change in future | |
90 | from being defined in the @file{.h} file to being part of the | |
91 | @code{targetm} structure. | |
92 | ||
93 | Similarly, there is a @code{targetcm} variable for hooks that are | |
94 | specific to front ends for C-family languages, documented as ``C | |
95 | Target Hook''. This is declared in @file{c-family/c-target.h}, the | |
96 | initializer @code{TARGETCM_INITIALIZER} in | |
97 | @file{c-family/c-target-def.h}. If targets initialize @code{targetcm} | |
98 | themselves, they should set @code{target_has_targetcm=yes} in | |
99 | @file{config.gcc}; otherwise a default definition is used. | |
100 | ||
101 | Similarly, there is a @code{targetm_common} variable for hooks that | |
102 | are shared between the compiler driver and the compilers proper, | |
103 | documented as ``Common Target Hook''. This is declared in | |
104 | @file{common/common-target.h}, the initializer | |
105 | @code{TARGETM_COMMON_INITIALIZER} in | |
106 | @file{common/common-target-def.h}. If targets initialize | |
107 | @code{targetm_common} themselves, they should set | |
108 | @code{target_has_targetm_common=yes} in @file{config.gcc}; otherwise a | |
109 | default definition is used. | |
110 | ||
111 | Similarly, there is a @code{targetdm} variable for hooks that are | |
112 | specific to the D language front end, documented as ``D Target Hook''. | |
113 | This is declared in @file{d/d-target.h}, the initializer | |
114 | @code{TARGETDM_INITIALIZER} in @file{d/d-target-def.h}. If targets | |
115 | initialize @code{targetdm} themselves, they should set | |
116 | @code{target_has_targetdm=yes} in @file{config.gcc}; otherwise a default | |
117 | definition is used. | |
118 | ||
b1c06fd9 IB |
119 | Similarly, there is a @code{targetrustm} variable for hooks that are |
120 | specific to the Rust language front end, documented as ``Rust Target | |
121 | Hook''. This is declared in @file{rust/rust-target.h}, the initializer | |
122 | @code{TARGETRUSTM_INITIALIZER} in @file{rust/rust-target-def.h}. | |
123 | If targets initialize @code{targetrustm} themselves, they should set | |
124 | @code{target_has_targetrustm=yes} in @file{config.gcc}; otherwise a | |
125 | default definition is used. | |
126 | ||
d77de738 ML |
127 | @node Driver |
128 | @section Controlling the Compilation Driver, @file{gcc} | |
129 | @cindex driver | |
130 | @cindex controlling the compilation driver | |
131 | ||
132 | @c prevent bad page break with this line | |
133 | You can control the compilation driver. | |
134 | ||
135 | @defmac DRIVER_SELF_SPECS | |
136 | A list of specs for the driver itself. It should be a suitable | |
137 | initializer for an array of strings, with no surrounding braces. | |
138 | ||
139 | The driver applies these specs to its own command line between loading | |
140 | default @file{specs} files (but not command-line specified ones) and | |
141 | choosing the multilib directory or running any subcommands. It | |
142 | applies them in the order given, so each spec can depend on the | |
143 | options added by earlier ones. It is also possible to remove options | |
144 | using @samp{%<@var{option}} in the usual way. | |
145 | ||
146 | This macro can be useful when a port has several interdependent target | |
147 | options. It provides a way of standardizing the command line so | |
148 | that the other specs are easier to write. | |
149 | ||
150 | Do not define this macro if it does not need to do anything. | |
151 | @end defmac | |
152 | ||
153 | @defmac OPTION_DEFAULT_SPECS | |
154 | A list of specs used to support configure-time default options (i.e.@: | |
155 | @option{--with} options) in the driver. It should be a suitable initializer | |
156 | for an array of structures, each containing two strings, without the | |
157 | outermost pair of surrounding braces. | |
158 | ||
159 | The first item in the pair is the name of the default. This must match | |
160 | the code in @file{config.gcc} for the target. The second item is a spec | |
161 | to apply if a default with this name was specified. The string | |
162 | @samp{%(VALUE)} in the spec will be replaced by the value of the default | |
163 | everywhere it occurs. | |
164 | ||
165 | The driver will apply these specs to its own command line between loading | |
166 | default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using | |
167 | the same mechanism as @code{DRIVER_SELF_SPECS}. | |
168 | ||
169 | Do not define this macro if it does not need to do anything. | |
170 | @end defmac | |
171 | ||
172 | @defmac CPP_SPEC | |
173 | A C string constant that tells the GCC driver program options to | |
174 | pass to CPP@. It can also specify how to translate options you | |
175 | give to GCC into options for GCC to pass to the CPP@. | |
176 | ||
177 | Do not define this macro if it does not need to do anything. | |
178 | @end defmac | |
179 | ||
180 | @defmac CPLUSPLUS_CPP_SPEC | |
181 | This macro is just like @code{CPP_SPEC}, but is used for C++, rather | |
182 | than C@. If you do not define this macro, then the value of | |
183 | @code{CPP_SPEC} (if any) will be used instead. | |
184 | @end defmac | |
185 | ||
186 | @defmac CC1_SPEC | |
187 | A C string constant that tells the GCC driver program options to | |
188 | pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language | |
189 | front ends. | |
190 | It can also specify how to translate options you give to GCC into options | |
191 | for GCC to pass to front ends. | |
192 | ||
193 | Do not define this macro if it does not need to do anything. | |
194 | @end defmac | |
195 | ||
196 | @defmac CC1PLUS_SPEC | |
197 | A C string constant that tells the GCC driver program options to | |
198 | pass to @code{cc1plus}. It can also specify how to translate options you | |
199 | give to GCC into options for GCC to pass to the @code{cc1plus}. | |
200 | ||
201 | Do not define this macro if it does not need to do anything. | |
202 | Note that everything defined in CC1_SPEC is already passed to | |
203 | @code{cc1plus} so there is no need to duplicate the contents of | |
204 | CC1_SPEC in CC1PLUS_SPEC@. | |
205 | @end defmac | |
206 | ||
207 | @defmac ASM_SPEC | |
208 | A C string constant that tells the GCC driver program options to | |
209 | pass to the assembler. It can also specify how to translate options | |
210 | you give to GCC into options for GCC to pass to the assembler. | |
211 | See the file @file{sun3.h} for an example of this. | |
212 | ||
213 | Do not define this macro if it does not need to do anything. | |
214 | @end defmac | |
215 | ||
216 | @defmac ASM_FINAL_SPEC | |
217 | A C string constant that tells the GCC driver program how to | |
218 | run any programs which cleanup after the normal assembler. | |
219 | Normally, this is not needed. See the file @file{mips.h} for | |
220 | an example of this. | |
221 | ||
222 | Do not define this macro if it does not need to do anything. | |
223 | @end defmac | |
224 | ||
225 | @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT | |
226 | Define this macro, with no value, if the driver should give the assembler | |
227 | an argument consisting of a single dash, @option{-}, to instruct it to | |
228 | read from its standard input (which will be a pipe connected to the | |
229 | output of the compiler proper). This argument is given after any | |
230 | @option{-o} option specifying the name of the output file. | |
231 | ||
232 | If you do not define this macro, the assembler is assumed to read its | |
233 | standard input if given no non-option arguments. If your assembler | |
234 | cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct; | |
235 | see @file{mips.h} for instance. | |
236 | @end defmac | |
237 | ||
238 | @defmac LINK_SPEC | |
239 | A C string constant that tells the GCC driver program options to | |
240 | pass to the linker. It can also specify how to translate options you | |
241 | give to GCC into options for GCC to pass to the linker. | |
242 | ||
243 | Do not define this macro if it does not need to do anything. | |
244 | @end defmac | |
245 | ||
246 | @defmac LIB_SPEC | |
247 | Another C string constant used much like @code{LINK_SPEC}. The difference | |
248 | between the two is that @code{LIB_SPEC} is used at the end of the | |
249 | command given to the linker. | |
250 | ||
251 | If this macro is not defined, a default is provided that | |
252 | loads the standard C library from the usual place. See @file{gcc.cc}. | |
253 | @end defmac | |
254 | ||
255 | @defmac LIBGCC_SPEC | |
256 | Another C string constant that tells the GCC driver program | |
257 | how and when to place a reference to @file{libgcc.a} into the | |
258 | linker command line. This constant is placed both before and after | |
259 | the value of @code{LIB_SPEC}. | |
260 | ||
261 | If this macro is not defined, the GCC driver provides a default that | |
262 | passes the string @option{-lgcc} to the linker. | |
263 | @end defmac | |
264 | ||
265 | @defmac REAL_LIBGCC_SPEC | |
266 | By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the | |
267 | @code{LIBGCC_SPEC} is not directly used by the driver program but is | |
268 | instead modified to refer to different versions of @file{libgcc.a} | |
269 | depending on the values of the command line flags @option{-static}, | |
270 | @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On | |
271 | targets where these modifications are inappropriate, define | |
272 | @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the | |
273 | driver how to place a reference to @file{libgcc} on the link command | |
274 | line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified. | |
275 | @end defmac | |
276 | ||
277 | @defmac USE_LD_AS_NEEDED | |
278 | A macro that controls the modifications to @code{LIBGCC_SPEC} | |
279 | mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be | |
280 | generated that uses @option{--as-needed} or equivalent options and the | |
281 | shared @file{libgcc} in place of the | |
282 | static exception handler library, when linking without any of | |
283 | @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}. | |
284 | @end defmac | |
285 | ||
286 | @defmac LINK_EH_SPEC | |
287 | If defined, this C string constant is added to @code{LINK_SPEC}. | |
288 | When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects | |
289 | the modifications to @code{LIBGCC_SPEC} mentioned in | |
290 | @code{REAL_LIBGCC_SPEC}. | |
291 | @end defmac | |
292 | ||
293 | @defmac STARTFILE_SPEC | |
294 | Another C string constant used much like @code{LINK_SPEC}. The | |
295 | difference between the two is that @code{STARTFILE_SPEC} is used at | |
296 | the very beginning of the command given to the linker. | |
297 | ||
298 | If this macro is not defined, a default is provided that loads the | |
299 | standard C startup file from the usual place. See @file{gcc.cc}. | |
300 | @end defmac | |
301 | ||
302 | @defmac ENDFILE_SPEC | |
303 | Another C string constant used much like @code{LINK_SPEC}. The | |
304 | difference between the two is that @code{ENDFILE_SPEC} is used at | |
305 | the very end of the command given to the linker. | |
306 | ||
307 | Do not define this macro if it does not need to do anything. | |
308 | @end defmac | |
309 | ||
310 | @defmac THREAD_MODEL_SPEC | |
311 | GCC @code{-v} will print the thread model GCC was configured to use. | |
312 | However, this doesn't work on platforms that are multilibbed on thread | |
313 | models, such as AIX 4.3. On such platforms, define | |
314 | @code{THREAD_MODEL_SPEC} such that it evaluates to a string without | |
315 | blanks that names one of the recognized thread models. @code{%*}, the | |
316 | default value of this macro, will expand to the value of | |
317 | @code{thread_file} set in @file{config.gcc}. | |
318 | @end defmac | |
319 | ||
320 | @defmac SYSROOT_SUFFIX_SPEC | |
321 | Define this macro to add a suffix to the target sysroot when GCC is | |
322 | configured with a sysroot. This will cause GCC to search for usr/lib, | |
323 | et al, within sysroot+suffix. | |
324 | @end defmac | |
325 | ||
326 | @defmac SYSROOT_HEADERS_SUFFIX_SPEC | |
327 | Define this macro to add a headers_suffix to the target sysroot when | |
328 | GCC is configured with a sysroot. This will cause GCC to pass the | |
329 | updated sysroot+headers_suffix to CPP, causing it to search for | |
330 | usr/include, et al, within sysroot+headers_suffix. | |
331 | @end defmac | |
332 | ||
333 | @defmac EXTRA_SPECS | |
334 | Define this macro to provide additional specifications to put in the | |
335 | @file{specs} file that can be used in various specifications like | |
336 | @code{CC1_SPEC}. | |
337 | ||
338 | The definition should be an initializer for an array of structures, | |
339 | containing a string constant, that defines the specification name, and a | |
340 | string constant that provides the specification. | |
341 | ||
342 | Do not define this macro if it does not need to do anything. | |
343 | ||
344 | @code{EXTRA_SPECS} is useful when an architecture contains several | |
345 | related targets, which have various @code{@dots{}_SPECS} which are similar | |
346 | to each other, and the maintainer would like one central place to keep | |
347 | these definitions. | |
348 | ||
349 | For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to | |
350 | define either @code{_CALL_SYSV} when the System V calling sequence is | |
351 | used or @code{_CALL_AIX} when the older AIX-based calling sequence is | |
352 | used. | |
353 | ||
354 | The @file{config/rs6000/rs6000.h} target file defines: | |
355 | ||
356 | @smallexample | |
357 | #define EXTRA_SPECS \ | |
358 | @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @}, | |
359 | ||
360 | #define CPP_SYS_DEFAULT "" | |
361 | @end smallexample | |
362 | ||
363 | The @file{config/rs6000/sysv.h} target file defines: | |
364 | @smallexample | |
365 | #undef CPP_SPEC | |
366 | #define CPP_SPEC \ | |
367 | "%@{posix: -D_POSIX_SOURCE @} \ | |
368 | %@{mcall-sysv: -D_CALL_SYSV @} \ | |
369 | %@{!mcall-sysv: %(cpp_sysv_default) @} \ | |
370 | %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}" | |
371 | ||
372 | #undef CPP_SYSV_DEFAULT | |
373 | #define CPP_SYSV_DEFAULT "-D_CALL_SYSV" | |
374 | @end smallexample | |
375 | ||
376 | while the @file{config/rs6000/eabiaix.h} target file defines | |
377 | @code{CPP_SYSV_DEFAULT} as: | |
378 | ||
379 | @smallexample | |
380 | #undef CPP_SYSV_DEFAULT | |
381 | #define CPP_SYSV_DEFAULT "-D_CALL_AIX" | |
382 | @end smallexample | |
383 | @end defmac | |
384 | ||
385 | @defmac LINK_LIBGCC_SPECIAL_1 | |
386 | Define this macro if the driver program should find the library | |
387 | @file{libgcc.a}. If you do not define this macro, the driver program will pass | |
388 | the argument @option{-lgcc} to tell the linker to do the search. | |
389 | @end defmac | |
390 | ||
391 | @defmac LINK_GCC_C_SEQUENCE_SPEC | |
392 | The sequence in which libgcc and libc are specified to the linker. | |
393 | By default this is @code{%G %L %G}. | |
394 | @end defmac | |
395 | ||
396 | @defmac POST_LINK_SPEC | |
397 | Define this macro to add additional steps to be executed after linker. | |
398 | The default value of this macro is empty string. | |
399 | @end defmac | |
400 | ||
401 | @defmac LINK_COMMAND_SPEC | |
402 | A C string constant giving the complete command line need to execute the | |
403 | linker. When you do this, you will need to update your port each time a | |
404 | change is made to the link command line within @file{gcc.cc}. Therefore, | |
405 | define this macro only if you need to completely redefine the command | |
406 | line for invoking the linker and there is no other way to accomplish | |
407 | the effect you need. Overriding this macro may be avoidable by overriding | |
408 | @code{LINK_GCC_C_SEQUENCE_SPEC} instead. | |
409 | @end defmac | |
410 | ||
411 | @hook TARGET_ALWAYS_STRIP_DOTDOT | |
412 | ||
413 | @defmac MULTILIB_DEFAULTS | |
414 | Define this macro as a C expression for the initializer of an array of | |
415 | string to tell the driver program which options are defaults for this | |
416 | target and thus do not need to be handled specially when using | |
417 | @code{MULTILIB_OPTIONS}. | |
418 | ||
419 | Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in | |
420 | the target makefile fragment or if none of the options listed in | |
421 | @code{MULTILIB_OPTIONS} are set by default. | |
422 | @xref{Target Fragment}. | |
423 | @end defmac | |
424 | ||
425 | @defmac RELATIVE_PREFIX_NOT_LINKDIR | |
426 | Define this macro to tell @command{gcc} that it should only translate | |
427 | a @option{-B} prefix into a @option{-L} linker option if the prefix | |
428 | indicates an absolute file name. | |
429 | @end defmac | |
430 | ||
431 | @defmac MD_EXEC_PREFIX | |
432 | If defined, this macro is an additional prefix to try after | |
433 | @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched | |
434 | when the compiler is built as a cross | |
435 | compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it | |
436 | to the list of directories used to find the assembler in @file{configure.ac}. | |
437 | @end defmac | |
438 | ||
439 | @defmac STANDARD_STARTFILE_PREFIX | |
440 | Define this macro as a C string constant if you wish to override the | |
441 | standard choice of @code{libdir} as the default prefix to | |
442 | try when searching for startup files such as @file{crt0.o}. | |
443 | @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler | |
444 | is built as a cross compiler. | |
445 | @end defmac | |
446 | ||
447 | @defmac STANDARD_STARTFILE_PREFIX_1 | |
448 | Define this macro as a C string constant if you wish to override the | |
449 | standard choice of @code{/lib} as a prefix to try after the default prefix | |
450 | when searching for startup files such as @file{crt0.o}. | |
451 | @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler | |
452 | is built as a cross compiler. | |
453 | @end defmac | |
454 | ||
455 | @defmac STANDARD_STARTFILE_PREFIX_2 | |
456 | Define this macro as a C string constant if you wish to override the | |
457 | standard choice of @code{/lib} as yet another prefix to try after the | |
458 | default prefix when searching for startup files such as @file{crt0.o}. | |
459 | @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler | |
460 | is built as a cross compiler. | |
461 | @end defmac | |
462 | ||
463 | @defmac MD_STARTFILE_PREFIX | |
464 | If defined, this macro supplies an additional prefix to try after the | |
465 | standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the | |
466 | compiler is built as a cross compiler. | |
467 | @end defmac | |
468 | ||
469 | @defmac MD_STARTFILE_PREFIX_1 | |
470 | If defined, this macro supplies yet another prefix to try after the | |
471 | standard prefixes. It is not searched when the compiler is built as a | |
472 | cross compiler. | |
473 | @end defmac | |
474 | ||
475 | @defmac INIT_ENVIRONMENT | |
476 | Define this macro as a C string constant if you wish to set environment | |
477 | variables for programs called by the driver, such as the assembler and | |
478 | loader. The driver passes the value of this macro to @code{putenv} to | |
479 | initialize the necessary environment variables. | |
480 | @end defmac | |
481 | ||
482 | @defmac LOCAL_INCLUDE_DIR | |
483 | Define this macro as a C string constant if you wish to override the | |
484 | standard choice of @file{/usr/local/include} as the default prefix to | |
485 | try when searching for local header files. @code{LOCAL_INCLUDE_DIR} | |
486 | comes before @code{NATIVE_SYSTEM_HEADER_DIR} (set in | |
487 | @file{config.gcc}, normally @file{/usr/include}) in the search order. | |
488 | ||
489 | Cross compilers do not search either @file{/usr/local/include} or its | |
490 | replacement. | |
491 | @end defmac | |
492 | ||
493 | @defmac NATIVE_SYSTEM_HEADER_COMPONENT | |
494 | The ``component'' corresponding to @code{NATIVE_SYSTEM_HEADER_DIR}. | |
495 | See @code{INCLUDE_DEFAULTS}, below, for the description of components. | |
496 | If you do not define this macro, no component is used. | |
497 | @end defmac | |
498 | ||
499 | @defmac INCLUDE_DEFAULTS | |
500 | Define this macro if you wish to override the entire default search path | |
501 | for include files. For a native compiler, the default search path | |
502 | usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR}, | |
503 | @code{GPLUSPLUS_INCLUDE_DIR}, and | |
504 | @code{NATIVE_SYSTEM_HEADER_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR} | |
505 | and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile}, | |
506 | and specify private search areas for GCC@. The directory | |
507 | @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs. | |
508 | ||
509 | The definition should be an initializer for an array of structures. | |
510 | Each array element should have four elements: the directory name (a | |
511 | string constant), the component name (also a string constant), a flag | |
512 | for C++-only directories, | |
513 | and a flag showing that the includes in the directory don't need to be | |
514 | wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of | |
515 | the array with a null element. | |
516 | ||
517 | The component name denotes what GNU package the include file is part of, | |
518 | if any, in all uppercase letters. For example, it might be @samp{GCC} | |
519 | or @samp{BINUTILS}. If the package is part of a vendor-supplied | |
520 | operating system, code the component name as @samp{0}. | |
521 | ||
522 | For example, here is the definition used for VAX/VMS: | |
523 | ||
524 | @smallexample | |
525 | #define INCLUDE_DEFAULTS \ | |
526 | @{ \ | |
527 | @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \ | |
528 | @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \ | |
529 | @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \ | |
530 | @{ ".", 0, 0, 0@}, \ | |
531 | @{ 0, 0, 0, 0@} \ | |
532 | @} | |
533 | @end smallexample | |
534 | @end defmac | |
535 | ||
536 | Here is the order of prefixes tried for exec files: | |
537 | ||
538 | @enumerate | |
539 | @item | |
540 | Any prefixes specified by the user with @option{-B}. | |
541 | ||
542 | @item | |
543 | The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX} | |
544 | is not set and the compiler has not been installed in the configure-time | |
545 | @var{prefix}, the location in which the compiler has actually been installed. | |
546 | ||
547 | @item | |
548 | The directories specified by the environment variable @code{COMPILER_PATH}. | |
549 | ||
550 | @item | |
551 | The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed | |
552 | in the configured-time @var{prefix}. | |
553 | ||
554 | @item | |
555 | The location @file{/usr/libexec/gcc/}, but only if this is a native compiler. | |
556 | ||
557 | @item | |
558 | The location @file{/usr/lib/gcc/}, but only if this is a native compiler. | |
559 | ||
560 | @item | |
561 | The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native | |
562 | compiler. | |
563 | @end enumerate | |
564 | ||
565 | Here is the order of prefixes tried for startfiles: | |
566 | ||
567 | @enumerate | |
568 | @item | |
569 | Any prefixes specified by the user with @option{-B}. | |
570 | ||
571 | @item | |
572 | The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined | |
573 | value based on the installed toolchain location. | |
574 | ||
575 | @item | |
576 | The directories specified by the environment variable @code{LIBRARY_PATH} | |
577 | (or port-specific name; native only, cross compilers do not use this). | |
578 | ||
579 | @item | |
580 | The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed | |
581 | in the configured @var{prefix} or this is a native compiler. | |
582 | ||
583 | @item | |
584 | The location @file{/usr/lib/gcc/}, but only if this is a native compiler. | |
585 | ||
586 | @item | |
587 | The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native | |
588 | compiler. | |
589 | ||
590 | @item | |
591 | The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a | |
592 | native compiler, or we have a target system root. | |
593 | ||
594 | @item | |
595 | The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a | |
596 | native compiler, or we have a target system root. | |
597 | ||
598 | @item | |
599 | The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications. | |
600 | If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and | |
601 | the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix. | |
602 | ||
603 | @item | |
604 | The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native | |
605 | compiler, or we have a target system root. The default for this macro is | |
606 | @file{/lib/}. | |
607 | ||
608 | @item | |
609 | The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native | |
610 | compiler, or we have a target system root. The default for this macro is | |
611 | @file{/usr/lib/}. | |
612 | @end enumerate | |
613 | ||
614 | @node Run-time Target | |
615 | @section Run-time Target Specification | |
616 | @cindex run-time target specification | |
617 | @cindex predefined macros | |
618 | @cindex target specifications | |
619 | ||
620 | @c prevent bad page break with this line | |
621 | Here are run-time target specifications. | |
622 | ||
623 | @defmac TARGET_CPU_CPP_BUILTINS () | |
624 | This function-like macro expands to a block of code that defines | |
625 | built-in preprocessor macros and assertions for the target CPU, using | |
626 | the functions @code{builtin_define}, @code{builtin_define_std} and | |
627 | @code{builtin_assert}. When the front end | |
628 | calls this macro it provides a trailing semicolon, and since it has | |
629 | finished command line option processing your code can use those | |
630 | results freely. | |
631 | ||
632 | @code{builtin_assert} takes a string in the form you pass to the | |
633 | command-line option @option{-A}, such as @code{cpu=mips}, and creates | |
634 | the assertion. @code{builtin_define} takes a string in the form | |
635 | accepted by option @option{-D} and unconditionally defines the macro. | |
636 | ||
637 | @code{builtin_define_std} takes a string representing the name of an | |
638 | object-like macro. If it doesn't lie in the user's namespace, | |
639 | @code{builtin_define_std} defines it unconditionally. Otherwise, it | |
640 | defines a version with two leading underscores, and another version | |
641 | with two leading and trailing underscores, and defines the original | |
642 | only if an ISO standard was not requested on the command line. For | |
643 | example, passing @code{unix} defines @code{__unix}, @code{__unix__} | |
644 | and possibly @code{unix}; passing @code{_mips} defines @code{__mips}, | |
645 | @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64} | |
646 | defines only @code{_ABI64}. | |
647 | ||
648 | You can also test for the C dialect being compiled. The variable | |
649 | @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus} | |
650 | or @code{clk_objective_c}. Note that if we are preprocessing | |
651 | assembler, this variable will be @code{clk_c} but the function-like | |
652 | macro @code{preprocessing_asm_p()} will return true, so you might want | |
653 | to check for that first. If you need to check for strict ANSI, the | |
654 | variable @code{flag_iso} can be used. The function-like macro | |
655 | @code{preprocessing_trad_p()} can be used to check for traditional | |
656 | preprocessing. | |
657 | @end defmac | |
658 | ||
659 | @defmac TARGET_OS_CPP_BUILTINS () | |
660 | Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional | |
661 | and is used for the target operating system instead. | |
662 | @end defmac | |
663 | ||
664 | @defmac TARGET_OBJFMT_CPP_BUILTINS () | |
665 | Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional | |
666 | and is used for the target object format. @file{elfos.h} uses this | |
667 | macro to define @code{__ELF__}, so you probably do not need to define | |
668 | it yourself. | |
669 | @end defmac | |
670 | ||
671 | @deftypevar {extern int} target_flags | |
672 | This variable is declared in @file{options.h}, which is included before | |
673 | any target-specific headers. | |
674 | @end deftypevar | |
675 | ||
676 | @hook TARGET_DEFAULT_TARGET_FLAGS | |
677 | This variable specifies the initial value of @code{target_flags}. | |
678 | Its default setting is 0. | |
679 | @end deftypevr | |
680 | ||
681 | @cindex optional hardware or system features | |
682 | @cindex features, optional, in system conventions | |
683 | ||
684 | @hook TARGET_HANDLE_OPTION | |
685 | This hook is called whenever the user specifies one of the | |
686 | target-specific options described by the @file{.opt} definition files | |
687 | (@pxref{Options}). It has the opportunity to do some option-specific | |
688 | processing and should return true if the option is valid. The default | |
689 | definition does nothing but return true. | |
690 | ||
691 | @var{decoded} specifies the option and its arguments. @var{opts} and | |
692 | @var{opts_set} are the @code{gcc_options} structures to be used for | |
693 | storing option state, and @var{loc} is the location at which the | |
694 | option was passed (@code{UNKNOWN_LOCATION} except for options passed | |
695 | via attributes). | |
696 | @end deftypefn | |
697 | ||
698 | @hook TARGET_HANDLE_C_OPTION | |
699 | This target hook is called whenever the user specifies one of the | |
700 | target-specific C language family options described by the @file{.opt} | |
701 | definition files(@pxref{Options}). It has the opportunity to do some | |
702 | option-specific processing and should return true if the option is | |
703 | valid. The arguments are like for @code{TARGET_HANDLE_OPTION}. The | |
704 | default definition does nothing but return false. | |
705 | ||
706 | In general, you should use @code{TARGET_HANDLE_OPTION} to handle | |
707 | options. However, if processing an option requires routines that are | |
708 | only available in the C (and related language) front ends, then you | |
709 | should use @code{TARGET_HANDLE_C_OPTION} instead. | |
710 | @end deftypefn | |
711 | ||
712 | @hook TARGET_OBJC_CONSTRUCT_STRING_OBJECT | |
713 | ||
714 | @hook TARGET_OBJC_DECLARE_UNRESOLVED_CLASS_REFERENCE | |
715 | ||
716 | @hook TARGET_OBJC_DECLARE_CLASS_DEFINITION | |
717 | ||
718 | @hook TARGET_STRING_OBJECT_REF_TYPE_P | |
719 | ||
720 | @hook TARGET_CHECK_STRING_OBJECT_FORMAT_ARG | |
721 | ||
722 | @hook TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE | |
723 | ||
724 | @defmac C_COMMON_OVERRIDE_OPTIONS | |
725 | This is similar to the @code{TARGET_OPTION_OVERRIDE} hook | |
726 | but is only used in the C | |
727 | language frontends (C, Objective-C, C++, Objective-C++) and so can be | |
728 | used to alter option flag variables which only exist in those | |
729 | frontends. | |
730 | @end defmac | |
731 | ||
732 | @hook TARGET_OPTION_OPTIMIZATION_TABLE | |
733 | Some machines may desire to change what optimizations are performed for | |
734 | various optimization levels. This variable, if defined, describes | |
735 | options to enable at particular sets of optimization levels. These | |
736 | options are processed once | |
737 | just after the optimization level is determined and before the remainder | |
738 | of the command options have been parsed, so may be overridden by other | |
739 | options passed explicitly. | |
740 | ||
741 | This processing is run once at program startup and when the optimization | |
742 | options are changed via @code{#pragma GCC optimize} or by using the | |
743 | @code{optimize} attribute. | |
744 | @end deftypevr | |
745 | ||
746 | @hook TARGET_OPTION_INIT_STRUCT | |
747 | ||
748 | @hook TARGET_COMPUTE_MULTILIB | |
749 | ||
750 | ||
751 | @defmac SWITCHABLE_TARGET | |
752 | Some targets need to switch between substantially different subtargets | |
753 | during compilation. For example, the MIPS target has one subtarget for | |
754 | the traditional MIPS architecture and another for MIPS16. Source code | |
755 | can switch between these two subarchitectures using the @code{mips16} | |
756 | and @code{nomips16} attributes. | |
757 | ||
758 | Such subtargets can differ in things like the set of available | |
759 | registers, the set of available instructions, the costs of various | |
760 | operations, and so on. GCC caches a lot of this type of information | |
761 | in global variables, and recomputing them for each subtarget takes a | |
762 | significant amount of time. The compiler therefore provides a facility | |
763 | for maintaining several versions of the global variables and quickly | |
764 | switching between them; see @file{target-globals.h} for details. | |
765 | ||
766 | Define this macro to 1 if your target needs this facility. The default | |
767 | is 0. | |
768 | @end defmac | |
769 | ||
770 | @hook TARGET_FLOAT_EXCEPTIONS_ROUNDING_SUPPORTED_P | |
771 | ||
772 | @node Per-Function Data | |
773 | @section Defining data structures for per-function information. | |
774 | @cindex per-function data | |
775 | @cindex data structures | |
776 | ||
777 | If the target needs to store information on a per-function basis, GCC | |
778 | provides a macro and a couple of variables to allow this. Note, just | |
779 | using statics to store the information is a bad idea, since GCC supports | |
780 | nested functions, so you can be halfway through encoding one function | |
781 | when another one comes along. | |
782 | ||
783 | GCC defines a data structure called @code{struct function} which | |
784 | contains all of the data specific to an individual function. This | |
785 | structure contains a field called @code{machine} whose type is | |
786 | @code{struct machine_function *}, which can be used by targets to point | |
787 | to their own specific data. | |
788 | ||
789 | If a target needs per-function specific data it should define the type | |
790 | @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}. | |
791 | This macro should be used to initialize the function pointer | |
792 | @code{init_machine_status}. This pointer is explained below. | |
793 | ||
794 | One typical use of per-function, target specific data is to create an | |
795 | RTX to hold the register containing the function's return address. This | |
796 | RTX can then be used to implement the @code{__builtin_return_address} | |
797 | function, for level 0. | |
798 | ||
799 | Note---earlier implementations of GCC used a single data area to hold | |
800 | all of the per-function information. Thus when processing of a nested | |
801 | function began the old per-function data had to be pushed onto a | |
802 | stack, and when the processing was finished, it had to be popped off the | |
803 | stack. GCC used to provide function pointers called | |
804 | @code{save_machine_status} and @code{restore_machine_status} to handle | |
805 | the saving and restoring of the target specific information. Since the | |
806 | single data area approach is no longer used, these pointers are no | |
807 | longer supported. | |
808 | ||
809 | @defmac INIT_EXPANDERS | |
810 | Macro called to initialize any target specific information. This macro | |
811 | is called once per function, before generation of any RTL has begun. | |
812 | The intention of this macro is to allow the initialization of the | |
813 | function pointer @code{init_machine_status}. | |
814 | @end defmac | |
815 | ||
816 | @deftypevar {void (*)(struct function *)} init_machine_status | |
817 | If this function pointer is non-@code{NULL} it will be called once per | |
818 | function, before function compilation starts, in order to allow the | |
819 | target to perform any target specific initialization of the | |
820 | @code{struct function} structure. It is intended that this would be | |
821 | used to initialize the @code{machine} of that structure. | |
822 | ||
823 | @code{struct machine_function} structures are expected to be freed by GC@. | |
824 | Generally, any memory that they reference must be allocated by using | |
825 | GC allocation, including the structure itself. | |
826 | @end deftypevar | |
827 | ||
828 | @node Storage Layout | |
829 | @section Storage Layout | |
830 | @cindex storage layout | |
831 | ||
832 | Note that the definitions of the macros in this table which are sizes or | |
833 | alignments measured in bits do not need to be constant. They can be C | |
834 | expressions that refer to static variables, such as the @code{target_flags}. | |
835 | @xref{Run-time Target}. | |
836 | ||
837 | @defmac BITS_BIG_ENDIAN | |
838 | Define this macro to have the value 1 if the most significant bit in a | |
839 | byte has the lowest number; otherwise define it to have the value zero. | |
840 | This means that bit-field instructions count from the most significant | |
841 | bit. If the machine has no bit-field instructions, then this must still | |
842 | be defined, but it doesn't matter which value it is defined to. This | |
843 | macro need not be a constant. | |
844 | ||
845 | This macro does not affect the way structure fields are packed into | |
846 | bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}. | |
847 | @end defmac | |
848 | ||
849 | @defmac BYTES_BIG_ENDIAN | |
850 | Define this macro to have the value 1 if the most significant byte in a | |
851 | word has the lowest number. This macro need not be a constant. | |
852 | @end defmac | |
853 | ||
854 | @defmac WORDS_BIG_ENDIAN | |
855 | Define this macro to have the value 1 if, in a multiword object, the | |
856 | most significant word has the lowest number. This applies to both | |
857 | memory locations and registers; see @code{REG_WORDS_BIG_ENDIAN} if the | |
858 | order of words in memory is not the same as the order in registers. This | |
859 | macro need not be a constant. | |
860 | @end defmac | |
861 | ||
862 | @defmac REG_WORDS_BIG_ENDIAN | |
863 | On some machines, the order of words in a multiword object differs between | |
864 | registers in memory. In such a situation, define this macro to describe | |
865 | the order of words in a register. The macro @code{WORDS_BIG_ENDIAN} controls | |
866 | the order of words in memory. | |
867 | @end defmac | |
868 | ||
869 | @defmac FLOAT_WORDS_BIG_ENDIAN | |
870 | Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or | |
871 | @code{TFmode} floating point numbers are stored in memory with the word | |
872 | containing the sign bit at the lowest address; otherwise define it to | |
873 | have the value 0. This macro need not be a constant. | |
874 | ||
875 | You need not define this macro if the ordering is the same as for | |
876 | multi-word integers. | |
877 | @end defmac | |
878 | ||
879 | @defmac BITS_PER_WORD | |
880 | Number of bits in a word. If you do not define this macro, the default | |
881 | is @code{BITS_PER_UNIT * UNITS_PER_WORD}. | |
882 | @end defmac | |
883 | ||
884 | @defmac MAX_BITS_PER_WORD | |
885 | Maximum number of bits in a word. If this is undefined, the default is | |
886 | @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the | |
887 | largest value that @code{BITS_PER_WORD} can have at run-time. | |
888 | @end defmac | |
889 | ||
890 | @defmac UNITS_PER_WORD | |
891 | Number of storage units in a word; normally the size of a general-purpose | |
892 | register, a power of two from 1 or 8. | |
893 | @end defmac | |
894 | ||
895 | @defmac MIN_UNITS_PER_WORD | |
896 | Minimum number of units in a word. If this is undefined, the default is | |
897 | @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the | |
898 | smallest value that @code{UNITS_PER_WORD} can have at run-time. | |
899 | @end defmac | |
900 | ||
901 | @defmac POINTER_SIZE | |
902 | Width of a pointer, in bits. You must specify a value no wider than the | |
903 | width of @code{Pmode}. If it is not equal to the width of @code{Pmode}, | |
904 | you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify | |
905 | a value the default is @code{BITS_PER_WORD}. | |
906 | @end defmac | |
907 | ||
908 | @defmac POINTERS_EXTEND_UNSIGNED | |
909 | A C expression that determines how pointers should be extended from | |
910 | @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is | |
911 | greater than zero if pointers should be zero-extended, zero if they | |
912 | should be sign-extended, and negative if some other sort of conversion | |
913 | is needed. In the last case, the extension is done by the target's | |
914 | @code{ptr_extend} instruction. | |
915 | ||
916 | You need not define this macro if the @code{ptr_mode}, @code{Pmode} | |
917 | and @code{word_mode} are all the same width. | |
918 | @end defmac | |
919 | ||
920 | @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type}) | |
921 | A macro to update @var{m} and @var{unsignedp} when an object whose type | |
922 | is @var{type} and which has the specified mode and signedness is to be | |
923 | stored in a register. This macro is only called when @var{type} is a | |
924 | scalar type. | |
925 | ||
926 | On most RISC machines, which only have operations that operate on a full | |
927 | register, define this macro to set @var{m} to @code{word_mode} if | |
928 | @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most | |
929 | cases, only integer modes should be widened because wider-precision | |
930 | floating-point operations are usually more expensive than their narrower | |
931 | counterparts. | |
932 | ||
933 | For most machines, the macro definition does not change @var{unsignedp}. | |
934 | However, some machines, have instructions that preferentially handle | |
935 | either signed or unsigned quantities of certain modes. For example, on | |
936 | the DEC Alpha, 32-bit loads from memory and 32-bit add instructions | |
937 | sign-extend the result to 64 bits. On such machines, set | |
938 | @var{unsignedp} according to which kind of extension is more efficient. | |
939 | ||
940 | Do not define this macro if it would never modify @var{m}. | |
941 | @end defmac | |
942 | ||
943 | @hook TARGET_C_EXCESS_PRECISION | |
944 | Return a value, with the same meaning as the C99 macro | |
945 | @code{FLT_EVAL_METHOD} that describes which excess precision should be | |
946 | applied. | |
947 | ||
4f4fa250 JJ |
948 | @hook TARGET_C_BITINT_TYPE_INFO |
949 | ||
55947b32 KL |
950 | @hook TARGET_C_MODE_FOR_FLOATING_TYPE |
951 | ||
d77de738 ML |
952 | @hook TARGET_PROMOTE_FUNCTION_MODE |
953 | ||
954 | @defmac PARM_BOUNDARY | |
955 | Normal alignment required for function parameters on the stack, in | |
956 | bits. All stack parameters receive at least this much alignment | |
957 | regardless of data type. On most machines, this is the same as the | |
958 | size of an integer. | |
959 | @end defmac | |
960 | ||
961 | @defmac STACK_BOUNDARY | |
962 | Define this macro to the minimum alignment enforced by hardware for the | |
963 | stack pointer on this machine. The definition is a C expression for the | |
964 | desired alignment (measured in bits). This value is used as a default | |
965 | if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines, | |
966 | this should be the same as @code{PARM_BOUNDARY}. | |
967 | @end defmac | |
968 | ||
969 | @defmac PREFERRED_STACK_BOUNDARY | |
970 | Define this macro if you wish to preserve a certain alignment for the | |
971 | stack pointer, greater than what the hardware enforces. The definition | |
972 | is a C expression for the desired alignment (measured in bits). This | |
973 | macro must evaluate to a value equal to or larger than | |
974 | @code{STACK_BOUNDARY}. | |
975 | @end defmac | |
976 | ||
977 | @defmac INCOMING_STACK_BOUNDARY | |
978 | Define this macro if the incoming stack boundary may be different | |
979 | from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate | |
980 | to a value equal to or larger than @code{STACK_BOUNDARY}. | |
981 | @end defmac | |
982 | ||
983 | @defmac FUNCTION_BOUNDARY | |
984 | Alignment required for a function entry point, in bits. | |
985 | @end defmac | |
986 | ||
987 | @defmac BIGGEST_ALIGNMENT | |
988 | Biggest alignment that any data type can require on this machine, in | |
989 | bits. Note that this is not the biggest alignment that is supported, | |
990 | just the biggest alignment that, when violated, may cause a fault. | |
991 | @end defmac | |
992 | ||
993 | @hook TARGET_ABSOLUTE_BIGGEST_ALIGNMENT | |
994 | ||
995 | @defmac MALLOC_ABI_ALIGNMENT | |
996 | Alignment, in bits, a C conformant malloc implementation has to | |
997 | provide. If not defined, the default value is @code{BITS_PER_WORD}. | |
998 | @end defmac | |
999 | ||
1000 | @defmac ATTRIBUTE_ALIGNED_VALUE | |
1001 | Alignment used by the @code{__attribute__ ((aligned))} construct. If | |
1002 | not defined, the default value is @code{BIGGEST_ALIGNMENT}. | |
1003 | @end defmac | |
1004 | ||
1005 | @defmac MINIMUM_ATOMIC_ALIGNMENT | |
1006 | If defined, the smallest alignment, in bits, that can be given to an | |
1007 | object that can be referenced in one operation, without disturbing any | |
1008 | nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger | |
1009 | on machines that don't have byte or half-word store operations. | |
1010 | @end defmac | |
1011 | ||
1012 | @defmac BIGGEST_FIELD_ALIGNMENT | |
1013 | Biggest alignment that any structure or union field can require on this | |
1014 | machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for | |
1015 | structure and union fields only, unless the field alignment has been set | |
1016 | by the @code{__attribute__ ((aligned (@var{n})))} construct. | |
1017 | @end defmac | |
1018 | ||
1019 | @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{type}, @var{computed}) | |
1020 | An expression for the alignment of a structure field @var{field} of | |
1021 | type @var{type} if the alignment computed in the usual way (including | |
1022 | applying of @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the | |
1023 | alignment) is @var{computed}. It overrides alignment only if the | |
1024 | field alignment has not been set by the | |
1025 | @code{__attribute__ ((aligned (@var{n})))} construct. Note that @var{field} | |
1026 | may be @code{NULL_TREE} in case we just query for the minimum alignment | |
1027 | of a field of type @var{type} in structure context. | |
1028 | @end defmac | |
1029 | ||
1030 | @defmac MAX_STACK_ALIGNMENT | |
1031 | Biggest stack alignment guaranteed by the backend. Use this macro | |
1032 | to specify the maximum alignment of a variable on stack. | |
1033 | ||
1034 | If not defined, the default value is @code{STACK_BOUNDARY}. | |
1035 | ||
1036 | @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}. | |
1037 | @c But the fix for PR 32893 indicates that we can only guarantee | |
1038 | @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not | |
1039 | @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported. | |
1040 | @end defmac | |
1041 | ||
1042 | @defmac MAX_OFILE_ALIGNMENT | |
1043 | Biggest alignment supported by the object file format of this machine. | |
1044 | Use this macro to limit the alignment which can be specified using the | |
1045 | @code{__attribute__ ((aligned (@var{n})))} construct for functions and | |
1046 | objects with static storage duration. The alignment of automatic | |
1047 | objects may exceed the object file format maximum up to the maximum | |
1048 | supported by GCC. If not defined, the default value is | |
1049 | @code{BIGGEST_ALIGNMENT}. | |
1050 | ||
1051 | On systems that use ELF, the default (in @file{config/elfos.h}) is | |
1052 | the largest supported 32-bit ELF section alignment representable on | |
1053 | a 32-bit host e.g.@: @samp{(((uint64_t) 1 << 28) * 8)}. | |
1054 | On 32-bit ELF the largest supported section alignment in bits is | |
1055 | @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts. | |
1056 | @end defmac | |
1057 | ||
1058 | @hook TARGET_LOWER_LOCAL_DECL_ALIGNMENT | |
1059 | ||
1060 | @hook TARGET_STATIC_RTX_ALIGNMENT | |
1061 | ||
1062 | @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align}) | |
1063 | If defined, a C expression to compute the alignment for a variable in | |
1064 | the static store. @var{type} is the data type, and @var{basic-align} is | |
1065 | the alignment that the object would ordinarily have. The value of this | |
1066 | macro is used instead of that alignment to align the object. | |
1067 | ||
1068 | If this macro is not defined, then @var{basic-align} is used. | |
1069 | ||
1070 | @findex strcpy | |
1071 | One use of this macro is to increase alignment of medium-size data to | |
1072 | make it all fit in fewer cache lines. Another is to cause character | |
1073 | arrays to be word-aligned so that @code{strcpy} calls that copy | |
1074 | constants to character arrays can be done inline. | |
1075 | @end defmac | |
1076 | ||
1077 | @defmac DATA_ABI_ALIGNMENT (@var{type}, @var{basic-align}) | |
1078 | Similar to @code{DATA_ALIGNMENT}, but for the cases where the ABI mandates | |
1079 | some alignment increase, instead of optimization only purposes. E.g.@ | |
1080 | AMD x86-64 psABI says that variables with array type larger than 15 bytes | |
1081 | must be aligned to 16 byte boundaries. | |
1082 | ||
1083 | If this macro is not defined, then @var{basic-align} is used. | |
1084 | @end defmac | |
1085 | ||
1086 | @hook TARGET_CONSTANT_ALIGNMENT | |
1087 | ||
1088 | @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align}) | |
1089 | If defined, a C expression to compute the alignment for a variable in | |
1090 | the local store. @var{type} is the data type, and @var{basic-align} is | |
1091 | the alignment that the object would ordinarily have. The value of this | |
1092 | macro is used instead of that alignment to align the object. | |
1093 | ||
1094 | If this macro is not defined, then @var{basic-align} is used. | |
1095 | ||
1096 | One use of this macro is to increase alignment of medium-size data to | |
1097 | make it all fit in fewer cache lines. | |
1098 | ||
1099 | If the value of this macro has a type, it should be an unsigned type. | |
1100 | @end defmac | |
1101 | ||
1102 | @hook TARGET_VECTOR_ALIGNMENT | |
1103 | ||
1104 | @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align}) | |
1105 | If defined, a C expression to compute the alignment for stack slot. | |
1106 | @var{type} is the data type, @var{mode} is the widest mode available, | |
1107 | and @var{basic-align} is the alignment that the slot would ordinarily | |
1108 | have. The value of this macro is used instead of that alignment to | |
1109 | align the slot. | |
1110 | ||
1111 | If this macro is not defined, then @var{basic-align} is used when | |
1112 | @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will | |
1113 | be used. | |
1114 | ||
1115 | This macro is to set alignment of stack slot to the maximum alignment | |
1116 | of all possible modes which the slot may have. | |
1117 | ||
1118 | If the value of this macro has a type, it should be an unsigned type. | |
1119 | @end defmac | |
1120 | ||
1121 | @defmac LOCAL_DECL_ALIGNMENT (@var{decl}) | |
1122 | If defined, a C expression to compute the alignment for a local | |
1123 | variable @var{decl}. | |
1124 | ||
1125 | If this macro is not defined, then | |
1126 | @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))} | |
1127 | is used. | |
1128 | ||
1129 | One use of this macro is to increase alignment of medium-size data to | |
1130 | make it all fit in fewer cache lines. | |
1131 | ||
1132 | If the value of this macro has a type, it should be an unsigned type. | |
1133 | @end defmac | |
1134 | ||
1135 | @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align}) | |
1136 | If defined, a C expression to compute the minimum required alignment | |
1137 | for dynamic stack realignment purposes for @var{exp} (a type or decl), | |
1138 | @var{mode}, assuming normal alignment @var{align}. | |
1139 | ||
1140 | If this macro is not defined, then @var{align} will be used. | |
1141 | @end defmac | |
1142 | ||
1143 | @defmac EMPTY_FIELD_BOUNDARY | |
1144 | Alignment in bits to be given to a structure bit-field that follows an | |
1145 | empty field such as @code{int : 0;}. | |
1146 | ||
1147 | If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro. | |
1148 | @end defmac | |
1149 | ||
1150 | @defmac STRUCTURE_SIZE_BOUNDARY | |
1151 | Number of bits which any structure or union's size must be a multiple of. | |
1152 | Each structure or union's size is rounded up to a multiple of this. | |
1153 | ||
1154 | If you do not define this macro, the default is the same as | |
1155 | @code{BITS_PER_UNIT}. | |
1156 | @end defmac | |
1157 | ||
1158 | @defmac STRICT_ALIGNMENT | |
1159 | Define this macro to be the value 1 if instructions will fail to work | |
1160 | if given data not on the nominal alignment. If instructions will merely | |
1161 | go slower in that case, define this macro as 0. | |
1162 | @end defmac | |
1163 | ||
1164 | @defmac PCC_BITFIELD_TYPE_MATTERS | |
1165 | Define this if you wish to imitate the way many other C compilers handle | |
1166 | alignment of bit-fields and the structures that contain them. | |
1167 | ||
1168 | The behavior is that the type written for a named bit-field (@code{int}, | |
1169 | @code{short}, or other integer type) imposes an alignment for the entire | |
1170 | structure, as if the structure really did contain an ordinary field of | |
1171 | that type. In addition, the bit-field is placed within the structure so | |
1172 | that it would fit within such a field, not crossing a boundary for it. | |
1173 | ||
1174 | Thus, on most machines, a named bit-field whose type is written as | |
1175 | @code{int} would not cross a four-byte boundary, and would force | |
1176 | four-byte alignment for the whole structure. (The alignment used may | |
1177 | not be four bytes; it is controlled by the other alignment parameters.) | |
1178 | ||
1179 | An unnamed bit-field will not affect the alignment of the containing | |
1180 | structure. | |
1181 | ||
1182 | If the macro is defined, its definition should be a C expression; | |
1183 | a nonzero value for the expression enables this behavior. | |
1184 | ||
1185 | Note that if this macro is not defined, or its value is zero, some | |
1186 | bit-fields may cross more than one alignment boundary. The compiler can | |
1187 | support such references if there are @samp{insv}, @samp{extv}, and | |
1188 | @samp{extzv} insns that can directly reference memory. | |
1189 | ||
1190 | The other known way of making bit-fields work is to define | |
1191 | @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}. | |
1192 | Then every structure can be accessed with fullwords. | |
1193 | ||
1194 | Unless the machine has bit-field instructions or you define | |
1195 | @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define | |
1196 | @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value. | |
1197 | ||
1198 | If your aim is to make GCC use the same conventions for laying out | |
1199 | bit-fields as are used by another compiler, here is how to investigate | |
1200 | what the other compiler does. Compile and run this program: | |
1201 | ||
1202 | @smallexample | |
1203 | struct foo1 | |
1204 | @{ | |
1205 | char x; | |
1206 | char :0; | |
1207 | char y; | |
1208 | @}; | |
1209 | ||
1210 | struct foo2 | |
1211 | @{ | |
1212 | char x; | |
1213 | int :0; | |
1214 | char y; | |
1215 | @}; | |
1216 | ||
1217 | main () | |
1218 | @{ | |
1219 | printf ("Size of foo1 is %d\n", | |
1220 | sizeof (struct foo1)); | |
1221 | printf ("Size of foo2 is %d\n", | |
1222 | sizeof (struct foo2)); | |
1223 | exit (0); | |
1224 | @} | |
1225 | @end smallexample | |
1226 | ||
1227 | If this prints 2 and 5, then the compiler's behavior is what you would | |
1228 | get from @code{PCC_BITFIELD_TYPE_MATTERS}. | |
1229 | @end defmac | |
1230 | ||
1231 | @defmac BITFIELD_NBYTES_LIMITED | |
1232 | Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited | |
1233 | to aligning a bit-field within the structure. | |
1234 | @end defmac | |
1235 | ||
1236 | @hook TARGET_ALIGN_ANON_BITFIELD | |
1237 | ||
1238 | @hook TARGET_NARROW_VOLATILE_BITFIELD | |
1239 | ||
1240 | @hook TARGET_MEMBER_TYPE_FORCES_BLK | |
1241 | ||
1242 | @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified}) | |
1243 | Define this macro as an expression for the alignment of a type (given | |
1244 | by @var{type} as a tree node) if the alignment computed in the usual | |
1245 | way is @var{computed} and the alignment explicitly specified was | |
1246 | @var{specified}. | |
1247 | ||
1248 | The default is to use @var{specified} if it is larger; otherwise, use | |
1249 | the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT} | |
1250 | @end defmac | |
1251 | ||
1252 | @defmac MAX_FIXED_MODE_SIZE | |
1253 | An integer expression for the size in bits of the largest integer | |
1254 | machine mode that should actually be used. All integer machine modes of | |
1255 | this size or smaller can be used for structures and unions with the | |
1256 | appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE | |
1257 | (DImode)} is assumed. | |
1258 | @end defmac | |
1259 | ||
1260 | @defmac STACK_SAVEAREA_MODE (@var{save_level}) | |
1261 | If defined, an expression of type @code{machine_mode} that | |
1262 | specifies the mode of the save area operand of a | |
1263 | @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}). | |
1264 | @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or | |
1265 | @code{SAVE_NONLOCAL} and selects which of the three named patterns is | |
1266 | having its mode specified. | |
1267 | ||
1268 | You need not define this macro if it always returns @code{Pmode}. You | |
1269 | would most commonly define this macro if the | |
1270 | @code{save_stack_@var{level}} patterns need to support both a 32- and a | |
1271 | 64-bit mode. | |
1272 | @end defmac | |
1273 | ||
1274 | @defmac STACK_SIZE_MODE | |
1275 | If defined, an expression of type @code{machine_mode} that | |
1276 | specifies the mode of the size increment operand of an | |
1277 | @code{allocate_stack} named pattern (@pxref{Standard Names}). | |
1278 | ||
1279 | You need not define this macro if it always returns @code{word_mode}. | |
1280 | You would most commonly define this macro if the @code{allocate_stack} | |
1281 | pattern needs to support both a 32- and a 64-bit mode. | |
1282 | @end defmac | |
1283 | ||
1284 | @hook TARGET_LIBGCC_CMP_RETURN_MODE | |
1285 | ||
1286 | @hook TARGET_LIBGCC_SHIFT_COUNT_MODE | |
1287 | ||
1288 | @hook TARGET_UNWIND_WORD_MODE | |
1289 | ||
1290 | @hook TARGET_MS_BITFIELD_LAYOUT_P | |
1291 | ||
1292 | @hook TARGET_DECIMAL_FLOAT_SUPPORTED_P | |
1293 | ||
1294 | @hook TARGET_FIXED_POINT_SUPPORTED_P | |
1295 | ||
1296 | @hook TARGET_EXPAND_TO_RTL_HOOK | |
1297 | ||
1298 | @hook TARGET_INSTANTIATE_DECLS | |
1299 | ||
1300 | @hook TARGET_MANGLE_TYPE | |
1301 | ||
6978df2c JJ |
1302 | @hook TARGET_EMIT_SUPPORT_TINFOS |
1303 | ||
d77de738 ML |
1304 | @node Type Layout |
1305 | @section Layout of Source Language Data Types | |
1306 | ||
1307 | These macros define the sizes and other characteristics of the standard | |
1308 | basic data types used in programs being compiled. Unlike the macros in | |
1309 | the previous section, these apply to specific features of C and related | |
1310 | languages, rather than to fundamental aspects of storage layout. | |
1311 | ||
1312 | @defmac INT_TYPE_SIZE | |
1313 | A C expression for the size in bits of the type @code{int} on the | |
1314 | target machine. If you don't define this, the default is one word. | |
1315 | @end defmac | |
1316 | ||
1317 | @defmac SHORT_TYPE_SIZE | |
1318 | A C expression for the size in bits of the type @code{short} on the | |
1319 | target machine. If you don't define this, the default is half a word. | |
1320 | (If this would be less than one storage unit, it is rounded up to one | |
1321 | unit.) | |
1322 | @end defmac | |
1323 | ||
1324 | @defmac LONG_TYPE_SIZE | |
1325 | A C expression for the size in bits of the type @code{long} on the | |
1326 | target machine. If you don't define this, the default is one word. | |
1327 | @end defmac | |
1328 | ||
1329 | @defmac ADA_LONG_TYPE_SIZE | |
1330 | On some machines, the size used for the Ada equivalent of the type | |
1331 | @code{long} by a native Ada compiler differs from that used by C@. In | |
1332 | that situation, define this macro to be a C expression to be used for | |
1333 | the size of that type. If you don't define this, the default is the | |
1334 | value of @code{LONG_TYPE_SIZE}. | |
1335 | @end defmac | |
1336 | ||
1337 | @defmac LONG_LONG_TYPE_SIZE | |
1338 | A C expression for the size in bits of the type @code{long long} on the | |
1339 | target machine. If you don't define this, the default is two | |
1340 | words. If you want to support GNU Ada on your machine, the value of this | |
1341 | macro must be at least 64. | |
1342 | @end defmac | |
1343 | ||
1344 | @defmac CHAR_TYPE_SIZE | |
1345 | A C expression for the size in bits of the type @code{char} on the | |
1346 | target machine. If you don't define this, the default is | |
1347 | @code{BITS_PER_UNIT}. | |
1348 | @end defmac | |
1349 | ||
1350 | @defmac BOOL_TYPE_SIZE | |
1351 | A C expression for the size in bits of the C++ type @code{bool} and | |
1352 | C99 type @code{_Bool} on the target machine. If you don't define | |
1353 | this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}. | |
1354 | @end defmac | |
1355 | ||
d77de738 ML |
1356 | @defmac SHORT_FRACT_TYPE_SIZE |
1357 | A C expression for the size in bits of the type @code{short _Fract} on | |
1358 | the target machine. If you don't define this, the default is | |
1359 | @code{BITS_PER_UNIT}. | |
1360 | @end defmac | |
1361 | ||
1362 | @defmac FRACT_TYPE_SIZE | |
1363 | A C expression for the size in bits of the type @code{_Fract} on | |
1364 | the target machine. If you don't define this, the default is | |
1365 | @code{BITS_PER_UNIT * 2}. | |
1366 | @end defmac | |
1367 | ||
1368 | @defmac LONG_FRACT_TYPE_SIZE | |
1369 | A C expression for the size in bits of the type @code{long _Fract} on | |
1370 | the target machine. If you don't define this, the default is | |
1371 | @code{BITS_PER_UNIT * 4}. | |
1372 | @end defmac | |
1373 | ||
1374 | @defmac LONG_LONG_FRACT_TYPE_SIZE | |
1375 | A C expression for the size in bits of the type @code{long long _Fract} on | |
1376 | the target machine. If you don't define this, the default is | |
1377 | @code{BITS_PER_UNIT * 8}. | |
1378 | @end defmac | |
1379 | ||
1380 | @defmac SHORT_ACCUM_TYPE_SIZE | |
1381 | A C expression for the size in bits of the type @code{short _Accum} on | |
1382 | the target machine. If you don't define this, the default is | |
1383 | @code{BITS_PER_UNIT * 2}. | |
1384 | @end defmac | |
1385 | ||
1386 | @defmac ACCUM_TYPE_SIZE | |
1387 | A C expression for the size in bits of the type @code{_Accum} on | |
1388 | the target machine. If you don't define this, the default is | |
1389 | @code{BITS_PER_UNIT * 4}. | |
1390 | @end defmac | |
1391 | ||
1392 | @defmac LONG_ACCUM_TYPE_SIZE | |
1393 | A C expression for the size in bits of the type @code{long _Accum} on | |
1394 | the target machine. If you don't define this, the default is | |
1395 | @code{BITS_PER_UNIT * 8}. | |
1396 | @end defmac | |
1397 | ||
1398 | @defmac LONG_LONG_ACCUM_TYPE_SIZE | |
1399 | A C expression for the size in bits of the type @code{long long _Accum} on | |
1400 | the target machine. If you don't define this, the default is | |
1401 | @code{BITS_PER_UNIT * 16}. | |
1402 | @end defmac | |
1403 | ||
1404 | @defmac LIBGCC2_GNU_PREFIX | |
1405 | This macro corresponds to the @code{TARGET_LIBFUNC_GNU_PREFIX} target | |
1406 | hook and should be defined if that hook is overriden to be true. It | |
1407 | causes function names in libgcc to be changed to use a @code{__gnu_} | |
1408 | prefix for their name rather than the default @code{__}. A port which | |
1409 | uses this macro should also arrange to use @file{t-gnu-prefix} in | |
1410 | the libgcc @file{config.host}. | |
1411 | @end defmac | |
1412 | ||
1413 | @defmac WIDEST_HARDWARE_FP_SIZE | |
1414 | A C expression for the size in bits of the widest floating-point format | |
1415 | supported by the hardware. If you define this macro, you must specify a | |
55947b32 KL |
1416 | value less than or equal to mode precision of the mode used for C type |
1417 | @code{long double} (from hook @code{targetm.c.mode_for_floating_type} | |
1418 | with argument @code{TI_LONG_DOUBLE_TYPE}). If you do not define this | |
1419 | macro, mode precision of the mode used for C type @code{long double} is | |
1420 | the default. | |
d77de738 ML |
1421 | @end defmac |
1422 | ||
1423 | @defmac DEFAULT_SIGNED_CHAR | |
1424 | An expression whose value is 1 or 0, according to whether the type | |
1425 | @code{char} should be signed or unsigned by default. The user can | |
1426 | always override this default with the options @option{-fsigned-char} | |
1427 | and @option{-funsigned-char}. | |
1428 | @end defmac | |
1429 | ||
1430 | @hook TARGET_DEFAULT_SHORT_ENUMS | |
1431 | ||
1432 | @defmac SIZE_TYPE | |
1433 | A C expression for a string describing the name of the data type to use | |
1434 | for size values. The typedef name @code{size_t} is defined using the | |
1435 | contents of the string. | |
1436 | ||
1437 | The string can contain more than one keyword. If so, separate them with | |
1438 | spaces, and write first any length keyword, then @code{unsigned} if | |
1439 | appropriate, and finally @code{int}. The string must exactly match one | |
1440 | of the data type names defined in the function | |
1441 | @code{c_common_nodes_and_builtins} in the file @file{c-family/c-common.cc}. | |
1442 | You may not omit @code{int} or change the order---that would cause the | |
1443 | compiler to crash on startup. | |
1444 | ||
1445 | If you don't define this macro, the default is @code{"long unsigned | |
1446 | int"}. | |
1447 | @end defmac | |
1448 | ||
1449 | @defmac SIZETYPE | |
1450 | GCC defines internal types (@code{sizetype}, @code{ssizetype}, | |
1451 | @code{bitsizetype} and @code{sbitsizetype}) for expressions | |
1452 | dealing with size. This macro is a C expression for a string describing | |
1453 | the name of the data type from which the precision of @code{sizetype} | |
1454 | is extracted. | |
1455 | ||
1456 | The string has the same restrictions as @code{SIZE_TYPE} string. | |
1457 | ||
1458 | If you don't define this macro, the default is @code{SIZE_TYPE}. | |
1459 | @end defmac | |
1460 | ||
1461 | @defmac PTRDIFF_TYPE | |
1462 | A C expression for a string describing the name of the data type to use | |
1463 | for the result of subtracting two pointers. The typedef name | |
1464 | @code{ptrdiff_t} is defined using the contents of the string. See | |
1465 | @code{SIZE_TYPE} above for more information. | |
1466 | ||
1467 | If you don't define this macro, the default is @code{"long int"}. | |
1468 | @end defmac | |
1469 | ||
1470 | @defmac WCHAR_TYPE | |
1471 | A C expression for a string describing the name of the data type to use | |
1472 | for wide characters. The typedef name @code{wchar_t} is defined using | |
1473 | the contents of the string. See @code{SIZE_TYPE} above for more | |
1474 | information. | |
1475 | ||
1476 | If you don't define this macro, the default is @code{"int"}. | |
1477 | @end defmac | |
1478 | ||
1479 | @defmac WCHAR_TYPE_SIZE | |
1480 | A C expression for the size in bits of the data type for wide | |
1481 | characters. This is used in @code{cpp}, which cannot make use of | |
1482 | @code{WCHAR_TYPE}. | |
1483 | @end defmac | |
1484 | ||
1485 | @defmac WINT_TYPE | |
1486 | A C expression for a string describing the name of the data type to | |
1487 | use for wide characters passed to @code{printf} and returned from | |
1488 | @code{getwc}. The typedef name @code{wint_t} is defined using the | |
1489 | contents of the string. See @code{SIZE_TYPE} above for more | |
1490 | information. | |
1491 | ||
1492 | If you don't define this macro, the default is @code{"unsigned int"}. | |
1493 | @end defmac | |
1494 | ||
1495 | @defmac INTMAX_TYPE | |
1496 | A C expression for a string describing the name of the data type that | |
1497 | can represent any value of any standard or extended signed integer type. | |
1498 | The typedef name @code{intmax_t} is defined using the contents of the | |
1499 | string. See @code{SIZE_TYPE} above for more information. | |
1500 | ||
1501 | If you don't define this macro, the default is the first of | |
1502 | @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as | |
1503 | much precision as @code{long long int}. | |
1504 | @end defmac | |
1505 | ||
1506 | @defmac UINTMAX_TYPE | |
1507 | A C expression for a string describing the name of the data type that | |
1508 | can represent any value of any standard or extended unsigned integer | |
1509 | type. The typedef name @code{uintmax_t} is defined using the contents | |
1510 | of the string. See @code{SIZE_TYPE} above for more information. | |
1511 | ||
1512 | If you don't define this macro, the default is the first of | |
1513 | @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long | |
1514 | unsigned int"} that has as much precision as @code{long long unsigned | |
1515 | int}. | |
1516 | @end defmac | |
1517 | ||
1518 | @defmac SIG_ATOMIC_TYPE | |
1519 | @defmacx INT8_TYPE | |
1520 | @defmacx INT16_TYPE | |
1521 | @defmacx INT32_TYPE | |
1522 | @defmacx INT64_TYPE | |
1523 | @defmacx UINT8_TYPE | |
1524 | @defmacx UINT16_TYPE | |
1525 | @defmacx UINT32_TYPE | |
1526 | @defmacx UINT64_TYPE | |
1527 | @defmacx INT_LEAST8_TYPE | |
1528 | @defmacx INT_LEAST16_TYPE | |
1529 | @defmacx INT_LEAST32_TYPE | |
1530 | @defmacx INT_LEAST64_TYPE | |
1531 | @defmacx UINT_LEAST8_TYPE | |
1532 | @defmacx UINT_LEAST16_TYPE | |
1533 | @defmacx UINT_LEAST32_TYPE | |
1534 | @defmacx UINT_LEAST64_TYPE | |
1535 | @defmacx INT_FAST8_TYPE | |
1536 | @defmacx INT_FAST16_TYPE | |
1537 | @defmacx INT_FAST32_TYPE | |
1538 | @defmacx INT_FAST64_TYPE | |
1539 | @defmacx UINT_FAST8_TYPE | |
1540 | @defmacx UINT_FAST16_TYPE | |
1541 | @defmacx UINT_FAST32_TYPE | |
1542 | @defmacx UINT_FAST64_TYPE | |
1543 | @defmacx INTPTR_TYPE | |
1544 | @defmacx UINTPTR_TYPE | |
1545 | C expressions for the standard types @code{sig_atomic_t}, | |
1546 | @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t}, | |
1547 | @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t}, | |
1548 | @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t}, | |
1549 | @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t}, | |
1550 | @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t}, | |
1551 | @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t}, | |
1552 | @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t}, | |
1553 | @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See | |
1554 | @code{SIZE_TYPE} above for more information. | |
1555 | ||
1556 | If any of these macros evaluates to a null pointer, the corresponding | |
1557 | type is not supported; if GCC is configured to provide | |
1558 | @code{<stdint.h>} in such a case, the header provided may not conform | |
1559 | to C99, depending on the type in question. The defaults for all of | |
1560 | these macros are null pointers. | |
1561 | @end defmac | |
1562 | ||
1563 | @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION | |
1564 | The C++ compiler represents a pointer-to-member-function with a struct | |
1565 | that looks like: | |
1566 | ||
1567 | @smallexample | |
1568 | struct @{ | |
1569 | union @{ | |
1570 | void (*fn)(); | |
1571 | ptrdiff_t vtable_index; | |
1572 | @}; | |
1573 | ptrdiff_t delta; | |
1574 | @}; | |
1575 | @end smallexample | |
1576 | ||
1577 | @noindent | |
1578 | The C++ compiler must use one bit to indicate whether the function that | |
1579 | will be called through a pointer-to-member-function is virtual. | |
1580 | Normally, we assume that the low-order bit of a function pointer must | |
1581 | always be zero. Then, by ensuring that the vtable_index is odd, we can | |
1582 | distinguish which variant of the union is in use. But, on some | |
1583 | platforms function pointers can be odd, and so this doesn't work. In | |
1584 | that case, we use the low-order bit of the @code{delta} field, and shift | |
1585 | the remainder of the @code{delta} field to the left. | |
1586 | ||
1587 | GCC will automatically make the right selection about where to store | |
1588 | this bit using the @code{FUNCTION_BOUNDARY} setting for your platform. | |
1589 | However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY} | |
1590 | set such that functions always start at even addresses, but the lowest | |
1591 | bit of pointers to functions indicate whether the function at that | |
1592 | address is in ARM or Thumb mode. If this is the case of your | |
1593 | architecture, you should define this macro to | |
1594 | @code{ptrmemfunc_vbit_in_delta}. | |
1595 | ||
1596 | In general, you should not have to define this macro. On architectures | |
1597 | in which function addresses are always even, according to | |
1598 | @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to | |
1599 | @code{ptrmemfunc_vbit_in_pfn}. | |
1600 | @end defmac | |
1601 | ||
1602 | @defmac TARGET_VTABLE_USES_DESCRIPTORS | |
1603 | Normally, the C++ compiler uses function pointers in vtables. This | |
1604 | macro allows the target to change to use ``function descriptors'' | |
1605 | instead. Function descriptors are found on targets for whom a | |
1606 | function pointer is actually a small data structure. Normally the | |
1607 | data structure consists of the actual code address plus a data | |
1608 | pointer to which the function's data is relative. | |
1609 | ||
1610 | If vtables are used, the value of this macro should be the number | |
1611 | of words that the function descriptor occupies. | |
1612 | @end defmac | |
1613 | ||
1614 | @defmac TARGET_VTABLE_ENTRY_ALIGN | |
1615 | By default, the vtable entries are void pointers, the so the alignment | |
1616 | is the same as pointer alignment. The value of this macro specifies | |
1617 | the alignment of the vtable entry in bits. It should be defined only | |
1618 | when special alignment is necessary. */ | |
1619 | @end defmac | |
1620 | ||
1621 | @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE | |
1622 | There are a few non-descriptor entries in the vtable at offsets below | |
1623 | zero. If these entries must be padded (say, to preserve the alignment | |
1624 | specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number | |
1625 | of words in each data entry. | |
1626 | @end defmac | |
1627 | ||
1628 | @node Registers | |
1629 | @section Register Usage | |
1630 | @cindex register usage | |
1631 | ||
1632 | This section explains how to describe what registers the target machine | |
1633 | has, and how (in general) they can be used. | |
1634 | ||
1635 | The description of which registers a specific instruction can use is | |
1636 | done with register classes; see @ref{Register Classes}. For information | |
1637 | on using registers to access a stack frame, see @ref{Frame Registers}. | |
1638 | For passing values in registers, see @ref{Register Arguments}. | |
1639 | For returning values in registers, see @ref{Scalar Return}. | |
1640 | ||
1641 | @menu | |
1642 | * Register Basics:: Number and kinds of registers. | |
1643 | * Allocation Order:: Order in which registers are allocated. | |
1644 | * Values in Registers:: What kinds of values each reg can hold. | |
1645 | * Leaf Functions:: Renumbering registers for leaf functions. | |
1646 | * Stack Registers:: Handling a register stack such as 80387. | |
1647 | @end menu | |
1648 | ||
1649 | @node Register Basics | |
1650 | @subsection Basic Characteristics of Registers | |
1651 | ||
1652 | @c prevent bad page break with this line | |
1653 | Registers have various characteristics. | |
1654 | ||
1655 | @defmac FIRST_PSEUDO_REGISTER | |
1656 | Number of hardware registers known to the compiler. They receive | |
1657 | numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first | |
1658 | pseudo register's number really is assigned the number | |
1659 | @code{FIRST_PSEUDO_REGISTER}. | |
1660 | @end defmac | |
1661 | ||
1662 | @defmac FIXED_REGISTERS | |
1663 | @cindex fixed register | |
1664 | An initializer that says which registers are used for fixed purposes | |
1665 | all throughout the compiled code and are therefore not available for | |
1666 | general allocation. These would include the stack pointer, the frame | |
1667 | pointer (except on machines where that can be used as a general | |
1668 | register when no frame pointer is needed), the program counter on | |
1669 | machines where that is considered one of the addressable registers, | |
1670 | and any other numbered register with a standard use. | |
1671 | ||
1672 | This information is expressed as a sequence of numbers, separated by | |
1673 | commas and surrounded by braces. The @var{n}th number is 1 if | |
1674 | register @var{n} is fixed, 0 otherwise. | |
1675 | ||
1676 | The table initialized from this macro, and the table initialized by | |
1677 | the following one, may be overridden at run time either automatically, | |
1678 | by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by | |
1679 | the user with the command options @option{-ffixed-@var{reg}}, | |
1680 | @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}. | |
1681 | @end defmac | |
1682 | ||
1683 | @defmac CALL_USED_REGISTERS | |
1684 | @cindex call-used register | |
1685 | @cindex call-clobbered register | |
1686 | @cindex call-saved register | |
1687 | Like @code{FIXED_REGISTERS} but has 1 for each register that is | |
1688 | clobbered (in general) by function calls as well as for fixed | |
1689 | registers. This macro therefore identifies the registers that are not | |
1690 | available for general allocation of values that must live across | |
1691 | function calls. | |
1692 | ||
1693 | If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler | |
1694 | automatically saves it on function entry and restores it on function | |
1695 | exit, if the register is used within the function. | |
1696 | ||
1697 | Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS} | |
1698 | must be defined. Modern ports should define @code{CALL_REALLY_USED_REGISTERS}. | |
1699 | @end defmac | |
1700 | ||
1701 | @defmac CALL_REALLY_USED_REGISTERS | |
1702 | @cindex call-used register | |
1703 | @cindex call-clobbered register | |
1704 | @cindex call-saved register | |
1705 | Like @code{CALL_USED_REGISTERS} except this macro doesn't require | |
1706 | that the entire set of @code{FIXED_REGISTERS} be included. | |
1707 | (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}). | |
1708 | ||
1709 | Exactly one of @code{CALL_USED_REGISTERS} and @code{CALL_REALLY_USED_REGISTERS} | |
1710 | must be defined. Modern ports should define @code{CALL_REALLY_USED_REGISTERS}. | |
1711 | @end defmac | |
1712 | ||
1713 | @cindex call-used register | |
1714 | @cindex call-clobbered register | |
1715 | @cindex call-saved register | |
1716 | @hook TARGET_FNTYPE_ABI | |
1717 | ||
1718 | @hook TARGET_INSN_CALLEE_ABI | |
1719 | ||
1720 | @cindex call-used register | |
1721 | @cindex call-clobbered register | |
1722 | @cindex call-saved register | |
1723 | @hook TARGET_HARD_REGNO_CALL_PART_CLOBBERED | |
1724 | ||
1725 | @hook TARGET_GET_MULTILIB_ABI_NAME | |
1726 | ||
1727 | @findex fixed_regs | |
1728 | @findex call_used_regs | |
1729 | @findex global_regs | |
1730 | @findex reg_names | |
1731 | @findex reg_class_contents | |
1732 | @hook TARGET_CONDITIONAL_REGISTER_USAGE | |
1733 | ||
1734 | @defmac INCOMING_REGNO (@var{out}) | |
1735 | Define this macro if the target machine has register windows. This C | |
1736 | expression returns the register number as seen by the called function | |
1737 | corresponding to the register number @var{out} as seen by the calling | |
1738 | function. Return @var{out} if register number @var{out} is not an | |
1739 | outbound register. | |
1740 | @end defmac | |
1741 | ||
1742 | @defmac OUTGOING_REGNO (@var{in}) | |
1743 | Define this macro if the target machine has register windows. This C | |
1744 | expression returns the register number as seen by the calling function | |
1745 | corresponding to the register number @var{in} as seen by the called | |
1746 | function. Return @var{in} if register number @var{in} is not an inbound | |
1747 | register. | |
1748 | @end defmac | |
1749 | ||
1750 | @defmac LOCAL_REGNO (@var{regno}) | |
1751 | Define this macro if the target machine has register windows. This C | |
1752 | expression returns true if the register is call-saved but is in the | |
1753 | register window. Unlike most call-saved registers, such registers | |
1754 | need not be explicitly restored on function exit or during non-local | |
1755 | gotos. | |
1756 | @end defmac | |
1757 | ||
1758 | @defmac PC_REGNUM | |
1759 | If the program counter has a register number, define this as that | |
1760 | register number. Otherwise, do not define it. | |
1761 | @end defmac | |
1762 | ||
1763 | @node Allocation Order | |
1764 | @subsection Order of Allocation of Registers | |
1765 | @cindex order of register allocation | |
1766 | @cindex register allocation order | |
1767 | ||
1768 | @c prevent bad page break with this line | |
1769 | Registers are allocated in order. | |
1770 | ||
1771 | @defmac REG_ALLOC_ORDER | |
1772 | If defined, an initializer for a vector of integers, containing the | |
1773 | numbers of hard registers in the order in which GCC should prefer | |
1774 | to use them (from most preferred to least). | |
1775 | ||
1776 | If this macro is not defined, registers are used lowest numbered first | |
1777 | (all else being equal). | |
1778 | ||
1779 | One use of this macro is on machines where the highest numbered | |
1780 | registers must always be saved and the save-multiple-registers | |
1781 | instruction supports only sequences of consecutive registers. On such | |
1782 | machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists | |
1783 | the highest numbered allocable register first. | |
1784 | @end defmac | |
1785 | ||
1786 | @defmac ADJUST_REG_ALLOC_ORDER | |
1787 | A C statement (sans semicolon) to choose the order in which to allocate | |
1788 | hard registers for pseudo-registers local to a basic block. | |
1789 | ||
1790 | Store the desired register order in the array @code{reg_alloc_order}. | |
1791 | Element 0 should be the register to allocate first; element 1, the next | |
1792 | register; and so on. | |
1793 | ||
1794 | The macro body should not assume anything about the contents of | |
1795 | @code{reg_alloc_order} before execution of the macro. | |
1796 | ||
1797 | On most machines, it is not necessary to define this macro. | |
1798 | @end defmac | |
1799 | ||
1800 | @defmac HONOR_REG_ALLOC_ORDER | |
1801 | Normally, IRA tries to estimate the costs for saving a register in the | |
1802 | prologue and restoring it in the epilogue. This discourages it from | |
1803 | using call-saved registers. If a machine wants to ensure that IRA | |
1804 | allocates registers in the order given by REG_ALLOC_ORDER even if some | |
1805 | call-saved registers appear earlier than call-used ones, then define this | |
1806 | macro as a C expression to nonzero. Default is 0. | |
1807 | @end defmac | |
1808 | ||
1809 | @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno}) | |
1810 | In some case register allocation order is not enough for the | |
1811 | Integrated Register Allocator (@acronym{IRA}) to generate a good code. | |
1812 | If this macro is defined, it should return a floating point value | |
1813 | based on @var{regno}. The cost of using @var{regno} for a pseudo will | |
1814 | be increased by approximately the pseudo's usage frequency times the | |
1815 | value returned by this macro. Not defining this macro is equivalent | |
1816 | to having it always return @code{0.0}. | |
1817 | ||
1818 | On most machines, it is not necessary to define this macro. | |
1819 | @end defmac | |
1820 | ||
1821 | @node Values in Registers | |
1822 | @subsection How Values Fit in Registers | |
1823 | ||
1824 | This section discusses the macros that describe which kinds of values | |
1825 | (specifically, which machine modes) each register can hold, and how many | |
1826 | consecutive registers are needed for a given mode. | |
1827 | ||
1828 | @hook TARGET_HARD_REGNO_NREGS | |
1829 | ||
1830 | @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode}) | |
1831 | A C expression that is nonzero if a value of mode @var{mode}, stored | |
1832 | in memory, ends with padding that causes it to take up more space than | |
1833 | in registers starting at register number @var{regno} (as determined by | |
1834 | multiplying GCC's notion of the size of the register when containing | |
1835 | this mode by the number of registers returned by | |
1836 | @code{TARGET_HARD_REGNO_NREGS}). By default this is zero. | |
1837 | ||
1838 | For example, if a floating-point value is stored in three 32-bit | |
1839 | registers but takes up 128 bits in memory, then this would be | |
1840 | nonzero. | |
1841 | ||
1842 | This macros only needs to be defined if there are cases where | |
1843 | @code{subreg_get_info} | |
1844 | would otherwise wrongly determine that a @code{subreg} can be | |
1845 | represented by an offset to the register number, when in fact such a | |
1846 | @code{subreg} would contain some of the padding not stored in | |
1847 | registers and so not be representable. | |
1848 | @end defmac | |
1849 | ||
1850 | @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode}) | |
1851 | For values of @var{regno} and @var{mode} for which | |
1852 | @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression | |
1853 | returning the greater number of registers required to hold the value | |
1854 | including any padding. In the example above, the value would be four. | |
1855 | @end defmac | |
1856 | ||
1857 | @defmac REGMODE_NATURAL_SIZE (@var{mode}) | |
1858 | Define this macro if the natural size of registers that hold values | |
1859 | of mode @var{mode} is not the word size. It is a C expression that | |
1860 | should give the natural size in bytes for the specified mode. It is | |
1861 | used by the register allocator to try to optimize its results. This | |
1862 | happens for example on SPARC 64-bit where the natural size of | |
1863 | floating-point registers is still 32-bit. | |
1864 | @end defmac | |
1865 | ||
1866 | @hook TARGET_HARD_REGNO_MODE_OK | |
1867 | ||
1868 | @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to}) | |
1869 | A C expression that is nonzero if it is OK to rename a hard register | |
1870 | @var{from} to another hard register @var{to}. | |
1871 | ||
1872 | One common use of this macro is to prevent renaming of a register to | |
1873 | another register that is not saved by a prologue in an interrupt | |
1874 | handler. | |
1875 | ||
1876 | The default is always nonzero. | |
1877 | @end defmac | |
1878 | ||
1879 | @hook TARGET_MODES_TIEABLE_P | |
1880 | ||
1881 | @hook TARGET_HARD_REGNO_SCRATCH_OK | |
1882 | ||
1883 | @defmac AVOID_CCMODE_COPIES | |
1884 | Define this macro if the compiler should avoid copies to/from @code{CCmode} | |
1885 | registers. You should only define this macro if support for copying to/from | |
1886 | @code{CCmode} is incomplete. | |
1887 | @end defmac | |
1888 | ||
1889 | @node Leaf Functions | |
1890 | @subsection Handling Leaf Functions | |
1891 | ||
1892 | @cindex leaf functions | |
1893 | @cindex functions, leaf | |
1894 | On some machines, a leaf function (i.e., one which makes no calls) can run | |
1895 | more efficiently if it does not make its own register window. Often this | |
1896 | means it is required to receive its arguments in the registers where they | |
1897 | are passed by the caller, instead of the registers where they would | |
1898 | normally arrive. | |
1899 | ||
1900 | The special treatment for leaf functions generally applies only when | |
1901 | other conditions are met; for example, often they may use only those | |
1902 | registers for its own variables and temporaries. We use the term ``leaf | |
1903 | function'' to mean a function that is suitable for this special | |
1904 | handling, so that functions with no calls are not necessarily ``leaf | |
1905 | functions''. | |
1906 | ||
1907 | GCC assigns register numbers before it knows whether the function is | |
1908 | suitable for leaf function treatment. So it needs to renumber the | |
1909 | registers in order to output a leaf function. The following macros | |
1910 | accomplish this. | |
1911 | ||
1912 | @defmac LEAF_REGISTERS | |
1913 | Name of a char vector, indexed by hard register number, which | |
1914 | contains 1 for a register that is allowable in a candidate for leaf | |
1915 | function treatment. | |
1916 | ||
1917 | If leaf function treatment involves renumbering the registers, then the | |
1918 | registers marked here should be the ones before renumbering---those that | |
1919 | GCC would ordinarily allocate. The registers which will actually be | |
1920 | used in the assembler code, after renumbering, should not be marked with 1 | |
1921 | in this vector. | |
1922 | ||
1923 | Define this macro only if the target machine offers a way to optimize | |
1924 | the treatment of leaf functions. | |
1925 | @end defmac | |
1926 | ||
1927 | @defmac LEAF_REG_REMAP (@var{regno}) | |
1928 | A C expression whose value is the register number to which @var{regno} | |
1929 | should be renumbered, when a function is treated as a leaf function. | |
1930 | ||
1931 | If @var{regno} is a register number which should not appear in a leaf | |
1932 | function before renumbering, then the expression should yield @minus{}1, which | |
1933 | will cause the compiler to abort. | |
1934 | ||
1935 | Define this macro only if the target machine offers a way to optimize the | |
1936 | treatment of leaf functions, and registers need to be renumbered to do | |
1937 | this. | |
1938 | @end defmac | |
1939 | ||
1940 | @findex current_function_is_leaf | |
1941 | @findex current_function_uses_only_leaf_regs | |
1942 | @code{TARGET_ASM_FUNCTION_PROLOGUE} and | |
1943 | @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions | |
1944 | specially. They can test the C variable @code{current_function_is_leaf} | |
1945 | which is nonzero for leaf functions. @code{current_function_is_leaf} is | |
1946 | set prior to local register allocation and is valid for the remaining | |
1947 | compiler passes. They can also test the C variable | |
1948 | @code{current_function_uses_only_leaf_regs} which is nonzero for leaf | |
1949 | functions which only use leaf registers. | |
1950 | @code{current_function_uses_only_leaf_regs} is valid after all passes | |
1951 | that modify the instructions have been run and is only useful if | |
1952 | @code{LEAF_REGISTERS} is defined. | |
1953 | @c changed this to fix overfull. ALSO: why the "it" at the beginning | |
1954 | @c of the next paragraph?! --mew 2feb93 | |
1955 | ||
1956 | @node Stack Registers | |
1957 | @subsection Registers That Form a Stack | |
1958 | ||
1959 | There are special features to handle computers where some of the | |
1960 | ``registers'' form a stack. Stack registers are normally written by | |
1961 | pushing onto the stack, and are numbered relative to the top of the | |
1962 | stack. | |
1963 | ||
1964 | Currently, GCC can only handle one group of stack-like registers, and | |
1965 | they must be consecutively numbered. Furthermore, the existing | |
1966 | support for stack-like registers is specific to the 80387 floating | |
1967 | point coprocessor. If you have a new architecture that uses | |
1968 | stack-like registers, you will need to do substantial work on | |
1969 | @file{reg-stack.cc} and write your machine description to cooperate | |
1970 | with it, as well as defining these macros. | |
1971 | ||
1972 | @defmac STACK_REGS | |
1973 | Define this if the machine has any stack-like registers. | |
1974 | @end defmac | |
1975 | ||
1976 | @defmac STACK_REG_COVER_CLASS | |
1977 | This is a cover class containing the stack registers. Define this if | |
1978 | the machine has any stack-like registers. | |
1979 | @end defmac | |
1980 | ||
1981 | @defmac FIRST_STACK_REG | |
1982 | The number of the first stack-like register. This one is the top | |
1983 | of the stack. | |
1984 | @end defmac | |
1985 | ||
1986 | @defmac LAST_STACK_REG | |
1987 | The number of the last stack-like register. This one is the bottom of | |
1988 | the stack. | |
1989 | @end defmac | |
1990 | ||
1991 | @node Register Classes | |
1992 | @section Register Classes | |
1993 | @cindex register class definitions | |
1994 | @cindex class definitions, register | |
1995 | ||
1996 | On many machines, the numbered registers are not all equivalent. | |
1997 | For example, certain registers may not be allowed for indexed addressing; | |
1998 | certain registers may not be allowed in some instructions. These machine | |
1999 | restrictions are described to the compiler using @dfn{register classes}. | |
2000 | ||
2001 | You define a number of register classes, giving each one a name and saying | |
2002 | which of the registers belong to it. Then you can specify register classes | |
2003 | that are allowed as operands to particular instruction patterns. | |
2004 | ||
2005 | @findex ALL_REGS | |
2006 | @findex NO_REGS | |
2007 | In general, each register will belong to several classes. In fact, one | |
2008 | class must be named @code{ALL_REGS} and contain all the registers. Another | |
2009 | class must be named @code{NO_REGS} and contain no registers. Often the | |
2010 | union of two classes will be another class; however, this is not required. | |
2011 | ||
2012 | @findex GENERAL_REGS | |
2013 | One of the classes must be named @code{GENERAL_REGS}. There is nothing | |
2014 | terribly special about the name, but the operand constraint letters | |
2015 | @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is | |
2016 | the same as @code{ALL_REGS}, just define it as a macro which expands | |
2017 | to @code{ALL_REGS}. | |
2018 | ||
2019 | Order the classes so that if class @var{x} is contained in class @var{y} | |
2020 | then @var{x} has a lower class number than @var{y}. | |
2021 | ||
2022 | The way classes other than @code{GENERAL_REGS} are specified in operand | |
2023 | constraints is through machine-dependent operand constraint letters. | |
2024 | You can define such letters to correspond to various classes, then use | |
2025 | them in operand constraints. | |
2026 | ||
2027 | You must define the narrowest register classes for allocatable | |
2028 | registers, so that each class either has no subclasses, or that for | |
2029 | some mode, the move cost between registers within the class is | |
2030 | cheaper than moving a register in the class to or from memory | |
2031 | (@pxref{Costs}). | |
2032 | ||
2033 | You should define a class for the union of two classes whenever some | |
2034 | instruction allows both classes. For example, if an instruction allows | |
2035 | either a floating point (coprocessor) register or a general register for a | |
2036 | certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS} | |
2037 | which includes both of them. Otherwise you will get suboptimal code, | |
2038 | or even internal compiler errors when reload cannot find a register in the | |
2039 | class computed via @code{reg_class_subunion}. | |
2040 | ||
2041 | You must also specify certain redundant information about the register | |
2042 | classes: for each class, which classes contain it and which ones are | |
2043 | contained in it; for each pair of classes, the largest class contained | |
2044 | in their union. | |
2045 | ||
2046 | When a value occupying several consecutive registers is expected in a | |
2047 | certain class, all the registers used must belong to that class. | |
2048 | Therefore, register classes cannot be used to enforce a requirement for | |
2049 | a register pair to start with an even-numbered register. The way to | |
09a85191 RS |
2050 | specify this requirement is with @code{TARGET_HARD_REGNO_MODE_OK}, |
2051 | or with a filter expression in a @code{define_register_constraint}. | |
d77de738 ML |
2052 | |
2053 | Register classes used for input-operands of bitwise-and or shift | |
2054 | instructions have a special requirement: each such class must have, for | |
2055 | each fixed-point machine mode, a subclass whose registers can transfer that | |
2056 | mode to or from memory. For example, on some machines, the operations for | |
2057 | single-byte values (@code{QImode}) are limited to certain registers. When | |
2058 | this is so, each register class that is used in a bitwise-and or shift | |
2059 | instruction must have a subclass consisting of registers from which | |
2060 | single-byte values can be loaded or stored. This is so that | |
2061 | @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return. | |
2062 | ||
2063 | @deftp {Data type} {enum reg_class} | |
2064 | An enumerated type that must be defined with all the register class names | |
2065 | as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS} | |
2066 | must be the last register class, followed by one more enumerated value, | |
2067 | @code{LIM_REG_CLASSES}, which is not a register class but rather | |
2068 | tells how many classes there are. | |
2069 | ||
2070 | Each register class has a number, which is the value of casting | |
2071 | the class name to type @code{int}. The number serves as an index | |
2072 | in many of the tables described below. | |
2073 | @end deftp | |
2074 | ||
2075 | @defmac N_REG_CLASSES | |
2076 | The number of distinct register classes, defined as follows: | |
2077 | ||
2078 | @smallexample | |
2079 | #define N_REG_CLASSES (int) LIM_REG_CLASSES | |
2080 | @end smallexample | |
2081 | @end defmac | |
2082 | ||
2083 | @defmac REG_CLASS_NAMES | |
2084 | An initializer containing the names of the register classes as C string | |
2085 | constants. These names are used in writing some of the debugging dumps. | |
2086 | @end defmac | |
2087 | ||
2088 | @defmac REG_CLASS_CONTENTS | |
2089 | An initializer containing the contents of the register classes, as integers | |
2090 | which are bit masks. The @var{n}th integer specifies the contents of class | |
2091 | @var{n}. The way the integer @var{mask} is interpreted is that | |
2092 | register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1. | |
2093 | ||
2094 | When the machine has more than 32 registers, an integer does not suffice. | |
2095 | Then the integers are replaced by sub-initializers, braced groupings containing | |
2096 | several integers. Each sub-initializer must be suitable as an initializer | |
2097 | for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}. | |
2098 | In this situation, the first integer in each sub-initializer corresponds to | |
2099 | registers 0 through 31, the second integer to registers 32 through 63, and | |
2100 | so on. | |
2101 | @end defmac | |
2102 | ||
2103 | @defmac REGNO_REG_CLASS (@var{regno}) | |
2104 | A C expression whose value is a register class containing hard register | |
2105 | @var{regno}. In general there is more than one such class; choose a class | |
2106 | which is @dfn{minimal}, meaning that no smaller class also contains the | |
2107 | register. | |
2108 | @end defmac | |
2109 | ||
2110 | @defmac BASE_REG_CLASS | |
2111 | A macro whose definition is the name of the class to which a valid | |
2112 | base register must belong. A base register is one used in an address | |
2113 | which is the register value plus a displacement. | |
2114 | @end defmac | |
2115 | ||
2116 | @defmac MODE_BASE_REG_CLASS (@var{mode}) | |
2117 | This is a variation of the @code{BASE_REG_CLASS} macro which allows | |
2118 | the selection of a base register in a mode dependent manner. If | |
2119 | @var{mode} is VOIDmode then it should return the same value as | |
2120 | @code{BASE_REG_CLASS}. | |
2121 | @end defmac | |
2122 | ||
2123 | @defmac MODE_BASE_REG_REG_CLASS (@var{mode}) | |
2124 | A C expression whose value is the register class to which a valid | |
2125 | base register must belong in order to be used in a base plus index | |
2126 | register address. You should define this macro if base plus index | |
2127 | addresses have different requirements than other base register uses. | |
2128 | @end defmac | |
2129 | ||
2130 | @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{address_space}, @var{outer_code}, @var{index_code}) | |
2131 | A C expression whose value is the register class to which a valid | |
2132 | base register for a memory reference in mode @var{mode} to address | |
2133 | space @var{address_space} must belong. @var{outer_code} and @var{index_code} | |
2134 | define the context in which the base register occurs. @var{outer_code} is | |
2135 | the code of the immediately enclosing expression (@code{MEM} for the top level | |
2136 | of an address, @code{ADDRESS} for something that occurs in an | |
2137 | @code{address_operand}). @var{index_code} is the code of the corresponding | |
2138 | index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise. | |
2139 | @end defmac | |
2140 | ||
bc4466b9 KL |
2141 | @defmac INSN_BASE_REG_CLASS (@var{insn}) |
2142 | A C expression whose value is the register class to which a valid | |
2143 | base register for a specified @var{insn} must belong. This macro is | |
2144 | used when some backend insns may have limited usage of base register | |
2145 | compared with other insns. If you define this macro, the compiler will | |
2146 | use it instead of all other defined macros that relate to | |
2147 | BASE_REG_CLASS. | |
2148 | @end defmac | |
2149 | ||
d77de738 ML |
2150 | @defmac INDEX_REG_CLASS |
2151 | A macro whose definition is the name of the class to which a valid | |
2152 | index register must belong. An index register is one used in an | |
2153 | address where its value is either multiplied by a scale factor or | |
2154 | added to another register (as well as added to a displacement). | |
2155 | @end defmac | |
2156 | ||
dfa15b49 HW |
2157 | @defmac INSN_INDEX_REG_CLASS (@var{insn}) |
2158 | A C expression whose value is the register class to which a valid | |
2159 | index register for a specified @var{insn} must belong. This macro is | |
2160 | used when some backend insns may have limited usage of index register | |
2161 | compared with other insns. If you defined this macro, the compiler | |
2162 | will use it instead of @code{INDEX_REG_CLASS}. | |
2163 | @end defmac | |
2164 | ||
d77de738 ML |
2165 | @defmac REGNO_OK_FOR_BASE_P (@var{num}) |
2166 | A C expression which is nonzero if register number @var{num} is | |
2167 | suitable for use as a base register in operand addresses. | |
2168 | @end defmac | |
2169 | ||
2170 | @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode}) | |
2171 | A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that | |
2172 | that expression may examine the mode of the memory reference in | |
2173 | @var{mode}. You should define this macro if the mode of the memory | |
2174 | reference affects whether a register may be used as a base register. If | |
2175 | you define this macro, the compiler will use it instead of | |
2176 | @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for | |
2177 | addresses that appear outside a @code{MEM}, i.e., as an | |
2178 | @code{address_operand}. | |
2179 | @end defmac | |
2180 | ||
2181 | @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode}) | |
2182 | A C expression which is nonzero if register number @var{num} is suitable for | |
2183 | use as a base register in base plus index operand addresses, accessing | |
2184 | memory in mode @var{mode}. It may be either a suitable hard register or a | |
2185 | pseudo register that has been allocated such a hard register. You should | |
2186 | define this macro if base plus index addresses have different requirements | |
2187 | than other base register uses. | |
2188 | ||
2189 | Use of this macro is deprecated; please use the more general | |
2190 | @code{REGNO_MODE_CODE_OK_FOR_BASE_P}. | |
2191 | @end defmac | |
2192 | ||
2193 | @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{address_space}, @var{outer_code}, @var{index_code}) | |
2194 | A C expression which is nonzero if register number @var{num} is | |
2195 | suitable for use as a base register in operand addresses, accessing | |
2196 | memory in mode @var{mode} in address space @var{address_space}. | |
2197 | This is similar to @code{REGNO_MODE_OK_FOR_BASE_P}, except | |
2198 | that that expression may examine the context in which the register | |
2199 | appears in the memory reference. @var{outer_code} is the code of the | |
2200 | immediately enclosing expression (@code{MEM} if at the top level of the | |
2201 | address, @code{ADDRESS} for something that occurs in an | |
2202 | @code{address_operand}). @var{index_code} is the code of the | |
2203 | corresponding index expression if @var{outer_code} is @code{PLUS}; | |
2204 | @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses | |
2205 | that appear outside a @code{MEM}, i.e., as an @code{address_operand}. | |
2206 | @end defmac | |
2207 | ||
bc4466b9 KL |
2208 | @defmac REGNO_OK_FOR_INSN_BASE_P (@var{num}, @var{insn}) |
2209 | A C expression which is nonzero if register number @var{num} is | |
2210 | suitable for use as a base register in operand addresses for a specified | |
2211 | @var{insn}. This macro is used when some backend insn may have limited | |
2212 | usage of base register compared with other insns. If you define this | |
2213 | macro, the compiler will use it instead of all other defined macros | |
2214 | that relate to REGNO_OK_FOR_BASE_P. | |
2215 | @end defmac | |
2216 | ||
d77de738 ML |
2217 | @defmac REGNO_OK_FOR_INDEX_P (@var{num}) |
2218 | A C expression which is nonzero if register number @var{num} is | |
2219 | suitable for use as an index register in operand addresses. It may be | |
2220 | either a suitable hard register or a pseudo register that has been | |
2221 | allocated such a hard register. | |
2222 | ||
2223 | The difference between an index register and a base register is that | |
2224 | the index register may be scaled. If an address involves the sum of | |
2225 | two registers, neither one of them scaled, then either one may be | |
2226 | labeled the ``base'' and the other the ``index''; but whichever | |
2227 | labeling is used must fit the machine's constraints of which registers | |
2228 | may serve in each capacity. The compiler will try both labelings, | |
2229 | looking for one that is valid, and will reload one or both registers | |
2230 | only if neither labeling works. | |
2231 | @end defmac | |
2232 | ||
2233 | @hook TARGET_PREFERRED_RENAME_CLASS | |
2234 | ||
2235 | @hook TARGET_PREFERRED_RELOAD_CLASS | |
2236 | ||
2237 | @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class}) | |
2238 | A C expression that places additional restrictions on the register class | |
2239 | to use when it is necessary to copy value @var{x} into a register in class | |
2240 | @var{class}. The value is a register class; perhaps @var{class}, or perhaps | |
2241 | another, smaller class. On many machines, the following definition is | |
2242 | safe: | |
2243 | ||
2244 | @smallexample | |
2245 | #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS | |
2246 | @end smallexample | |
2247 | ||
2248 | Sometimes returning a more restrictive class makes better code. For | |
2249 | example, on the 68000, when @var{x} is an integer constant that is in range | |
2250 | for a @samp{moveq} instruction, the value of this macro is always | |
2251 | @code{DATA_REGS} as long as @var{class} includes the data registers. | |
2252 | Requiring a data register guarantees that a @samp{moveq} will be used. | |
2253 | ||
2254 | One case where @code{PREFERRED_RELOAD_CLASS} must not return | |
2255 | @var{class} is if @var{x} is a legitimate constant which cannot be | |
2256 | loaded into some register class. By returning @code{NO_REGS} you can | |
2257 | force @var{x} into a memory location. For example, rs6000 can load | |
2258 | immediate values into general-purpose registers, but does not have an | |
2259 | instruction for loading an immediate value into a floating-point | |
2260 | register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when | |
2261 | @var{x} is a floating-point constant. If the constant cannot be loaded | |
2262 | into any kind of register, code generation will be better if | |
2263 | @code{TARGET_LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead | |
2264 | of using @code{TARGET_PREFERRED_RELOAD_CLASS}. | |
2265 | ||
2266 | If an insn has pseudos in it after register allocation, reload will go | |
2267 | through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS} | |
2268 | to find the best one. Returning @code{NO_REGS}, in this case, makes | |
2269 | reload add a @code{!} in front of the constraint: the x86 back-end uses | |
2270 | this feature to discourage usage of 387 registers when math is done in | |
2271 | the SSE registers (and vice versa). | |
2272 | @end defmac | |
2273 | ||
2274 | @hook TARGET_PREFERRED_OUTPUT_RELOAD_CLASS | |
2275 | ||
2276 | @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class}) | |
2277 | A C expression that places additional restrictions on the register class | |
2278 | to use when it is necessary to be able to hold a value of mode | |
2279 | @var{mode} in a reload register for which class @var{class} would | |
2280 | ordinarily be used. | |
2281 | ||
2282 | Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when | |
2283 | there are certain modes that simply cannot go in certain reload classes. | |
2284 | ||
2285 | The value is a register class; perhaps @var{class}, or perhaps another, | |
2286 | smaller class. | |
2287 | ||
2288 | Don't define this macro unless the target machine has limitations which | |
2289 | require the macro to do something nontrivial. | |
2290 | @end defmac | |
2291 | ||
2292 | @hook TARGET_SECONDARY_RELOAD | |
2293 | ||
2294 | @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) | |
2295 | @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) | |
2296 | @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) | |
2297 | These macros are obsolete, new ports should use the target hook | |
2298 | @code{TARGET_SECONDARY_RELOAD} instead. | |
2299 | ||
2300 | These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD} | |
2301 | target hook. Older ports still define these macros to indicate to the | |
2302 | reload phase that it may | |
2303 | need to allocate at least one register for a reload in addition to the | |
2304 | register to contain the data. Specifically, if copying @var{x} to a | |
2305 | register @var{class} in @var{mode} requires an intermediate register, | |
2306 | you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the | |
2307 | largest register class all of whose registers can be used as | |
2308 | intermediate registers or scratch registers. | |
2309 | ||
2310 | If copying a register @var{class} in @var{mode} to @var{x} requires an | |
2311 | intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS} | |
2312 | was supposed to be defined to return the largest register | |
2313 | class required. If the | |
2314 | requirements for input and output reloads were the same, the macro | |
2315 | @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both | |
2316 | macros identically. | |
2317 | ||
2318 | The values returned by these macros are often @code{GENERAL_REGS}. | |
2319 | Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x} | |
2320 | can be directly copied to or from a register of @var{class} in | |
2321 | @var{mode} without requiring a scratch register. Do not define this | |
2322 | macro if it would always return @code{NO_REGS}. | |
2323 | ||
2324 | If a scratch register is required (either with or without an | |
2325 | intermediate register), you were supposed to define patterns for | |
2326 | @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required | |
2327 | (@pxref{Standard Names}. These patterns, which were normally | |
2328 | implemented with a @code{define_expand}, should be similar to the | |
2329 | @samp{mov@var{m}} patterns, except that operand 2 is the scratch | |
2330 | register. | |
2331 | ||
2332 | These patterns need constraints for the reload register and scratch | |
2333 | register that | |
2334 | contain a single register class. If the original reload register (whose | |
2335 | class is @var{class}) can meet the constraint given in the pattern, the | |
2336 | value returned by these macros is used for the class of the scratch | |
2337 | register. Otherwise, two additional reload registers are required. | |
2338 | Their classes are obtained from the constraints in the insn pattern. | |
2339 | ||
2340 | @var{x} might be a pseudo-register or a @code{subreg} of a | |
2341 | pseudo-register, which could either be in a hard register or in memory. | |
2342 | Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is | |
2343 | in memory and the hard register number if it is in a register. | |
2344 | ||
2345 | These macros should not be used in the case where a particular class of | |
2346 | registers can only be copied to memory and not to another class of | |
2347 | registers. In that case, secondary reload registers are not needed and | |
2348 | would not be helpful. Instead, a stack location must be used to perform | |
2349 | the copy and the @code{mov@var{m}} pattern should use memory as an | |
2350 | intermediate storage. This case often occurs between floating-point and | |
2351 | general registers. | |
2352 | @end defmac | |
2353 | ||
2354 | @hook TARGET_SECONDARY_MEMORY_NEEDED | |
2355 | ||
2356 | @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode}) | |
2357 | Normally when @code{TARGET_SECONDARY_MEMORY_NEEDED} is defined, the compiler | |
2358 | allocates a stack slot for a memory location needed for register copies. | |
2359 | If this macro is defined, the compiler instead uses the memory location | |
2360 | defined by this macro. | |
2361 | ||
2362 | Do not define this macro if you do not define | |
2363 | @code{TARGET_SECONDARY_MEMORY_NEEDED}. | |
2364 | @end defmac | |
2365 | ||
2366 | @hook TARGET_SECONDARY_MEMORY_NEEDED_MODE | |
2367 | ||
2368 | @hook TARGET_SELECT_EARLY_REMAT_MODES | |
2369 | ||
2370 | @hook TARGET_CLASS_LIKELY_SPILLED_P | |
2371 | ||
2372 | @hook TARGET_CLASS_MAX_NREGS | |
2373 | ||
2374 | @defmac CLASS_MAX_NREGS (@var{class}, @var{mode}) | |
2375 | A C expression for the maximum number of consecutive registers | |
2376 | of class @var{class} needed to hold a value of mode @var{mode}. | |
2377 | ||
2378 | This is closely related to the macro @code{TARGET_HARD_REGNO_NREGS}. In fact, | |
2379 | the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})} | |
2380 | should be the maximum value of @code{TARGET_HARD_REGNO_NREGS (@var{regno}, | |
2381 | @var{mode})} for all @var{regno} values in the class @var{class}. | |
2382 | ||
2383 | This macro helps control the handling of multiple-word values | |
2384 | in the reload pass. | |
2385 | @end defmac | |
2386 | ||
2387 | @hook TARGET_CAN_CHANGE_MODE_CLASS | |
2388 | ||
2389 | @hook TARGET_IRA_CHANGE_PSEUDO_ALLOCNO_CLASS | |
2390 | ||
2391 | @hook TARGET_LRA_P | |
2392 | ||
2393 | @hook TARGET_REGISTER_PRIORITY | |
2394 | ||
2395 | @hook TARGET_REGISTER_USAGE_LEVELING_P | |
2396 | ||
2397 | @hook TARGET_DIFFERENT_ADDR_DISPLACEMENT_P | |
2398 | ||
2399 | @hook TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P | |
2400 | ||
2401 | @hook TARGET_LEGITIMIZE_ADDRESS_DISPLACEMENT | |
2402 | ||
2403 | @hook TARGET_SPILL_CLASS | |
2404 | ||
2405 | @hook TARGET_ADDITIONAL_ALLOCNO_CLASS_P | |
2406 | ||
2407 | @hook TARGET_CSTORE_MODE | |
2408 | ||
2409 | @hook TARGET_COMPUTE_PRESSURE_CLASSES | |
2410 | ||
2411 | @node Stack and Calling | |
2412 | @section Stack Layout and Calling Conventions | |
2413 | @cindex calling conventions | |
2414 | ||
2415 | @c prevent bad page break with this line | |
2416 | This describes the stack layout and calling conventions. | |
2417 | ||
2418 | @menu | |
2419 | * Frame Layout:: | |
2420 | * Exception Handling:: | |
2421 | * Stack Checking:: | |
2422 | * Frame Registers:: | |
2423 | * Elimination:: | |
2424 | * Stack Arguments:: | |
2425 | * Register Arguments:: | |
2426 | * Scalar Return:: | |
2427 | * Aggregate Return:: | |
2428 | * Caller Saves:: | |
2429 | * Function Entry:: | |
2430 | * Profiling:: | |
2431 | * Tail Calls:: | |
2432 | * Shrink-wrapping separate components:: | |
2433 | * Stack Smashing Protection:: | |
2434 | * Miscellaneous Register Hooks:: | |
2435 | @end menu | |
2436 | ||
2437 | @node Frame Layout | |
2438 | @subsection Basic Stack Layout | |
2439 | @cindex stack frame layout | |
2440 | @cindex frame layout | |
2441 | ||
2442 | @c prevent bad page break with this line | |
2443 | Here is the basic stack layout. | |
2444 | ||
2445 | @defmac STACK_GROWS_DOWNWARD | |
2446 | Define this macro to be true if pushing a word onto the stack moves the stack | |
2447 | pointer to a smaller address, and false otherwise. | |
2448 | @end defmac | |
2449 | ||
2450 | @defmac STACK_PUSH_CODE | |
2451 | This macro defines the operation used when something is pushed | |
2452 | on the stack. In RTL, a push operation will be | |
2453 | @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})} | |
2454 | ||
2455 | The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC}, | |
2456 | and @code{POST_INC}. Which of these is correct depends on | |
2457 | the stack direction and on whether the stack pointer points | |
2458 | to the last item on the stack or whether it points to the | |
2459 | space for the next item on the stack. | |
2460 | ||
2461 | The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is | |
2462 | true, which is almost always right, and @code{PRE_INC} otherwise, | |
2463 | which is often wrong. | |
2464 | @end defmac | |
2465 | ||
2466 | @defmac FRAME_GROWS_DOWNWARD | |
2467 | Define this macro to nonzero value if the addresses of local variable slots | |
2468 | are at negative offsets from the frame pointer. | |
2469 | @end defmac | |
2470 | ||
2471 | @defmac ARGS_GROW_DOWNWARD | |
2472 | Define this macro if successive arguments to a function occupy decreasing | |
2473 | addresses on the stack. | |
2474 | @end defmac | |
2475 | ||
2476 | @hook TARGET_STARTING_FRAME_OFFSET | |
2477 | ||
2478 | @defmac STACK_ALIGNMENT_NEEDED | |
2479 | Define to zero to disable final alignment of the stack during reload. | |
2480 | The nonzero default for this macro is suitable for most ports. | |
2481 | ||
2482 | On ports where @code{TARGET_STARTING_FRAME_OFFSET} is nonzero or where there | |
2483 | is a register save block following the local block that doesn't require | |
2484 | alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable | |
2485 | stack alignment and do it in the backend. | |
2486 | @end defmac | |
2487 | ||
2488 | @defmac STACK_POINTER_OFFSET | |
2489 | Offset from the stack pointer register to the first location at which | |
2490 | outgoing arguments are placed. If not specified, the default value of | |
2491 | zero is used. This is the proper value for most machines. | |
2492 | ||
2493 | If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above | |
2494 | the first location at which outgoing arguments are placed. | |
2495 | @end defmac | |
2496 | ||
2497 | @defmac FIRST_PARM_OFFSET (@var{fundecl}) | |
2498 | Offset from the argument pointer register to the first argument's | |
2499 | address. On some machines it may depend on the data type of the | |
2500 | function. | |
2501 | ||
2502 | If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above | |
2503 | the first argument's address. | |
2504 | @end defmac | |
2505 | ||
2506 | @defmac STACK_DYNAMIC_OFFSET (@var{fundecl}) | |
2507 | Offset from the stack pointer register to an item dynamically allocated | |
2508 | on the stack, e.g., by @code{alloca}. | |
2509 | ||
2510 | The default value for this macro is @code{STACK_POINTER_OFFSET} plus the | |
2511 | length of the outgoing arguments. The default is correct for most | |
2512 | machines. See @file{function.cc} for details. | |
2513 | @end defmac | |
2514 | ||
2515 | @defmac INITIAL_FRAME_ADDRESS_RTX | |
2516 | A C expression whose value is RTL representing the address of the initial | |
2517 | stack frame. This address is passed to @code{RETURN_ADDR_RTX} and | |
2518 | @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable | |
2519 | default value will be used. Define this macro in order to make frame pointer | |
2520 | elimination work in the presence of @code{__builtin_frame_address (count)} and | |
2521 | @code{__builtin_return_address (count)} for @code{count} not equal to zero. | |
2522 | @end defmac | |
2523 | ||
2524 | @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr}) | |
2525 | A C expression whose value is RTL representing the address in a stack | |
2526 | frame where the pointer to the caller's frame is stored. Assume that | |
2527 | @var{frameaddr} is an RTL expression for the address of the stack frame | |
2528 | itself. | |
2529 | ||
2530 | If you don't define this macro, the default is to return the value | |
2531 | of @var{frameaddr}---that is, the stack frame address is also the | |
2532 | address of the stack word that points to the previous frame. | |
2533 | @end defmac | |
2534 | ||
2535 | @defmac SETUP_FRAME_ADDRESSES | |
2536 | A C expression that produces the machine-specific code to | |
2537 | setup the stack so that arbitrary frames can be accessed. For example, | |
2538 | on the SPARC, we must flush all of the register windows to the stack | |
2539 | before we can access arbitrary stack frames. You will seldom need to | |
2540 | define this macro. The default is to do nothing. | |
2541 | @end defmac | |
2542 | ||
2543 | @hook TARGET_BUILTIN_SETJMP_FRAME_VALUE | |
2544 | ||
2545 | @defmac FRAME_ADDR_RTX (@var{frameaddr}) | |
2546 | A C expression whose value is RTL representing the value of the frame | |
2547 | address for the current frame. @var{frameaddr} is the frame pointer | |
2548 | of the current frame. This is used for __builtin_frame_address. | |
2549 | You need only define this macro if the frame address is not the same | |
2550 | as the frame pointer. Most machines do not need to define it. | |
2551 | @end defmac | |
2552 | ||
2553 | @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr}) | |
2554 | A C expression whose value is RTL representing the value of the return | |
2555 | address for the frame @var{count} steps up from the current frame, after | |
2556 | the prologue. @var{frameaddr} is the frame pointer of the @var{count} | |
2557 | frame, or the frame pointer of the @var{count} @minus{} 1 frame if | |
2558 | @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is nonzero. | |
2559 | ||
2560 | The value of the expression must always be the correct address when | |
2561 | @var{count} is zero, but may be @code{NULL_RTX} if there is no way to | |
2562 | determine the return address of other frames. | |
2563 | @end defmac | |
2564 | ||
2565 | @defmac RETURN_ADDR_IN_PREVIOUS_FRAME | |
2566 | Define this macro to nonzero value if the return address of a particular | |
2567 | stack frame is accessed from the frame pointer of the previous stack | |
2568 | frame. The zero default for this macro is suitable for most ports. | |
2569 | @end defmac | |
2570 | ||
2571 | @defmac INCOMING_RETURN_ADDR_RTX | |
2572 | A C expression whose value is RTL representing the location of the | |
2573 | incoming return address at the beginning of any function, before the | |
2574 | prologue. This RTL is either a @code{REG}, indicating that the return | |
2575 | value is saved in @samp{REG}, or a @code{MEM} representing a location in | |
2576 | the stack. | |
2577 | ||
2578 | You only need to define this macro if you want to support call frame | |
2579 | debugging information like that provided by DWARF 2. | |
2580 | ||
2581 | If this RTL is a @code{REG}, you should also define | |
2582 | @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}. | |
2583 | @end defmac | |
2584 | ||
2585 | @defmac DWARF_ALT_FRAME_RETURN_COLUMN | |
2586 | A C expression whose value is an integer giving a DWARF 2 column | |
2587 | number that may be used as an alternative return column. The column | |
2588 | must not correspond to any gcc hard register (that is, it must not | |
2589 | be in the range of @code{DWARF_FRAME_REGNUM}). | |
2590 | ||
2591 | This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a | |
2592 | general register, but an alternative column needs to be used for signal | |
2593 | frames. Some targets have also used different frame return columns | |
2594 | over time. | |
2595 | @end defmac | |
2596 | ||
2597 | @defmac DWARF_ZERO_REG | |
2598 | A C expression whose value is an integer giving a DWARF 2 register | |
2599 | number that is considered to always have the value zero. This should | |
2600 | only be defined if the target has an architected zero register, and | |
2601 | someone decided it was a good idea to use that register number to | |
2602 | terminate the stack backtrace. New ports should avoid this. | |
2603 | @end defmac | |
2604 | ||
2605 | @defmac DWARF_VERSION_DEFAULT | |
2606 | A C expression whose value is the default dwarf standard version we'll honor | |
2607 | and advertise when generating dwarf debug information, in absence of | |
2608 | an explicit @option{-gdwarf-@var{version}} option on the command line. | |
2609 | @end defmac | |
2610 | ||
2611 | @hook TARGET_DWARF_HANDLE_FRAME_UNSPEC | |
2612 | ||
2613 | @hook TARGET_DWARF_POLY_INDETERMINATE_VALUE | |
2614 | ||
2615 | @defmac INCOMING_FRAME_SP_OFFSET | |
2616 | A C expression whose value is an integer giving the offset, in bytes, | |
2617 | from the value of the stack pointer register to the top of the stack | |
2618 | frame at the beginning of any function, before the prologue. The top of | |
2619 | the frame is defined to be the value of the stack pointer in the | |
2620 | previous frame, just before the call instruction. | |
2621 | ||
2622 | You only need to define this macro if you want to support call frame | |
2623 | debugging information like that provided by DWARF 2. | |
2624 | @end defmac | |
2625 | ||
2626 | @defmac DEFAULT_INCOMING_FRAME_SP_OFFSET | |
2627 | Like @code{INCOMING_FRAME_SP_OFFSET}, but must be the same for all | |
2628 | functions of the same ABI, and when using GAS @code{.cfi_*} directives | |
2629 | must also agree with the default CFI GAS emits. Define this macro | |
2630 | only if @code{INCOMING_FRAME_SP_OFFSET} can have different values | |
2631 | between different functions of the same ABI or when | |
2632 | @code{INCOMING_FRAME_SP_OFFSET} does not agree with GAS default CFI. | |
2633 | @end defmac | |
2634 | ||
2635 | @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl}) | |
2636 | A C expression whose value is an integer giving the offset, in bytes, | |
2637 | from the argument pointer to the canonical frame address (cfa). The | |
2638 | final value should coincide with that calculated by | |
2639 | @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable | |
2640 | during virtual register instantiation. | |
2641 | ||
2642 | The default value for this macro is | |
2643 | @code{FIRST_PARM_OFFSET (fundecl) + crtl->args.pretend_args_size}, | |
2644 | which is correct for most machines; in general, the arguments are found | |
2645 | immediately before the stack frame. Note that this is not the case on | |
2646 | some targets that save registers into the caller's frame, such as SPARC | |
2647 | and rs6000, and so such targets need to define this macro. | |
2648 | ||
2649 | You only need to define this macro if the default is incorrect, and you | |
2650 | want to support call frame debugging information like that provided by | |
2651 | DWARF 2. | |
2652 | @end defmac | |
2653 | ||
2654 | @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl}) | |
2655 | If defined, a C expression whose value is an integer giving the offset | |
2656 | in bytes from the frame pointer to the canonical frame address (cfa). | |
2657 | The final value should coincide with that calculated by | |
2658 | @code{INCOMING_FRAME_SP_OFFSET}. | |
2659 | ||
2660 | Normally the CFA is calculated as an offset from the argument pointer, | |
2661 | via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is | |
2662 | variable due to the ABI, this may not be possible. If this macro is | |
2663 | defined, it implies that the virtual register instantiation should be | |
2664 | based on the frame pointer instead of the argument pointer. Only one | |
2665 | of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET} | |
2666 | should be defined. | |
2667 | @end defmac | |
2668 | ||
2669 | @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl}) | |
2670 | If defined, a C expression whose value is an integer giving the offset | |
2671 | in bytes from the canonical frame address (cfa) to the frame base used | |
2672 | in DWARF 2 debug information. The default is zero. A different value | |
2673 | may reduce the size of debug information on some ports. | |
2674 | @end defmac | |
2675 | ||
f908368d AO |
2676 | @hook TARGET_HAVE_STRUB_SUPPORT_FOR |
2677 | ||
320fb976 AO |
2678 | @defmac STACK_ADDRESS_OFFSET |
2679 | Offset from the stack pointer register to the boundary address between | |
2680 | the stack area claimed by an active function, and stack ranges that | |
2681 | could get clobbered if it called another function. It should NOT | |
2682 | encompass any stack red zone, that is used in leaf functions. | |
2683 | ||
2684 | This value is added to the stack pointer register to compute the address | |
2685 | returned by @code{__builtin_stack_address}, and this is its only use. | |
2686 | If this macro is not defined, no offset is added. Defining it like | |
2687 | @code{STACK_POINTER_OFFSET} may be appropriate for many machines, but | |
2688 | not all. | |
2689 | ||
2690 | On SPARC, for example, the register save area is *not* considered active | |
2691 | or used by the active function, but rather as akin to the area in which | |
2692 | call-preserved registers are saved by callees, so the stack address is | |
2693 | above that area, even though the (unbiased) stack pointer points below | |
2694 | it. This enables @code{__strub_leave} to clear what would otherwise | |
2695 | overlap with its own register save area. | |
2696 | ||
2697 | On PowerPC, @code{STACK_POINTER_OFFSET} also reserves space for a save | |
2698 | area, but that area is used by the caller rather than the callee, so the | |
2699 | boundary address is below it. | |
2700 | ||
2701 | If the address is computed too high or too low, parts of a stack range | |
2702 | that should be scrubbed may be left unscrubbed, scrubbing may corrupt | |
2703 | active portions of the stack frame, and stack ranges may be | |
2704 | doubly-scrubbed by caller and callee. | |
2705 | @end defmac | |
2706 | ||
f0a90c7d AO |
2707 | @defmac TARGET_STRUB_USE_DYNAMIC_ARRAY |
2708 | If defined to nonzero, @code{__strub_leave} will allocate a dynamic | |
2709 | array covering the stack range that needs scrubbing before clearing it. | |
2710 | Allocating the array tends to make scrubbing slower, but it enables the | |
2711 | scrubbing to be safely implemented with a @code{memset} call, which | |
2712 | could make up for the difference. | |
2713 | @end defmac | |
2714 | ||
2715 | @defmac TARGET_STRUB_MAY_USE_MEMSET | |
2716 | If defined to nonzero, enable @code{__strub_leave} to be optimized so as | |
2717 | to call @code{memset} for stack scrubbing. This is only enabled by | |
2718 | default if @code{TARGET_STRUB_USE_DYNAMIC_ARRAY} is enabled; it's not | |
2719 | advisable to enable it otherwise, since @code{memset} would then likely | |
2720 | overwrite its own stack frame, but it might work if the target ABI | |
2721 | enables @code{memset} to not use the stack at all, not even for | |
2722 | arguments or its return address, and its implementation is trivial | |
2723 | enough that it doesn't use a stack frame. | |
2724 | @end defmac | |
2725 | ||
d77de738 ML |
2726 | @node Exception Handling |
2727 | @subsection Exception Handling Support | |
2728 | @cindex exception handling | |
2729 | ||
2730 | @defmac EH_RETURN_DATA_REGNO (@var{N}) | |
2731 | A C expression whose value is the @var{N}th register number used for | |
2732 | data by exception handlers, or @code{INVALID_REGNUM} if fewer than | |
2733 | @var{N} registers are usable. | |
2734 | ||
2735 | The exception handling library routines communicate with the exception | |
2736 | handlers via a set of agreed upon registers. Ideally these registers | |
2737 | should be call-clobbered; it is possible to use call-saved registers, | |
2738 | but may negatively impact code size. The target must support at least | |
2739 | 2 data registers, but should define 4 if there are enough free registers. | |
2740 | ||
2741 | You must define this macro if you want to support call frame exception | |
2742 | handling like that provided by DWARF 2. | |
2743 | @end defmac | |
2744 | ||
2745 | @defmac EH_RETURN_STACKADJ_RTX | |
2746 | A C expression whose value is RTL representing a location in which | |
2747 | to store a stack adjustment to be applied before function return. | |
2748 | This is used to unwind the stack to an exception handler's call frame. | |
2749 | It will be assigned zero on code paths that return normally. | |
2750 | ||
2751 | Typically this is a call-clobbered hard register that is otherwise | |
2752 | untouched by the epilogue, but could also be a stack slot. | |
2753 | ||
2754 | Do not define this macro if the stack pointer is saved and restored | |
2755 | by the regular prolog and epilog code in the call frame itself; in | |
2756 | this case, the exception handling library routines will update the | |
2757 | stack location to be restored in place. Otherwise, you must define | |
2758 | this macro if you want to support call frame exception handling like | |
2759 | that provided by DWARF 2. | |
2760 | @end defmac | |
2761 | ||
2762 | @defmac EH_RETURN_HANDLER_RTX | |
2763 | A C expression whose value is RTL representing a location in which | |
2764 | to store the address of an exception handler to which we should | |
2765 | return. It will not be assigned on code paths that return normally. | |
2766 | ||
2767 | Typically this is the location in the call frame at which the normal | |
2768 | return address is stored. For targets that return by popping an | |
2769 | address off the stack, this might be a memory address just below | |
2770 | the @emph{target} call frame rather than inside the current call | |
2771 | frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already | |
2772 | been assigned, so it may be used to calculate the location of the | |
2773 | target call frame. | |
2774 | ||
2775 | Some targets have more complex requirements than storing to an | |
2776 | address calculable during initial code generation. In that case | |
2777 | the @code{eh_return} instruction pattern should be used instead. | |
2778 | ||
2779 | If you want to support call frame exception handling, you must | |
2780 | define either this macro or the @code{eh_return} instruction pattern. | |
2781 | @end defmac | |
2782 | ||
426fddcb SN |
2783 | @defmac EH_RETURN_TAKEN_RTX |
2784 | A C expression whose value is RTL representing a location in which | |
2785 | to store if the EH return path was taken instead of a normal return. | |
2786 | This macro allows conditionally executing different code in the | |
2787 | epilogue for the EH and normal return cases. | |
2788 | ||
2789 | When this macro is defined, the macros @code{EH_RETURN_STACKADJ_RTX} | |
2790 | and @code{EH_RETURN_HANDLER_RTX} are only meaningful in the epilogue | |
2791 | when 1 is stored to the specified location. The value 0 means normal | |
2792 | return. | |
2793 | @end defmac | |
2794 | ||
d77de738 ML |
2795 | @defmac RETURN_ADDR_OFFSET |
2796 | If defined, an integer-valued C expression for which rtl will be generated | |
2797 | to add it to the exception handler address before it is searched in the | |
2798 | exception handling tables, and to subtract it again from the address before | |
2799 | using it to return to the exception handler. | |
2800 | @end defmac | |
2801 | ||
2802 | @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global}) | |
2803 | This macro chooses the encoding of pointers embedded in the exception | |
2804 | handling sections. If at all possible, this should be defined such | |
2805 | that the exception handling section will not require dynamic relocations, | |
2806 | and so may be read-only. | |
2807 | ||
2808 | @var{code} is 0 for data, 1 for code labels, 2 for function pointers. | |
2809 | @var{global} is true if the symbol may be affected by dynamic relocations. | |
2810 | The macro should return a combination of the @code{DW_EH_PE_*} defines | |
2811 | as found in @file{dwarf2.h}. | |
2812 | ||
2813 | If this macro is not defined, pointers will not be encoded but | |
2814 | represented directly. | |
2815 | @end defmac | |
2816 | ||
2817 | @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done}) | |
2818 | This macro allows the target to emit whatever special magic is required | |
2819 | to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}. | |
2820 | Generic code takes care of pc-relative and indirect encodings; this must | |
2821 | be defined if the target uses text-relative or data-relative encodings. | |
2822 | ||
2823 | This is a C statement that branches to @var{done} if the format was | |
2824 | handled. @var{encoding} is the format chosen, @var{size} is the number | |
2825 | of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF} | |
2826 | to be emitted. | |
2827 | @end defmac | |
2828 | ||
2829 | @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs}) | |
2830 | This macro allows the target to add CPU and operating system specific | |
2831 | code to the call-frame unwinder for use when there is no unwind data | |
2832 | available. The most common reason to implement this macro is to unwind | |
2833 | through signal frames. | |
2834 | ||
2835 | This macro is called from @code{uw_frame_state_for} in | |
2836 | @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and | |
2837 | @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context}; | |
2838 | @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra} | |
2839 | for the address of the code being executed and @code{context->cfa} for | |
2840 | the stack pointer value. If the frame can be decoded, the register | |
2841 | save addresses should be updated in @var{fs} and the macro should | |
2842 | evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded, | |
2843 | the macro should evaluate to @code{_URC_END_OF_STACK}. | |
2844 | ||
2845 | For proper signal handling in Java this macro is accompanied by | |
2846 | @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers. | |
2847 | @end defmac | |
2848 | ||
2849 | @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs}) | |
2850 | This macro allows the target to add operating system specific code to the | |
2851 | call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive, | |
2852 | usually used for signal or interrupt frames. | |
2853 | ||
2854 | This macro is called from @code{uw_update_context} in libgcc's | |
2855 | @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context}; | |
2856 | @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi} | |
2857 | for the abi and context in the @code{.unwabi} directive. If the | |
2858 | @code{.unwabi} directive can be handled, the register save addresses should | |
2859 | be updated in @var{fs}. | |
2860 | @end defmac | |
2861 | ||
2862 | @defmac TARGET_USES_WEAK_UNWIND_INFO | |
2863 | A C expression that evaluates to true if the target requires unwind | |
2864 | info to be given comdat linkage. Define it to be @code{1} if comdat | |
2865 | linkage is necessary. The default is @code{0}. | |
2866 | @end defmac | |
2867 | ||
2868 | @node Stack Checking | |
2869 | @subsection Specifying How Stack Checking is Done | |
2870 | ||
2871 | GCC will check that stack references are within the boundaries of the | |
2872 | stack, if the option @option{-fstack-check} is specified, in one of | |
2873 | three ways: | |
2874 | ||
2875 | @enumerate | |
2876 | @item | |
2877 | If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC | |
2878 | will assume that you have arranged for full stack checking to be done | |
2879 | at appropriate places in the configuration files. GCC will not do | |
2880 | other special processing. | |
2881 | ||
2882 | @item | |
2883 | If @code{STACK_CHECK_BUILTIN} is zero and the value of the | |
2884 | @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume | |
2885 | that you have arranged for static stack checking (checking of the | |
2886 | static stack frame of functions) to be done at appropriate places | |
2887 | in the configuration files. GCC will only emit code to do dynamic | |
2888 | stack checking (checking on dynamic stack allocations) using the third | |
2889 | approach below. | |
2890 | ||
2891 | @item | |
2892 | If neither of the above are true, GCC will generate code to periodically | |
2893 | ``probe'' the stack pointer using the values of the macros defined below. | |
2894 | @end enumerate | |
2895 | ||
2896 | If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined, | |
2897 | GCC will change its allocation strategy for large objects if the option | |
2898 | @option{-fstack-check} is specified: they will always be allocated | |
2899 | dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes. | |
2900 | ||
2901 | @defmac STACK_CHECK_BUILTIN | |
2902 | A nonzero value if stack checking is done by the configuration files in a | |
2903 | machine-dependent manner. You should define this macro if stack checking | |
2904 | is required by the ABI of your machine or if you would like to do stack | |
2905 | checking in some more efficient way than the generic approach. The default | |
2906 | value of this macro is zero. | |
2907 | @end defmac | |
2908 | ||
2909 | @defmac STACK_CHECK_STATIC_BUILTIN | |
2910 | A nonzero value if static stack checking is done by the configuration files | |
2911 | in a machine-dependent manner. You should define this macro if you would | |
2912 | like to do static stack checking in some more efficient way than the generic | |
2913 | approach. The default value of this macro is zero. | |
2914 | @end defmac | |
2915 | ||
2916 | @defmac STACK_CHECK_PROBE_INTERVAL_EXP | |
2917 | An integer specifying the interval at which GCC must generate stack probe | |
2918 | instructions, defined as 2 raised to this integer. You will normally | |
2919 | define this macro so that the interval be no larger than the size of | |
2920 | the ``guard pages'' at the end of a stack area. The default value | |
2921 | of 12 (4096-byte interval) is suitable for most systems. | |
2922 | @end defmac | |
2923 | ||
2924 | @defmac STACK_CHECK_MOVING_SP | |
2925 | An integer which is nonzero if GCC should move the stack pointer page by page | |
2926 | when doing probes. This can be necessary on systems where the stack pointer | |
2927 | contains the bottom address of the memory area accessible to the executing | |
2928 | thread at any point in time. In this situation an alternate signal stack | |
2929 | is required in order to be able to recover from a stack overflow. The | |
2930 | default value of this macro is zero. | |
2931 | @end defmac | |
2932 | ||
2933 | @defmac STACK_CHECK_PROTECT | |
2934 | The number of bytes of stack needed to recover from a stack overflow, for | |
2935 | languages where such a recovery is supported. The default value of 4KB/8KB | |
2936 | with the @code{setjmp}/@code{longjmp}-based exception handling mechanism and | |
2937 | 8KB/12KB with other exception handling mechanisms should be adequate for most | |
2938 | architectures and operating systems. | |
2939 | @end defmac | |
2940 | ||
2941 | The following macros are relevant only if neither STACK_CHECK_BUILTIN | |
2942 | nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether | |
2943 | in the opposite case. | |
2944 | ||
2945 | @defmac STACK_CHECK_MAX_FRAME_SIZE | |
2946 | The maximum size of a stack frame, in bytes. GCC will generate probe | |
2947 | instructions in non-leaf functions to ensure at least this many bytes of | |
2948 | stack are available. If a stack frame is larger than this size, stack | |
2949 | checking will not be reliable and GCC will issue a warning. The | |
2950 | default is chosen so that GCC only generates one instruction on most | |
2951 | systems. You should normally not change the default value of this macro. | |
2952 | @end defmac | |
2953 | ||
2954 | @defmac STACK_CHECK_FIXED_FRAME_SIZE | |
2955 | GCC uses this value to generate the above warning message. It | |
2956 | represents the amount of fixed frame used by a function, not including | |
2957 | space for any callee-saved registers, temporaries and user variables. | |
2958 | You need only specify an upper bound for this amount and will normally | |
2959 | use the default of four words. | |
2960 | @end defmac | |
2961 | ||
2962 | @defmac STACK_CHECK_MAX_VAR_SIZE | |
2963 | The maximum size, in bytes, of an object that GCC will place in the | |
2964 | fixed area of the stack frame when the user specifies | |
2965 | @option{-fstack-check}. | |
2966 | GCC computed the default from the values of the above macros and you will | |
2967 | normally not need to override that default. | |
2968 | @end defmac | |
2969 | ||
2970 | @hook TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE | |
2971 | ||
2972 | @need 2000 | |
2973 | @node Frame Registers | |
2974 | @subsection Registers That Address the Stack Frame | |
2975 | ||
2976 | @c prevent bad page break with this line | |
2977 | This discusses registers that address the stack frame. | |
2978 | ||
2979 | @defmac STACK_POINTER_REGNUM | |
2980 | The register number of the stack pointer register, which must also be a | |
2981 | fixed register according to @code{FIXED_REGISTERS}. On most machines, | |
2982 | the hardware determines which register this is. | |
2983 | @end defmac | |
2984 | ||
2985 | @defmac FRAME_POINTER_REGNUM | |
2986 | The register number of the frame pointer register, which is used to | |
2987 | access automatic variables in the stack frame. On some machines, the | |
2988 | hardware determines which register this is. On other machines, you can | |
2989 | choose any register you wish for this purpose. | |
2990 | @end defmac | |
2991 | ||
2992 | @defmac HARD_FRAME_POINTER_REGNUM | |
2993 | On some machines the offset between the frame pointer and starting | |
2994 | offset of the automatic variables is not known until after register | |
2995 | allocation has been done (for example, because the saved registers are | |
2996 | between these two locations). On those machines, define | |
2997 | @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to | |
2998 | be used internally until the offset is known, and define | |
2999 | @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number | |
3000 | used for the frame pointer. | |
3001 | ||
3002 | You should define this macro only in the very rare circumstances when it | |
3003 | is not possible to calculate the offset between the frame pointer and | |
3004 | the automatic variables until after register allocation has been | |
3005 | completed. When this macro is defined, you must also indicate in your | |
3006 | definition of @code{ELIMINABLE_REGS} how to eliminate | |
3007 | @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM} | |
3008 | or @code{STACK_POINTER_REGNUM}. | |
3009 | ||
3010 | Do not define this macro if it would be the same as | |
3011 | @code{FRAME_POINTER_REGNUM}. | |
3012 | @end defmac | |
3013 | ||
3014 | @defmac ARG_POINTER_REGNUM | |
3015 | The register number of the arg pointer register, which is used to access | |
3016 | the function's argument list. On some machines, this is the same as the | |
3017 | frame pointer register. On some machines, the hardware determines which | |
3018 | register this is. On other machines, you can choose any register you | |
3019 | wish for this purpose. If this is not the same register as the frame | |
3020 | pointer register, then you must mark it as a fixed register according to | |
3021 | @code{FIXED_REGISTERS}, or arrange to be able to eliminate it | |
3022 | (@pxref{Elimination}). | |
3023 | @end defmac | |
3024 | ||
3025 | @defmac HARD_FRAME_POINTER_IS_FRAME_POINTER | |
3026 | Define this to a preprocessor constant that is nonzero if | |
3027 | @code{hard_frame_pointer_rtx} and @code{frame_pointer_rtx} should be | |
3028 | the same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM | |
3029 | == FRAME_POINTER_REGNUM)}; you only need to define this macro if that | |
3030 | definition is not suitable for use in preprocessor conditionals. | |
3031 | @end defmac | |
3032 | ||
3033 | @defmac HARD_FRAME_POINTER_IS_ARG_POINTER | |
3034 | Define this to a preprocessor constant that is nonzero if | |
3035 | @code{hard_frame_pointer_rtx} and @code{arg_pointer_rtx} should be the | |
3036 | same. The default definition is @samp{(HARD_FRAME_POINTER_REGNUM == | |
3037 | ARG_POINTER_REGNUM)}; you only need to define this macro if that | |
3038 | definition is not suitable for use in preprocessor conditionals. | |
3039 | @end defmac | |
3040 | ||
3041 | @defmac RETURN_ADDRESS_POINTER_REGNUM | |
3042 | The register number of the return address pointer register, which is used to | |
3043 | access the current function's return address from the stack. On some | |
3044 | machines, the return address is not at a fixed offset from the frame | |
3045 | pointer or stack pointer or argument pointer. This register can be defined | |
3046 | to point to the return address on the stack, and then be converted by | |
3047 | @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer. | |
3048 | ||
3049 | Do not define this macro unless there is no other way to get the return | |
3050 | address from the stack. | |
3051 | @end defmac | |
3052 | ||
3053 | @defmac STATIC_CHAIN_REGNUM | |
3054 | @defmacx STATIC_CHAIN_INCOMING_REGNUM | |
3055 | Register numbers used for passing a function's static chain pointer. If | |
3056 | register windows are used, the register number as seen by the called | |
3057 | function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register | |
3058 | number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If | |
3059 | these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need | |
3060 | not be defined. | |
3061 | ||
3062 | The static chain register need not be a fixed register. | |
3063 | ||
3064 | If the static chain is passed in memory, these macros should not be | |
3065 | defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used. | |
3066 | @end defmac | |
3067 | ||
3068 | @hook TARGET_STATIC_CHAIN | |
3069 | ||
3070 | @defmac DWARF_FRAME_REGISTERS | |
3071 | This macro specifies the maximum number of hard registers that can be | |
3072 | saved in a call frame. This is used to size data structures used in | |
3073 | DWARF2 exception handling. | |
3074 | ||
3075 | Prior to GCC 3.0, this macro was needed in order to establish a stable | |
3076 | exception handling ABI in the face of adding new hard registers for ISA | |
3077 | extensions. In GCC 3.0 and later, the EH ABI is insulated from changes | |
3078 | in the number of hard registers. Nevertheless, this macro can still be | |
3079 | used to reduce the runtime memory requirements of the exception handling | |
3080 | routines, which can be substantial if the ISA contains a lot of | |
3081 | registers that are not call-saved. | |
3082 | ||
3083 | If this macro is not defined, it defaults to | |
3084 | @code{FIRST_PSEUDO_REGISTER}. | |
3085 | @end defmac | |
3086 | ||
3087 | @defmac PRE_GCC3_DWARF_FRAME_REGISTERS | |
3088 | ||
3089 | This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided | |
3090 | for backward compatibility in pre GCC 3.0 compiled code. | |
3091 | ||
3092 | If this macro is not defined, it defaults to | |
3093 | @code{DWARF_FRAME_REGISTERS}. | |
3094 | @end defmac | |
3095 | ||
3096 | @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno}) | |
3097 | ||
3098 | Define this macro if the target's representation for dwarf registers | |
3099 | is different than the internal representation for unwind column. | |
3100 | Given a dwarf register, this macro should return the internal unwind | |
3101 | column number to use instead. | |
3102 | @end defmac | |
3103 | ||
3104 | @defmac DWARF_FRAME_REGNUM (@var{regno}) | |
3105 | ||
3106 | Define this macro if the target's representation for dwarf registers | |
3107 | used in .eh_frame or .debug_frame is different from that used in other | |
3108 | debug info sections. Given a GCC hard register number, this macro | |
3109 | should return the .eh_frame register number. The default is | |
3110 | @code{DEBUGGER_REGNO (@var{regno})}. | |
d77de738 ML |
3111 | @end defmac |
3112 | ||
3113 | @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh}) | |
3114 | ||
3115 | Define this macro to map register numbers held in the call frame info | |
3116 | that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that | |
3117 | should be output in .debug_frame (@code{@var{for_eh}} is zero) and | |
3118 | .eh_frame (@code{@var{for_eh}} is nonzero). The default is to | |
3119 | return @code{@var{regno}}. | |
3120 | ||
3121 | @end defmac | |
3122 | ||
3123 | @defmac REG_VALUE_IN_UNWIND_CONTEXT | |
3124 | ||
3125 | Define this macro if the target stores register values as | |
3126 | @code{_Unwind_Word} type in unwind context. It should be defined if | |
3127 | target register size is larger than the size of @code{void *}. The | |
3128 | default is to store register values as @code{void *} type. | |
3129 | ||
3130 | @end defmac | |
3131 | ||
3132 | @defmac ASSUME_EXTENDED_UNWIND_CONTEXT | |
3133 | ||
3134 | Define this macro to be 1 if the target always uses extended unwind | |
3135 | context with version, args_size and by_value fields. If it is undefined, | |
3136 | it will be defined to 1 when @code{REG_VALUE_IN_UNWIND_CONTEXT} is | |
3137 | defined and 0 otherwise. | |
3138 | ||
3139 | @end defmac | |
3140 | ||
3141 | @defmac DWARF_LAZY_REGISTER_VALUE (@var{regno}, @var{value}) | |
3142 | Define this macro if the target has pseudo DWARF registers whose | |
3143 | values need to be computed lazily on demand by the unwinder (such as when | |
3144 | referenced in a CFA expression). The macro returns true if @var{regno} | |
3145 | is such a register and stores its value in @samp{*@var{value}} if so. | |
3146 | @end defmac | |
3147 | ||
3148 | @node Elimination | |
3149 | @subsection Eliminating Frame Pointer and Arg Pointer | |
3150 | ||
3151 | @c prevent bad page break with this line | |
3152 | This is about eliminating the frame pointer and arg pointer. | |
3153 | ||
3154 | @hook TARGET_FRAME_POINTER_REQUIRED | |
3155 | ||
3156 | @defmac ELIMINABLE_REGS | |
3157 | This macro specifies a table of register pairs used to eliminate | |
3158 | unneeded registers that point into the stack frame. | |
3159 | ||
3160 | The definition of this macro is a list of structure initializations, each | |
3161 | of which specifies an original and replacement register. | |
3162 | ||
3163 | On some machines, the position of the argument pointer is not known until | |
3164 | the compilation is completed. In such a case, a separate hard register | |
3165 | must be used for the argument pointer. This register can be eliminated by | |
3166 | replacing it with either the frame pointer or the argument pointer, | |
3167 | depending on whether or not the frame pointer has been eliminated. | |
3168 | ||
3169 | In this case, you might specify: | |
3170 | @smallexample | |
3171 | #define ELIMINABLE_REGS \ | |
3172 | @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \ | |
3173 | @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \ | |
3174 | @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@} | |
3175 | @end smallexample | |
3176 | ||
3177 | Note that the elimination of the argument pointer with the stack pointer is | |
3178 | specified first since that is the preferred elimination. | |
3179 | @end defmac | |
3180 | ||
5bfb91c1 GJL |
3181 | @defmac RELOAD_ELIMINABLE_REGS |
3182 | Like @code{ELIMINABLE_REGS}, but only used in the old reload framework where | |
3183 | it takes precedence over @code{ELIMINABLE_REGS}. This macro can be useful | |
3184 | during the transition to LRA because there are cases where reload and LRA | |
3185 | disagree on how eliminable registers should be represented. For an example, | |
3186 | see @file{avr.h}. | |
3187 | @end defmac | |
3188 | ||
d77de738 ML |
3189 | @hook TARGET_CAN_ELIMINATE |
3190 | ||
3191 | @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var}) | |
3192 | This macro returns the initial difference between the specified pair | |
3193 | of registers. The value would be computed from information | |
3194 | such as the result of @code{get_frame_size ()} and the tables of | |
3195 | registers @code{df_regs_ever_live_p} and @code{call_used_regs}. | |
3196 | @end defmac | |
3197 | ||
3198 | @hook TARGET_COMPUTE_FRAME_LAYOUT | |
3199 | ||
3200 | @node Stack Arguments | |
3201 | @subsection Passing Function Arguments on the Stack | |
3202 | @cindex arguments on stack | |
3203 | @cindex stack arguments | |
3204 | ||
3205 | The macros in this section control how arguments are passed | |
3206 | on the stack. See the following section for other macros that | |
3207 | control passing certain arguments in registers. | |
3208 | ||
3209 | @hook TARGET_PROMOTE_PROTOTYPES | |
3210 | ||
3211 | @hook TARGET_PUSH_ARGUMENT | |
3212 | ||
3213 | @defmac PUSH_ARGS_REVERSED | |
3214 | A C expression. If nonzero, function arguments will be evaluated from | |
3215 | last to first, rather than from first to last. If this macro is not | |
3216 | defined, it defaults to @code{PUSH_ARGS} on targets where the stack | |
3217 | and args grow in opposite directions, and 0 otherwise. | |
3218 | @end defmac | |
3219 | ||
3220 | @defmac PUSH_ROUNDING (@var{npushed}) | |
3221 | A C expression that is the number of bytes actually pushed onto the | |
3222 | stack when an instruction attempts to push @var{npushed} bytes. | |
3223 | ||
3224 | On some machines, the definition | |
3225 | ||
3226 | @smallexample | |
3227 | #define PUSH_ROUNDING(BYTES) (BYTES) | |
3228 | @end smallexample | |
3229 | ||
3230 | @noindent | |
3231 | will suffice. But on other machines, instructions that appear | |
3232 | to push one byte actually push two bytes in an attempt to maintain | |
3233 | alignment. Then the definition should be | |
3234 | ||
3235 | @smallexample | |
3236 | #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1) | |
3237 | @end smallexample | |
3238 | ||
3239 | If the value of this macro has a type, it should be an unsigned type. | |
3240 | @end defmac | |
3241 | ||
3242 | @findex outgoing_args_size | |
3243 | @findex crtl->outgoing_args_size | |
3244 | @defmac ACCUMULATE_OUTGOING_ARGS | |
3245 | A C expression. If nonzero, the maximum amount of space required for outgoing arguments | |
3246 | will be computed and placed into | |
3247 | @code{crtl->outgoing_args_size}. No space will be pushed | |
3248 | onto the stack for each call; instead, the function prologue should | |
3249 | increase the stack frame size by this amount. | |
3250 | ||
3251 | Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS} | |
3252 | is not proper. | |
3253 | @end defmac | |
3254 | ||
3255 | @defmac REG_PARM_STACK_SPACE (@var{fndecl}) | |
3256 | Define this macro if functions should assume that stack space has been | |
3257 | allocated for arguments even when their values are passed in | |
3258 | registers. | |
3259 | ||
3260 | The value of this macro is the size, in bytes, of the area reserved for | |
3261 | arguments passed in registers for the function represented by @var{fndecl}, | |
3262 | which can be zero if GCC is calling a library function. | |
3263 | The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself | |
3264 | of the function. | |
3265 | ||
3266 | This space can be allocated by the caller, or be a part of the | |
3267 | machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says | |
3268 | which. | |
3269 | @end defmac | |
3270 | @c above is overfull. not sure what to do. --mew 5feb93 did | |
3271 | @c something, not sure if it looks good. --mew 10feb93 | |
3272 | ||
3273 | @defmac INCOMING_REG_PARM_STACK_SPACE (@var{fndecl}) | |
3274 | Like @code{REG_PARM_STACK_SPACE}, but for incoming register arguments. | |
3275 | Define this macro if space guaranteed when compiling a function body | |
3276 | is different to space required when making a call, a situation that | |
3277 | can arise with K&R style function definitions. | |
3278 | @end defmac | |
3279 | ||
3280 | @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype}) | |
3281 | Define this to a nonzero value if it is the responsibility of the | |
3282 | caller to allocate the area reserved for arguments passed in registers | |
3283 | when calling a function of @var{fntype}. @var{fntype} may be NULL | |
3284 | if the function called is a library function. | |
3285 | ||
3286 | If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls | |
3287 | whether the space for these arguments counts in the value of | |
3288 | @code{crtl->outgoing_args_size}. | |
3289 | @end defmac | |
3290 | ||
3291 | @defmac STACK_PARMS_IN_REG_PARM_AREA | |
3292 | Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the | |
3293 | stack parameters don't skip the area specified by it. | |
3294 | @c i changed this, makes more sens and it should have taken care of the | |
3295 | @c overfull.. not as specific, tho. --mew 5feb93 | |
3296 | ||
3297 | Normally, when a parameter is not passed in registers, it is placed on the | |
3298 | stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro | |
3299 | suppresses this behavior and causes the parameter to be passed on the | |
3300 | stack in its natural location. | |
3301 | @end defmac | |
3302 | ||
3303 | @hook TARGET_RETURN_POPS_ARGS | |
3304 | ||
3305 | @defmac CALL_POPS_ARGS (@var{cum}) | |
3306 | A C expression that should indicate the number of bytes a call sequence | |
3307 | pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS} | |
3308 | when compiling a function call. | |
3309 | ||
3310 | @var{cum} is the variable in which all arguments to the called function | |
3311 | have been accumulated. | |
3312 | ||
3313 | On certain architectures, such as the SH5, a call trampoline is used | |
3314 | that pops certain registers off the stack, depending on the arguments | |
3315 | that have been passed to the function. Since this is a property of the | |
3316 | call site, not of the called function, @code{RETURN_POPS_ARGS} is not | |
3317 | appropriate. | |
3318 | @end defmac | |
3319 | ||
3320 | @node Register Arguments | |
3321 | @subsection Passing Arguments in Registers | |
3322 | @cindex arguments in registers | |
3323 | @cindex registers arguments | |
3324 | ||
3325 | This section describes the macros which let you control how various | |
3326 | types of arguments are passed in registers or how they are arranged in | |
3327 | the stack. | |
3328 | ||
3329 | @hook TARGET_FUNCTION_ARG | |
3330 | ||
3331 | @hook TARGET_MUST_PASS_IN_STACK | |
3332 | ||
3333 | @hook TARGET_FUNCTION_INCOMING_ARG | |
3334 | ||
3335 | @hook TARGET_USE_PSEUDO_PIC_REG | |
3336 | ||
3337 | @hook TARGET_INIT_PIC_REG | |
3338 | ||
3339 | @hook TARGET_ARG_PARTIAL_BYTES | |
3340 | ||
3341 | @hook TARGET_PASS_BY_REFERENCE | |
3342 | ||
3343 | @hook TARGET_CALLEE_COPIES | |
3344 | ||
3345 | @defmac CUMULATIVE_ARGS | |
3346 | A C type for declaring a variable that is used as the first argument | |
3347 | of @code{TARGET_FUNCTION_ARG} and other related values. For some | |
3348 | target machines, the type @code{int} suffices and can hold the number | |
3349 | of bytes of argument so far. | |
3350 | ||
3351 | There is no need to record in @code{CUMULATIVE_ARGS} anything about the | |
3352 | arguments that have been passed on the stack. The compiler has other | |
3353 | variables to keep track of that. For target machines on which all | |
3354 | arguments are passed on the stack, there is no need to store anything in | |
3355 | @code{CUMULATIVE_ARGS}; however, the data structure must exist and | |
3356 | should not be empty, so use @code{int}. | |
3357 | @end defmac | |
3358 | ||
3359 | @defmac OVERRIDE_ABI_FORMAT (@var{fndecl}) | |
3360 | If defined, this macro is called before generating any code for a | |
3361 | function, but after the @var{cfun} descriptor for the function has been | |
3362 | created. The back end may use this macro to update @var{cfun} to | |
3363 | reflect an ABI other than that which would normally be used by default. | |
3364 | If the compiler is generating code for a compiler-generated function, | |
3365 | @var{fndecl} may be @code{NULL}. | |
3366 | @end defmac | |
3367 | ||
3368 | @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args}) | |
3369 | A C statement (sans semicolon) for initializing the variable | |
3370 | @var{cum} for the state at the beginning of the argument list. The | |
3371 | variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype} | |
3372 | is the tree node for the data type of the function which will receive | |
3373 | the args, or 0 if the args are to a compiler support library function. | |
3374 | For direct calls that are not libcalls, @var{fndecl} contain the | |
3375 | declaration node of the function. @var{fndecl} is also set when | |
3376 | @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function | |
3377 | being compiled. @var{n_named_args} is set to the number of named | |
3378 | arguments, including a structure return address if it is passed as a | |
3379 | parameter, when making a call. When processing incoming arguments, | |
3380 | @var{n_named_args} is set to @minus{}1. | |
3381 | ||
3382 | When processing a call to a compiler support library function, | |
3383 | @var{libname} identifies which one. It is a @code{symbol_ref} rtx which | |
3384 | contains the name of the function, as a string. @var{libname} is 0 when | |
3385 | an ordinary C function call is being processed. Thus, each time this | |
3386 | macro is called, either @var{libname} or @var{fntype} is nonzero, but | |
3387 | never both of them at once. | |
3388 | @end defmac | |
3389 | ||
3390 | @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname}) | |
3391 | Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls, | |
3392 | it gets a @code{MODE} argument instead of @var{fntype}, that would be | |
3393 | @code{NULL}. @var{indirect} would always be zero, too. If this macro | |
3394 | is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname, | |
3395 | 0)} is used instead. | |
3396 | @end defmac | |
3397 | ||
3398 | @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname}) | |
3399 | Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of | |
3400 | finding the arguments for the function being compiled. If this macro is | |
3401 | undefined, @code{INIT_CUMULATIVE_ARGS} is used instead. | |
3402 | ||
3403 | The value passed for @var{libname} is always 0, since library routines | |
3404 | with special calling conventions are never compiled with GCC@. The | |
3405 | argument @var{libname} exists for symmetry with | |
3406 | @code{INIT_CUMULATIVE_ARGS}. | |
3407 | @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe. | |
3408 | @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93 | |
3409 | @end defmac | |
3410 | ||
3411 | @hook TARGET_FUNCTION_ARG_ADVANCE | |
3412 | ||
3413 | @hook TARGET_FUNCTION_ARG_OFFSET | |
3414 | ||
3415 | @hook TARGET_FUNCTION_ARG_PADDING | |
3416 | ||
3417 | @defmac PAD_VARARGS_DOWN | |
3418 | If defined, a C expression which determines whether the default | |
3419 | implementation of va_arg will attempt to pad down before reading the | |
3420 | next argument, if that argument is smaller than its aligned space as | |
3421 | controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such | |
3422 | arguments are padded down if @code{BYTES_BIG_ENDIAN} is true. | |
3423 | @end defmac | |
3424 | ||
3425 | @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first}) | |
3426 | Specify padding for the last element of a block move between registers and | |
3427 | memory. @var{first} is nonzero if this is the only element. Defining this | |
3428 | macro allows better control of register function parameters on big-endian | |
3429 | machines, without using @code{PARALLEL} rtl. In particular, | |
3430 | @code{MUST_PASS_IN_STACK} need not test padding and mode of types in | |
3431 | registers, as there is no longer a "wrong" part of a register; For example, | |
3432 | a three byte aggregate may be passed in the high part of a register if so | |
3433 | required. | |
3434 | @end defmac | |
3435 | ||
3436 | @hook TARGET_FUNCTION_ARG_BOUNDARY | |
3437 | ||
3438 | @hook TARGET_FUNCTION_ARG_ROUND_BOUNDARY | |
3439 | ||
3440 | @defmac FUNCTION_ARG_REGNO_P (@var{regno}) | |
3441 | A C expression that is nonzero if @var{regno} is the number of a hard | |
3442 | register in which function arguments are sometimes passed. This does | |
3443 | @emph{not} include implicit arguments such as the static chain and | |
3444 | the structure-value address. On many machines, no registers can be | |
3445 | used for this purpose since all function arguments are pushed on the | |
3446 | stack. | |
3447 | @end defmac | |
3448 | ||
3449 | @hook TARGET_SPLIT_COMPLEX_ARG | |
3450 | ||
3451 | @hook TARGET_BUILD_BUILTIN_VA_LIST | |
3452 | ||
3453 | @hook TARGET_ENUM_VA_LIST_P | |
3454 | ||
3455 | @hook TARGET_FN_ABI_VA_LIST | |
3456 | ||
3457 | @hook TARGET_CANONICAL_VA_LIST_TYPE | |
3458 | ||
3459 | @hook TARGET_GIMPLIFY_VA_ARG_EXPR | |
3460 | ||
3461 | @hook TARGET_VALID_POINTER_MODE | |
3462 | ||
3463 | @hook TARGET_REF_MAY_ALIAS_ERRNO | |
3464 | ||
b3974356 RB |
3465 | @hook TARGET_MODE_CAN_TRANSFER_BITS |
3466 | ||
d77de738 ML |
3467 | @hook TARGET_TRANSLATE_MODE_ATTRIBUTE |
3468 | ||
3469 | @hook TARGET_SCALAR_MODE_SUPPORTED_P | |
3470 | ||
3471 | @hook TARGET_VECTOR_MODE_SUPPORTED_P | |
3472 | ||
84152985 KT |
3473 | @hook TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P |
3474 | ||
d77de738 ML |
3475 | @hook TARGET_COMPATIBLE_VECTOR_TYPES_P |
3476 | ||
3477 | @hook TARGET_ARRAY_MODE | |
3478 | ||
3479 | @hook TARGET_ARRAY_MODE_SUPPORTED_P | |
3480 | ||
3481 | @hook TARGET_LIBGCC_FLOATING_MODE_SUPPORTED_P | |
3482 | ||
3483 | @hook TARGET_FLOATN_MODE | |
3484 | ||
3485 | @hook TARGET_FLOATN_BUILTIN_P | |
3486 | ||
3487 | @hook TARGET_SMALL_REGISTER_CLASSES_FOR_MODE_P | |
3488 | ||
3489 | @node Scalar Return | |
3490 | @subsection How Scalar Function Values Are Returned | |
3491 | @cindex return values in registers | |
3492 | @cindex values, returned by functions | |
3493 | @cindex scalars, returned as values | |
3494 | ||
3495 | This section discusses the macros that control returning scalars as | |
3496 | values---values that can fit in registers. | |
3497 | ||
3498 | @hook TARGET_FUNCTION_VALUE | |
3499 | ||
3500 | @defmac FUNCTION_VALUE (@var{valtype}, @var{func}) | |
3501 | This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for | |
3502 | a new target instead. | |
3503 | @end defmac | |
3504 | ||
3505 | @defmac LIBCALL_VALUE (@var{mode}) | |
3506 | A C expression to create an RTX representing the place where a library | |
3507 | function returns a value of mode @var{mode}. | |
3508 | ||
3509 | Note that ``library function'' in this context means a compiler | |
3510 | support routine, used to perform arithmetic, whose name is known | |
3511 | specially by the compiler and was not mentioned in the C code being | |
3512 | compiled. | |
3513 | @end defmac | |
3514 | ||
3515 | @hook TARGET_LIBCALL_VALUE | |
3516 | ||
3517 | @defmac FUNCTION_VALUE_REGNO_P (@var{regno}) | |
3518 | A C expression that is nonzero if @var{regno} is the number of a hard | |
3519 | register in which the values of called function may come back. | |
3520 | ||
3521 | A register whose use for returning values is limited to serving as the | |
3522 | second of a pair (for a value of type @code{double}, say) need not be | |
3523 | recognized by this macro. So for most machines, this definition | |
3524 | suffices: | |
3525 | ||
3526 | @smallexample | |
3527 | #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0) | |
3528 | @end smallexample | |
3529 | ||
3530 | If the machine has register windows, so that the caller and the called | |
3531 | function use different registers for the return value, this macro | |
3532 | should recognize only the caller's register numbers. | |
3533 | ||
3534 | This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE_REGNO_P} | |
3535 | for a new target instead. | |
3536 | @end defmac | |
3537 | ||
3538 | @hook TARGET_FUNCTION_VALUE_REGNO_P | |
3539 | ||
3540 | @defmac APPLY_RESULT_SIZE | |
3541 | Define this macro if @samp{untyped_call} and @samp{untyped_return} | |
3542 | need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for | |
3543 | saving and restoring an arbitrary return value. | |
3544 | @end defmac | |
3545 | ||
3546 | @hook TARGET_OMIT_STRUCT_RETURN_REG | |
3547 | ||
3548 | @hook TARGET_RETURN_IN_MSB | |
3549 | ||
3550 | @node Aggregate Return | |
3551 | @subsection How Large Values Are Returned | |
3552 | @cindex aggregates as return values | |
3553 | @cindex large return values | |
3554 | @cindex returning aggregate values | |
3555 | @cindex structure value address | |
3556 | ||
3557 | When a function value's mode is @code{BLKmode} (and in some other | |
3558 | cases), the value is not returned according to | |
3559 | @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the | |
3560 | caller passes the address of a block of memory in which the value | |
3561 | should be stored. This address is called the @dfn{structure value | |
3562 | address}. | |
3563 | ||
3564 | This section describes how to control returning structure values in | |
3565 | memory. | |
3566 | ||
3567 | @hook TARGET_RETURN_IN_MEMORY | |
3568 | ||
3569 | @defmac DEFAULT_PCC_STRUCT_RETURN | |
3570 | Define this macro to be 1 if all structure and union return values must be | |
3571 | in memory. Since this results in slower code, this should be defined | |
3572 | only if needed for compatibility with other compilers or with an ABI@. | |
3573 | If you define this macro to be 0, then the conventions used for structure | |
3574 | and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY} | |
3575 | target hook. | |
3576 | ||
3577 | If not defined, this defaults to the value 1. | |
3578 | @end defmac | |
3579 | ||
3580 | @hook TARGET_STRUCT_VALUE_RTX | |
3581 | ||
3582 | @defmac PCC_STATIC_STRUCT_RETURN | |
3583 | Define this macro if the usual system convention on the target machine | |
3584 | for returning structures and unions is for the called function to return | |
3585 | the address of a static variable containing the value. | |
3586 | ||
3587 | Do not define this if the usual system convention is for the caller to | |
3588 | pass an address to the subroutine. | |
3589 | ||
3590 | This macro has effect in @option{-fpcc-struct-return} mode, but it does | |
3591 | nothing when you use @option{-freg-struct-return} mode. | |
3592 | @end defmac | |
3593 | ||
3594 | @hook TARGET_GET_RAW_RESULT_MODE | |
3595 | ||
3596 | @hook TARGET_GET_RAW_ARG_MODE | |
3597 | ||
3598 | @hook TARGET_EMPTY_RECORD_P | |
3599 | ||
3600 | @hook TARGET_WARN_PARAMETER_PASSING_ABI | |
3601 | ||
3602 | @node Caller Saves | |
3603 | @subsection Caller-Saves Register Allocation | |
3604 | ||
3605 | If you enable it, GCC can save registers around function calls. This | |
3606 | makes it possible to use call-clobbered registers to hold variables that | |
3607 | must live across calls. | |
3608 | ||
3609 | @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs}) | |
3610 | A C expression specifying which mode is required for saving @var{nregs} | |
3611 | of a pseudo-register in call-clobbered hard register @var{regno}. If | |
3612 | @var{regno} is unsuitable for caller save, @code{VOIDmode} should be | |
3613 | returned. For most machines this macro need not be defined since GCC | |
3614 | will select the smallest suitable mode. | |
3615 | @end defmac | |
3616 | ||
3617 | @node Function Entry | |
3618 | @subsection Function Entry and Exit | |
3619 | @cindex function entry and exit | |
3620 | @cindex prologue | |
3621 | @cindex epilogue | |
3622 | ||
3623 | This section describes the macros that output function entry | |
3624 | (@dfn{prologue}) and exit (@dfn{epilogue}) code. | |
3625 | ||
3626 | @hook TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY | |
3627 | ||
3628 | @hook TARGET_ASM_FUNCTION_PROLOGUE | |
3629 | ||
3630 | @hook TARGET_ASM_FUNCTION_END_PROLOGUE | |
3631 | ||
3632 | @hook TARGET_ASM_FUNCTION_BEGIN_EPILOGUE | |
3633 | ||
3634 | @hook TARGET_ASM_FUNCTION_EPILOGUE | |
3635 | ||
3636 | @itemize @bullet | |
d77de738 ML |
3637 | @findex pretend_args_size |
3638 | @findex crtl->args.pretend_args_size | |
f33d7a88 | 3639 | @item |
d77de738 ML |
3640 | A region of @code{crtl->args.pretend_args_size} bytes of |
3641 | uninitialized space just underneath the first argument arriving on the | |
3642 | stack. (This may not be at the very start of the allocated stack region | |
3643 | if the calling sequence has pushed anything else since pushing the stack | |
3644 | arguments. But usually, on such machines, nothing else has been pushed | |
3645 | yet, because the function prologue itself does all the pushing.) This | |
3646 | region is used on machines where an argument may be passed partly in | |
3647 | registers and partly in memory, and, in some cases to support the | |
3648 | features in @code{<stdarg.h>}. | |
3649 | ||
3650 | @item | |
3651 | An area of memory used to save certain registers used by the function. | |
3652 | The size of this area, which may also include space for such things as | |
3653 | the return address and pointers to previous stack frames, is | |
3654 | machine-specific and usually depends on which registers have been used | |
3655 | in the function. Machines with register windows often do not require | |
3656 | a save area. | |
3657 | ||
3658 | @item | |
3659 | A region of at least @var{size} bytes, possibly rounded up to an allocation | |
3660 | boundary, to contain the local variables of the function. On some machines, | |
3661 | this region and the save area may occur in the opposite order, with the | |
3662 | save area closer to the top of the stack. | |
3663 | ||
d77de738 | 3664 | @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames |
f33d7a88 | 3665 | @item |
d77de738 ML |
3666 | Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of |
3667 | @code{crtl->outgoing_args_size} bytes to be used for outgoing | |
3668 | argument lists of the function. @xref{Stack Arguments}. | |
3669 | @end itemize | |
3670 | ||
3671 | @defmac EXIT_IGNORE_STACK | |
3672 | Define this macro as a C expression that is nonzero if the return | |
3673 | instruction or the function epilogue ignores the value of the stack | |
3674 | pointer; in other words, if it is safe to delete an instruction to | |
3675 | adjust the stack pointer before a return from the function. The | |
3676 | default is 0. | |
3677 | ||
3678 | Note that this macro's value is relevant only for functions for which | |
3679 | frame pointers are maintained. It is never safe to delete a final | |
3680 | stack adjustment in a function that has no frame pointer, and the | |
3681 | compiler knows this regardless of @code{EXIT_IGNORE_STACK}. | |
3682 | @end defmac | |
3683 | ||
3684 | @defmac EPILOGUE_USES (@var{regno}) | |
3685 | Define this macro as a C expression that is nonzero for registers that are | |
3686 | used by the epilogue or the @samp{return} pattern. The stack and frame | |
3687 | pointer registers are already assumed to be used as needed. | |
3688 | @end defmac | |
3689 | ||
3690 | @defmac EH_USES (@var{regno}) | |
3691 | Define this macro as a C expression that is nonzero for registers that are | |
3692 | used by the exception handling mechanism, and so should be considered live | |
3693 | on entry to an exception edge. | |
3694 | @end defmac | |
3695 | ||
3696 | @hook TARGET_ASM_OUTPUT_MI_THUNK | |
3697 | ||
3698 | @hook TARGET_ASM_CAN_OUTPUT_MI_THUNK | |
3699 | ||
3700 | @node Profiling | |
3701 | @subsection Generating Code for Profiling | |
3702 | @cindex profiling, code generation | |
3703 | ||
3704 | These macros will help you generate code for profiling. | |
3705 | ||
3706 | @defmac FUNCTION_PROFILER (@var{file}, @var{labelno}) | |
3707 | A C statement or compound statement to output to @var{file} some | |
3708 | assembler code to call the profiling subroutine @code{mcount}. | |
3709 | ||
3710 | @findex mcount | |
3711 | The details of how @code{mcount} expects to be called are determined by | |
3712 | your operating system environment, not by GCC@. To figure them out, | |
3713 | compile a small program for profiling using the system's installed C | |
3714 | compiler and look at the assembler code that results. | |
3715 | ||
3716 | Older implementations of @code{mcount} expect the address of a counter | |
3717 | variable to be loaded into some register. The name of this variable is | |
3718 | @samp{LP} followed by the number @var{labelno}, so you would generate | |
3719 | the name using @samp{LP%d} in a @code{fprintf}. | |
3720 | @end defmac | |
3721 | ||
3722 | @defmac PROFILE_HOOK | |
3723 | A C statement or compound statement to output to @var{file} some assembly | |
3724 | code to call the profiling subroutine @code{mcount} even the target does | |
3725 | not support profiling. | |
3726 | @end defmac | |
3727 | ||
3728 | @defmac NO_PROFILE_COUNTERS | |
3729 | Define this macro to be an expression with a nonzero value if the | |
3730 | @code{mcount} subroutine on your system does not need a counter variable | |
3731 | allocated for each function. This is true for almost all modern | |
3732 | implementations. If you define this macro, you must not use the | |
3733 | @var{labelno} argument to @code{FUNCTION_PROFILER}. | |
3734 | @end defmac | |
3735 | ||
3736 | @defmac PROFILE_BEFORE_PROLOGUE | |
3737 | Define this macro if the code for function profiling should come before | |
3738 | the function prologue. Normally, the profiling code comes after. | |
3739 | @end defmac | |
3740 | ||
3741 | @hook TARGET_KEEP_LEAF_WHEN_PROFILED | |
3742 | ||
3743 | @node Tail Calls | |
3744 | @subsection Permitting tail calls | |
3745 | @cindex tail calls | |
3746 | ||
3747 | @hook TARGET_FUNCTION_OK_FOR_SIBCALL | |
3748 | ||
3749 | @hook TARGET_EXTRA_LIVE_ON_ENTRY | |
3750 | ||
3751 | @hook TARGET_SET_UP_BY_PROLOGUE | |
3752 | ||
3753 | @hook TARGET_WARN_FUNC_RETURN | |
3754 | ||
3755 | @node Shrink-wrapping separate components | |
3756 | @subsection Shrink-wrapping separate components | |
3757 | @cindex shrink-wrapping separate components | |
3758 | ||
3759 | The prologue may perform a variety of target dependent tasks such as | |
3760 | saving callee-saved registers, saving the return address, aligning the | |
3761 | stack, creating a stack frame, initializing the PIC register, setting | |
3762 | up the static chain, etc. | |
3763 | ||
3764 | On some targets some of these tasks may be independent of others and | |
3765 | thus may be shrink-wrapped separately. These independent tasks are | |
3766 | referred to as components and are handled generically by the target | |
3767 | independent parts of GCC. | |
3768 | ||
3769 | Using the following hooks those prologue or epilogue components can be | |
3770 | shrink-wrapped separately, so that the initialization (and possibly | |
3771 | teardown) those components do is not done as frequently on execution | |
3772 | paths where this would unnecessary. | |
3773 | ||
3774 | What exactly those components are is up to the target code; the generic | |
3775 | code treats them abstractly, as a bit in an @code{sbitmap}. These | |
3776 | @code{sbitmap}s are allocated by the @code{shrink_wrap.get_separate_components} | |
3777 | and @code{shrink_wrap.components_for_bb} hooks, and deallocated by the | |
3778 | generic code. | |
3779 | ||
3780 | @hook TARGET_SHRINK_WRAP_GET_SEPARATE_COMPONENTS | |
3781 | ||
3782 | @hook TARGET_SHRINK_WRAP_COMPONENTS_FOR_BB | |
3783 | ||
3784 | @hook TARGET_SHRINK_WRAP_DISQUALIFY_COMPONENTS | |
3785 | ||
3786 | @hook TARGET_SHRINK_WRAP_EMIT_PROLOGUE_COMPONENTS | |
3787 | ||
3788 | @hook TARGET_SHRINK_WRAP_EMIT_EPILOGUE_COMPONENTS | |
3789 | ||
3790 | @hook TARGET_SHRINK_WRAP_SET_HANDLED_COMPONENTS | |
3791 | ||
3792 | @node Stack Smashing Protection | |
3793 | @subsection Stack smashing protection | |
3794 | @cindex stack smashing protection | |
3795 | ||
3796 | @hook TARGET_STACK_PROTECT_GUARD | |
3797 | ||
3798 | @hook TARGET_STACK_PROTECT_FAIL | |
3799 | ||
3800 | @hook TARGET_STACK_PROTECT_RUNTIME_ENABLED_P | |
3801 | ||
3802 | @hook TARGET_SUPPORTS_SPLIT_STACK | |
3803 | ||
3804 | @hook TARGET_GET_VALID_OPTION_VALUES | |
3805 | ||
3806 | @node Miscellaneous Register Hooks | |
3807 | @subsection Miscellaneous register hooks | |
3808 | @cindex miscellaneous register hooks | |
3809 | ||
3810 | @hook TARGET_CALL_FUSAGE_CONTAINS_NON_CALLEE_CLOBBERS | |
3811 | ||
3812 | @node Varargs | |
3813 | @section Implementing the Varargs Macros | |
3814 | @cindex varargs implementation | |
3815 | ||
3816 | GCC comes with an implementation of @code{<varargs.h>} and | |
3817 | @code{<stdarg.h>} that work without change on machines that pass arguments | |
3818 | on the stack. Other machines require their own implementations of | |
3819 | varargs, and the two machine independent header files must have | |
3820 | conditionals to include it. | |
3821 | ||
3822 | ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in | |
3823 | the calling convention for @code{va_start}. The traditional | |
3824 | implementation takes just one argument, which is the variable in which | |
3825 | to store the argument pointer. The ISO implementation of | |
3826 | @code{va_start} takes an additional second argument. The user is | |
3827 | supposed to write the last named argument of the function here. | |
3828 | ||
3829 | However, @code{va_start} should not use this argument. The way to find | |
3830 | the end of the named arguments is with the built-in functions described | |
3831 | below. | |
3832 | ||
3833 | @defmac __builtin_saveregs () | |
3834 | Use this built-in function to save the argument registers in memory so | |
3835 | that the varargs mechanism can access them. Both ISO and traditional | |
3836 | versions of @code{va_start} must use @code{__builtin_saveregs}, unless | |
3837 | you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead. | |
3838 | ||
3839 | On some machines, @code{__builtin_saveregs} is open-coded under the | |
3840 | control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On | |
3841 | other machines, it calls a routine written in assembler language, | |
3842 | found in @file{libgcc2.c}. | |
3843 | ||
3844 | Code generated for the call to @code{__builtin_saveregs} appears at the | |
3845 | beginning of the function, as opposed to where the call to | |
3846 | @code{__builtin_saveregs} is written, regardless of what the code is. | |
3847 | This is because the registers must be saved before the function starts | |
3848 | to use them for its own purposes. | |
3849 | @c i rewrote the first sentence above to fix an overfull hbox. --mew | |
3850 | @c 10feb93 | |
3851 | @end defmac | |
3852 | ||
3853 | @defmac __builtin_next_arg (@var{lastarg}) | |
3854 | This builtin returns the address of the first anonymous stack | |
3855 | argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it | |
3856 | returns the address of the location above the first anonymous stack | |
3857 | argument. Use it in @code{va_start} to initialize the pointer for | |
3858 | fetching arguments from the stack. Also use it in @code{va_start} to | |
3859 | verify that the second parameter @var{lastarg} is the last named argument | |
3860 | of the current function. | |
3861 | @end defmac | |
3862 | ||
3863 | @defmac __builtin_classify_type (@var{object}) | |
3864 | Since each machine has its own conventions for which data types are | |
3865 | passed in which kind of register, your implementation of @code{va_arg} | |
3866 | has to embody these conventions. The easiest way to categorize the | |
3867 | specified data type is to use @code{__builtin_classify_type} together | |
3868 | with @code{sizeof} and @code{__alignof__}. | |
3869 | ||
3870 | @code{__builtin_classify_type} ignores the value of @var{object}, | |
3871 | considering only its data type. It returns an integer describing what | |
3872 | kind of type that is---integer, floating, pointer, structure, and so on. | |
3873 | ||
3874 | The file @file{typeclass.h} defines an enumeration that you can use to | |
3875 | interpret the values of @code{__builtin_classify_type}. | |
3876 | @end defmac | |
3877 | ||
3878 | These machine description macros help implement varargs: | |
3879 | ||
3880 | @hook TARGET_EXPAND_BUILTIN_SAVEREGS | |
3881 | ||
3882 | @hook TARGET_SETUP_INCOMING_VARARGS | |
3883 | ||
3884 | @hook TARGET_STRICT_ARGUMENT_NAMING | |
3885 | ||
3472b574 AO |
3886 | @hook TARGET_CALL_OFFSET_RETURN_LABEL |
3887 | ||
672fad57 RS |
3888 | @hook TARGET_START_CALL_ARGS |
3889 | ||
d77de738 ML |
3890 | @hook TARGET_CALL_ARGS |
3891 | ||
3892 | @hook TARGET_END_CALL_ARGS | |
3893 | ||
3894 | @hook TARGET_PRETEND_OUTGOING_VARARGS_NAMED | |
3895 | ||
3896 | @node Trampolines | |
3897 | @section Support for Nested Functions | |
3898 | @cindex support for nested functions | |
3899 | @cindex trampolines for nested functions | |
3900 | @cindex descriptors for nested functions | |
3901 | @cindex nested functions, support for | |
3902 | ||
3903 | Taking the address of a nested function requires special compiler | |
3904 | handling to ensure that the static chain register is loaded when | |
3905 | the function is invoked via an indirect call. | |
3906 | ||
3907 | GCC has traditionally supported nested functions by creating an | |
3908 | executable @dfn{trampoline} at run time when the address of a nested | |
3909 | function is taken. This is a small piece of code which normally | |
3910 | resides on the stack, in the stack frame of the containing function. | |
3911 | The trampoline loads the static chain register and then jumps to the | |
3912 | real address of the nested function. | |
3913 | ||
3914 | The use of trampolines requires an executable stack, which is a | |
3915 | security risk. To avoid this problem, GCC also supports another | |
3916 | strategy: using descriptors for nested functions. Under this model, | |
3917 | taking the address of a nested function results in a pointer to a | |
3918 | non-executable function descriptor object. Initializing the static chain | |
3919 | from the descriptor is handled at indirect call sites. | |
3920 | ||
3921 | On some targets, including HPPA and IA-64, function descriptors may be | |
3922 | mandated by the ABI or be otherwise handled in a target-specific way | |
3923 | by the back end in its code generation strategy for indirect calls. | |
3924 | GCC also provides its own generic descriptor implementation to support the | |
3925 | @option{-fno-trampolines} option. In this case runtime detection of | |
3926 | function descriptors at indirect call sites relies on descriptor | |
3927 | pointers being tagged with a bit that is never set in bare function | |
3928 | addresses. Since GCC's generic function descriptors are | |
3929 | not ABI-compliant, this option is typically used only on a | |
3930 | per-language basis (notably by Ada) or when it can otherwise be | |
3931 | applied to the whole program. | |
3932 | ||
3933 | For languages other than Ada, the @code{-ftrampolines} and | |
3934 | @code{-fno-trampolines} options currently have no effect, and | |
3935 | trampolines are always generated on platforms that need them | |
3936 | for nested functions. | |
3937 | ||
3938 | Define the following hook if your backend either implements ABI-specified | |
3939 | descriptor support, or can use GCC's generic descriptor implementation | |
3940 | for nested functions. | |
3941 | ||
3942 | @hook TARGET_CUSTOM_FUNCTION_DESCRIPTORS | |
3943 | ||
3944 | The following macros tell GCC how to generate code to allocate and | |
3945 | initialize an executable trampoline. You can also use this interface | |
3946 | if your back end needs to create ABI-specified non-executable descriptors; in | |
3947 | this case the "trampoline" created is the descriptor containing data only. | |
3948 | ||
3949 | The instructions in an executable trampoline must do two things: load | |
3950 | a constant address into the static chain register, and jump to the real | |
3951 | address of the nested function. On CISC machines such as the m68k, | |
3952 | this requires two instructions, a move immediate and a jump. Then the | |
3953 | two addresses exist in the trampoline as word-long immediate operands. | |
3954 | On RISC machines, it is often necessary to load each address into a | |
3955 | register in two parts. Then pieces of each address form separate | |
3956 | immediate operands. | |
3957 | ||
3958 | The code generated to initialize the trampoline must store the variable | |
3959 | parts---the static chain value and the function address---into the | |
3960 | immediate operands of the instructions. On a CISC machine, this is | |
3961 | simply a matter of copying each address to a memory reference at the | |
3962 | proper offset from the start of the trampoline. On a RISC machine, it | |
3963 | may be necessary to take out pieces of the address and store them | |
3964 | separately. | |
3965 | ||
3966 | @hook TARGET_ASM_TRAMPOLINE_TEMPLATE | |
3967 | ||
3968 | @defmac TRAMPOLINE_SECTION | |
3969 | Return the section into which the trampoline template is to be placed | |
3970 | (@pxref{Sections}). The default value is @code{readonly_data_section}. | |
3971 | @end defmac | |
3972 | ||
3973 | @defmac TRAMPOLINE_SIZE | |
3974 | A C expression for the size in bytes of the trampoline, as an integer. | |
3975 | @end defmac | |
3976 | ||
3977 | @defmac TRAMPOLINE_ALIGNMENT | |
3978 | Alignment required for trampolines, in bits. | |
3979 | ||
3980 | If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT} | |
3981 | is used for aligning trampolines. | |
3982 | @end defmac | |
3983 | ||
3984 | @hook TARGET_TRAMPOLINE_INIT | |
3985 | ||
3986 | @hook TARGET_EMIT_CALL_BUILTIN___CLEAR_CACHE | |
3987 | ||
3988 | @hook TARGET_TRAMPOLINE_ADJUST_ADDRESS | |
3989 | ||
3990 | Implementing trampolines is difficult on many machines because they have | |
3991 | separate instruction and data caches. Writing into a stack location | |
3992 | fails to clear the memory in the instruction cache, so when the program | |
3993 | jumps to that location, it executes the old contents. | |
3994 | ||
3995 | Here are two possible solutions. One is to clear the relevant parts of | |
3996 | the instruction cache whenever a trampoline is set up. The other is to | |
3997 | make all trampolines identical, by having them jump to a standard | |
3998 | subroutine. The former technique makes trampoline execution faster; the | |
3999 | latter makes initialization faster. | |
4000 | ||
4001 | To clear the instruction cache when a trampoline is initialized, define | |
4002 | the following macro. | |
4003 | ||
4004 | @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end}) | |
4005 | If defined, expands to a C expression clearing the @emph{instruction | |
4006 | cache} in the specified interval. The definition of this macro would | |
4007 | typically be a series of @code{asm} statements. Both @var{beg} and | |
4008 | @var{end} are pointer expressions. | |
4009 | @end defmac | |
4010 | ||
4011 | To use a standard subroutine, define the following macro. In addition, | |
4012 | you must make sure that the instructions in a trampoline fill an entire | |
4013 | cache line with identical instructions, or else ensure that the | |
4014 | beginning of the trampoline code is always aligned at the same point in | |
4015 | its cache line. Look in @file{m68k.h} as a guide. | |
4016 | ||
4017 | @defmac TRANSFER_FROM_TRAMPOLINE | |
4018 | Define this macro if trampolines need a special subroutine to do their | |
4019 | work. The macro should expand to a series of @code{asm} statements | |
4020 | which will be compiled with GCC@. They go in a library function named | |
4021 | @code{__transfer_from_trampoline}. | |
4022 | ||
4023 | If you need to avoid executing the ordinary prologue code of a compiled | |
4024 | C function when you jump to the subroutine, you can do so by placing a | |
4025 | special label of your own in the assembler code. Use one @code{asm} | |
4026 | statement to generate an assembler label, and another to make the label | |
4027 | global. Then trampolines can use that label to jump directly to your | |
4028 | special assembler code. | |
4029 | @end defmac | |
4030 | ||
4031 | @node Library Calls | |
4032 | @section Implicit Calls to Library Routines | |
4033 | @cindex library subroutine names | |
4034 | @cindex @file{libgcc.a} | |
4035 | ||
4036 | @c prevent bad page break with this line | |
4037 | Here is an explanation of implicit calls to library routines. | |
4038 | ||
4039 | @defmac DECLARE_LIBRARY_RENAMES | |
4040 | This macro, if defined, should expand to a piece of C code that will get | |
4041 | expanded when compiling functions for libgcc.a. It can be used to | |
4042 | provide alternate names for GCC's internal library functions if there | |
4043 | are ABI-mandated names that the compiler should provide. | |
4044 | @end defmac | |
4045 | ||
4046 | @findex set_optab_libfunc | |
4047 | @findex init_one_libfunc | |
4048 | @hook TARGET_INIT_LIBFUNCS | |
4049 | ||
4050 | @hook TARGET_LIBFUNC_GNU_PREFIX | |
4051 | ||
4052 | @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison}) | |
4053 | This macro should return @code{true} if the library routine that | |
4054 | implements the floating point comparison operator @var{comparison} in | |
4055 | mode @var{mode} will return a boolean, and @var{false} if it will | |
4056 | return a tristate. | |
4057 | ||
4058 | GCC's own floating point libraries return tristates from the | |
4059 | comparison operators, so the default returns false always. Most ports | |
4060 | don't need to define this macro. | |
4061 | @end defmac | |
4062 | ||
4063 | @defmac TARGET_LIB_INT_CMP_BIASED | |
4064 | This macro should evaluate to @code{true} if the integer comparison | |
4065 | functions (like @code{__cmpdi2}) return 0 to indicate that the first | |
4066 | operand is smaller than the second, 1 to indicate that they are equal, | |
4067 | and 2 to indicate that the first operand is greater than the second. | |
4068 | If this macro evaluates to @code{false} the comparison functions return | |
4069 | @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines | |
4070 | in @file{libgcc.a}, you do not need to define this macro. | |
4071 | @end defmac | |
4072 | ||
4073 | @defmac TARGET_HAS_NO_HW_DIVIDE | |
4074 | This macro should be defined if the target has no hardware divide | |
4075 | instructions. If this macro is defined, GCC will use an algorithm which | |
4076 | make use of simple logical and arithmetic operations for 64-bit | |
4077 | division. If the macro is not defined, GCC will use an algorithm which | |
4078 | make use of a 64-bit by 32-bit divide primitive. | |
4079 | @end defmac | |
4080 | ||
4081 | @cindex @code{EDOM}, implicit usage | |
4082 | @findex matherr | |
4083 | @defmac TARGET_EDOM | |
4084 | The value of @code{EDOM} on the target machine, as a C integer constant | |
4085 | expression. If you don't define this macro, GCC does not attempt to | |
4086 | deposit the value of @code{EDOM} into @code{errno} directly. Look in | |
4087 | @file{/usr/include/errno.h} to find the value of @code{EDOM} on your | |
4088 | system. | |
4089 | ||
4090 | If you do not define @code{TARGET_EDOM}, then compiled code reports | |
4091 | domain errors by calling the library function and letting it report the | |
4092 | error. If mathematical functions on your system use @code{matherr} when | |
4093 | there is an error, then you should leave @code{TARGET_EDOM} undefined so | |
4094 | that @code{matherr} is used normally. | |
4095 | @end defmac | |
4096 | ||
4097 | @cindex @code{errno}, implicit usage | |
4098 | @defmac GEN_ERRNO_RTX | |
4099 | Define this macro as a C expression to create an rtl expression that | |
4100 | refers to the global ``variable'' @code{errno}. (On certain systems, | |
4101 | @code{errno} may not actually be a variable.) If you don't define this | |
4102 | macro, a reasonable default is used. | |
4103 | @end defmac | |
4104 | ||
4105 | @hook TARGET_LIBC_HAS_FUNCTION | |
4106 | ||
4107 | @hook TARGET_LIBC_HAS_FAST_FUNCTION | |
4108 | ||
24592abd MP |
4109 | @hook TARGET_FORTIFY_SOURCE_DEFAULT_LEVEL |
4110 | ||
bd539c9b JJ |
4111 | @hook TARGET_LIBM_FUNCTION_MAX_ERROR |
4112 | ||
d77de738 ML |
4113 | @defmac NEXT_OBJC_RUNTIME |
4114 | Set this macro to 1 to use the "NeXT" Objective-C message sending conventions | |
4115 | by default. This calling convention involves passing the object, the selector | |
4116 | and the method arguments all at once to the method-lookup library function. | |
a335cf24 | 4117 | This is the usual setting when targeting Darwin / macOS systems, which have |
d77de738 ML |
4118 | the NeXT runtime installed. |
4119 | ||
4120 | If the macro is set to 0, the "GNU" Objective-C message sending convention | |
4121 | will be used by default. This convention passes just the object and the | |
4122 | selector to the method-lookup function, which returns a pointer to the method. | |
4123 | ||
4124 | In either case, it remains possible to select code-generation for the alternate | |
4125 | scheme, by means of compiler command line switches. | |
4126 | @end defmac | |
4127 | ||
4128 | @node Addressing Modes | |
4129 | @section Addressing Modes | |
4130 | @cindex addressing modes | |
4131 | ||
4132 | @c prevent bad page break with this line | |
4133 | This is about addressing modes. | |
4134 | ||
4135 | @defmac HAVE_PRE_INCREMENT | |
4136 | @defmacx HAVE_PRE_DECREMENT | |
4137 | @defmacx HAVE_POST_INCREMENT | |
4138 | @defmacx HAVE_POST_DECREMENT | |
4139 | A C expression that is nonzero if the machine supports pre-increment, | |
4140 | pre-decrement, post-increment, or post-decrement addressing respectively. | |
4141 | @end defmac | |
4142 | ||
4143 | @defmac HAVE_PRE_MODIFY_DISP | |
4144 | @defmacx HAVE_POST_MODIFY_DISP | |
4145 | A C expression that is nonzero if the machine supports pre- or | |
4146 | post-address side-effect generation involving constants other than | |
4147 | the size of the memory operand. | |
4148 | @end defmac | |
4149 | ||
4150 | @defmac HAVE_PRE_MODIFY_REG | |
4151 | @defmacx HAVE_POST_MODIFY_REG | |
4152 | A C expression that is nonzero if the machine supports pre- or | |
4153 | post-address side-effect generation involving a register displacement. | |
4154 | @end defmac | |
4155 | ||
4156 | @defmac CONSTANT_ADDRESS_P (@var{x}) | |
4157 | A C expression that is 1 if the RTX @var{x} is a constant which | |
4158 | is a valid address. On most machines the default definition of | |
4159 | @code{(CONSTANT_P (@var{x}) && GET_CODE (@var{x}) != CONST_DOUBLE)} | |
4160 | is acceptable, but a few machines are more restrictive as to which | |
4161 | constant addresses are supported. | |
4162 | @end defmac | |
4163 | ||
4164 | @defmac CONSTANT_P (@var{x}) | |
4165 | @code{CONSTANT_P}, which is defined by target-independent code, | |
4166 | accepts integer-values expressions whose values are not explicitly | |
4167 | known, such as @code{symbol_ref}, @code{label_ref}, and @code{high} | |
4168 | expressions and @code{const} arithmetic expressions, in addition to | |
4169 | @code{const_int} and @code{const_double} expressions. | |
4170 | @end defmac | |
4171 | ||
4172 | @defmac MAX_REGS_PER_ADDRESS | |
4173 | A number, the maximum number of registers that can appear in a valid | |
4174 | memory address. Note that it is up to you to specify a value equal to | |
4175 | the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever | |
4176 | accept. | |
4177 | @end defmac | |
4178 | ||
4179 | @hook TARGET_LEGITIMATE_ADDRESS_P | |
4180 | ||
4181 | @defmac TARGET_MEM_CONSTRAINT | |
4182 | A single character to be used instead of the default @code{'m'} | |
4183 | character for general memory addresses. This defines the constraint | |
4184 | letter which matches the memory addresses accepted by | |
4185 | @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to | |
4186 | support new address formats in your back end without changing the | |
4187 | semantics of the @code{'m'} constraint. This is necessary in order to | |
4188 | preserve functionality of inline assembly constructs using the | |
4189 | @code{'m'} constraint. | |
4190 | @end defmac | |
4191 | ||
4192 | @defmac FIND_BASE_TERM (@var{x}) | |
4193 | A C expression to determine the base term of address @var{x}, | |
4194 | or to provide a simplified version of @var{x} from which @file{alias.cc} | |
4195 | can easily find the base term. This macro is used in only two places: | |
4196 | @code{find_base_value} and @code{find_base_term} in @file{alias.cc}. | |
4197 | ||
4198 | It is always safe for this macro to not be defined. It exists so | |
4199 | that alias analysis can understand machine-dependent addresses. | |
4200 | ||
4201 | The typical use of this macro is to handle addresses containing | |
4202 | a label_ref or symbol_ref within an UNSPEC@. | |
4203 | @end defmac | |
4204 | ||
4205 | @hook TARGET_LEGITIMIZE_ADDRESS | |
4206 | ||
4207 | @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win}) | |
4208 | A C compound statement that attempts to replace @var{x}, which is an address | |
4209 | that needs reloading, with a valid memory address for an operand of mode | |
4210 | @var{mode}. @var{win} will be a C statement label elsewhere in the code. | |
4211 | It is not necessary to define this macro, but it might be useful for | |
4212 | performance reasons. | |
4213 | ||
4214 | For example, on the i386, it is sometimes possible to use a single | |
4215 | reload register instead of two by reloading a sum of two pseudo | |
4216 | registers into a register. On the other hand, for number of RISC | |
4217 | processors offsets are limited so that often an intermediate address | |
4218 | needs to be generated in order to address a stack slot. By defining | |
4219 | @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses | |
4220 | generated for adjacent some stack slots can be made identical, and thus | |
4221 | be shared. | |
4222 | ||
4223 | @emph{Note}: This macro should be used with caution. It is necessary | |
4224 | to know something of how reload works in order to effectively use this, | |
4225 | and it is quite easy to produce macros that build in too much knowledge | |
4226 | of reload internals. | |
4227 | ||
4228 | @emph{Note}: This macro must be able to reload an address created by a | |
4229 | previous invocation of this macro. If it fails to handle such addresses | |
4230 | then the compiler may generate incorrect code or abort. | |
4231 | ||
4232 | @findex push_reload | |
4233 | The macro definition should use @code{push_reload} to indicate parts that | |
4234 | need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually | |
4235 | suitable to be passed unaltered to @code{push_reload}. | |
4236 | ||
4237 | The code generated by this macro must not alter the substructure of | |
4238 | @var{x}. If it transforms @var{x} into a more legitimate form, it | |
4239 | should assign @var{x} (which will always be a C variable) a new value. | |
4240 | This also applies to parts that you change indirectly by calling | |
4241 | @code{push_reload}. | |
4242 | ||
4243 | @findex strict_memory_address_p | |
4244 | The macro definition may use @code{strict_memory_address_p} to test if | |
4245 | the address has become legitimate. | |
4246 | ||
4247 | @findex copy_rtx | |
4248 | If you want to change only a part of @var{x}, one standard way of doing | |
4249 | this is to use @code{copy_rtx}. Note, however, that it unshares only a | |
4250 | single level of rtl. Thus, if the part to be changed is not at the | |
4251 | top level, you'll need to replace first the top level. | |
4252 | It is not necessary for this macro to come up with a legitimate | |
4253 | address; but often a machine-dependent strategy can generate better code. | |
4254 | @end defmac | |
4255 | ||
4256 | @hook TARGET_MODE_DEPENDENT_ADDRESS_P | |
4257 | ||
4258 | @hook TARGET_LEGITIMATE_CONSTANT_P | |
4259 | ||
4260 | @hook TARGET_PRECOMPUTE_TLS_P | |
4261 | ||
4262 | @hook TARGET_DELEGITIMIZE_ADDRESS | |
4263 | ||
4264 | @hook TARGET_CONST_NOT_OK_FOR_DEBUG_P | |
4265 | ||
4266 | @hook TARGET_CANNOT_FORCE_CONST_MEM | |
4267 | ||
4268 | @hook TARGET_USE_BLOCKS_FOR_CONSTANT_P | |
4269 | ||
4270 | @hook TARGET_USE_BLOCKS_FOR_DECL_P | |
4271 | ||
4272 | @hook TARGET_BUILTIN_RECIPROCAL | |
4273 | ||
4274 | @hook TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD | |
4275 | ||
4276 | @hook TARGET_VECTORIZE_BUILTIN_VECTORIZATION_COST | |
4277 | ||
4278 | @hook TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT | |
4279 | ||
4280 | @hook TARGET_VECTORIZE_VECTOR_ALIGNMENT_REACHABLE | |
4281 | ||
4282 | @hook TARGET_VECTORIZE_VEC_PERM_CONST | |
4283 | ||
81fd62d1 | 4284 | @hook TARGET_VECTORIZE_PREFERRED_DIV_AS_SHIFTS_OVER_MULT |
8beff04a | 4285 | |
d77de738 ML |
4286 | @hook TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION |
4287 | ||
4288 | @hook TARGET_VECTORIZE_BUILTIN_MD_VECTORIZED_FUNCTION | |
4289 | ||
4290 | @hook TARGET_VECTORIZE_SUPPORT_VECTOR_MISALIGNMENT | |
4291 | ||
4292 | @hook TARGET_VECTORIZE_PREFERRED_SIMD_MODE | |
4293 | ||
4294 | @hook TARGET_VECTORIZE_SPLIT_REDUCTION | |
4295 | ||
4296 | @hook TARGET_VECTORIZE_AUTOVECTORIZE_VECTOR_MODES | |
4297 | ||
4298 | @hook TARGET_VECTORIZE_RELATED_MODE | |
4299 | ||
4300 | @hook TARGET_VECTORIZE_GET_MASK_MODE | |
4301 | ||
af792f02 TC |
4302 | @hook TARGET_VECTORIZE_CONDITIONAL_OPERATION_IS_EXPENSIVE |
4303 | ||
d77de738 ML |
4304 | @hook TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE |
4305 | ||
4306 | @hook TARGET_VECTORIZE_CREATE_COSTS | |
4307 | ||
4308 | @hook TARGET_VECTORIZE_BUILTIN_GATHER | |
4309 | ||
4310 | @hook TARGET_VECTORIZE_BUILTIN_SCATTER | |
4311 | ||
4312 | @hook TARGET_SIMD_CLONE_COMPUTE_VECSIZE_AND_SIMDLEN | |
4313 | ||
4314 | @hook TARGET_SIMD_CLONE_ADJUST | |
4315 | ||
4316 | @hook TARGET_SIMD_CLONE_USABLE | |
4317 | ||
4318 | @hook TARGET_SIMT_VF | |
4319 | ||
4320 | @hook TARGET_OMP_DEVICE_KIND_ARCH_ISA | |
4321 | ||
4322 | @hook TARGET_GOACC_VALIDATE_DIMS | |
4323 | ||
4324 | @hook TARGET_GOACC_DIM_LIMIT | |
4325 | ||
4326 | @hook TARGET_GOACC_FORK_JOIN | |
4327 | ||
4328 | @hook TARGET_GOACC_REDUCTION | |
4329 | ||
4330 | @hook TARGET_PREFERRED_ELSE_VALUE | |
4331 | ||
4332 | @hook TARGET_GOACC_ADJUST_PRIVATE_DECL | |
4333 | ||
4334 | @hook TARGET_GOACC_EXPAND_VAR_DECL | |
4335 | ||
4336 | @hook TARGET_GOACC_CREATE_WORKER_BROADCAST_RECORD | |
4337 | ||
4338 | @hook TARGET_GOACC_SHARED_MEM_LAYOUT | |
4339 | ||
4340 | @node Anchored Addresses | |
4341 | @section Anchored Addresses | |
4342 | @cindex anchored addresses | |
4343 | @cindex @option{-fsection-anchors} | |
4344 | ||
4345 | GCC usually addresses every static object as a separate entity. | |
4346 | For example, if we have: | |
4347 | ||
4348 | @smallexample | |
4349 | static int a, b, c; | |
4350 | int foo (void) @{ return a + b + c; @} | |
4351 | @end smallexample | |
4352 | ||
4353 | the code for @code{foo} will usually calculate three separate symbolic | |
4354 | addresses: those of @code{a}, @code{b} and @code{c}. On some targets, | |
4355 | it would be better to calculate just one symbolic address and access | |
4356 | the three variables relative to it. The equivalent pseudocode would | |
4357 | be something like: | |
4358 | ||
4359 | @smallexample | |
4360 | int foo (void) | |
4361 | @{ | |
4362 | register int *xr = &x; | |
4363 | return xr[&a - &x] + xr[&b - &x] + xr[&c - &x]; | |
4364 | @} | |
4365 | @end smallexample | |
4366 | ||
4367 | (which isn't valid C). We refer to shared addresses like @code{x} as | |
4368 | ``section anchors''. Their use is controlled by @option{-fsection-anchors}. | |
4369 | ||
4370 | The hooks below describe the target properties that GCC needs to know | |
4371 | in order to make effective use of section anchors. It won't use | |
4372 | section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET} | |
4373 | or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value. | |
4374 | ||
4375 | @hook TARGET_MIN_ANCHOR_OFFSET | |
4376 | ||
4377 | @hook TARGET_MAX_ANCHOR_OFFSET | |
4378 | ||
4379 | @hook TARGET_ASM_OUTPUT_ANCHOR | |
4380 | ||
4381 | @hook TARGET_USE_ANCHORS_FOR_SYMBOL_P | |
4382 | ||
4383 | @node Condition Code | |
4384 | @section Condition Code Status | |
4385 | @cindex condition code status | |
4386 | ||
4387 | Condition codes in GCC are represented as registers, | |
4388 | which provides better schedulability for | |
4389 | architectures that do have a condition code register, but on which | |
4390 | most instructions do not affect it. The latter category includes | |
4391 | most RISC machines. | |
4392 | ||
4393 | Implicit clobbering would pose a strong restriction on the placement of | |
4394 | the definition and use of the condition code. In the past the definition | |
4395 | and use were always adjacent. However, recent changes to support trapping | |
4396 | arithmetic may result in the definition and user being in different blocks. | |
4397 | Thus, there may be a @code{NOTE_INSN_BASIC_BLOCK} between them. Additionally, | |
4398 | the definition may be the source of exception handling edges. | |
4399 | ||
4400 | These restrictions can prevent important | |
4401 | optimizations on some machines. For example, on the IBM RS/6000, there | |
4402 | is a delay for taken branches unless the condition code register is set | |
4403 | three instructions earlier than the conditional branch. The instruction | |
4404 | scheduler cannot perform this optimization if it is not permitted to | |
4405 | separate the definition and use of the condition code register. | |
4406 | ||
4407 | If there is a specific | |
4408 | condition code register in the machine, use a hard register. If the | |
4409 | condition code or comparison result can be placed in any general register, | |
4410 | or if there are multiple condition registers, use a pseudo register. | |
4411 | Registers used to store the condition code value will usually have a mode | |
4412 | that is in class @code{MODE_CC}. | |
4413 | ||
4414 | Alternatively, you can use @code{BImode} if the comparison operator is | |
4415 | specified already in the compare instruction. In this case, you are not | |
4416 | interested in most macros in this section. | |
4417 | ||
4418 | @menu | |
4419 | * MODE_CC Condition Codes:: Modern representation of condition codes. | |
4420 | @end menu | |
4421 | ||
4422 | @node MODE_CC Condition Codes | |
4423 | @subsection Representation of condition codes using registers | |
4424 | @findex CCmode | |
4425 | @findex MODE_CC | |
4426 | ||
4427 | @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y}) | |
4428 | On many machines, the condition code may be produced by other instructions | |
4429 | than compares, for example the branch can use directly the condition | |
4430 | code set by a subtract instruction. However, on some machines | |
4431 | when the condition code is set this way some bits (such as the overflow | |
4432 | bit) are not set in the same way as a test instruction, so that a different | |
4433 | branch instruction must be used for some conditional branches. When | |
4434 | this happens, use the machine mode of the condition code register to | |
4435 | record different formats of the condition code register. Modes can | |
4436 | also be used to record which compare instruction (e.g.@: a signed or an | |
4437 | unsigned comparison) produced the condition codes. | |
4438 | ||
4439 | If other modes than @code{CCmode} are required, add them to | |
4440 | @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose | |
4441 | a mode given an operand of a compare. This is needed because the modes | |
4442 | have to be chosen not only during RTL generation but also, for example, | |
4443 | by instruction combination. The result of @code{SELECT_CC_MODE} should | |
4444 | be consistent with the mode used in the patterns; for example to support | |
4445 | the case of the add on the SPARC discussed above, we have the pattern | |
4446 | ||
4447 | @smallexample | |
4448 | (define_insn "" | |
4449 | [(set (reg:CCNZ 0) | |
4450 | (compare:CCNZ | |
4451 | (plus:SI (match_operand:SI 0 "register_operand" "%r") | |
4452 | (match_operand:SI 1 "arith_operand" "rI")) | |
4453 | (const_int 0)))] | |
4454 | "" | |
4455 | "@dots{}") | |
4456 | @end smallexample | |
4457 | ||
4458 | @noindent | |
4459 | together with a @code{SELECT_CC_MODE} that returns @code{CCNZmode} | |
4460 | for comparisons whose argument is a @code{plus}: | |
4461 | ||
4462 | @smallexample | |
4463 | #define SELECT_CC_MODE(OP,X,Y) \ | |
4464 | (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \ | |
4465 | ? ((OP == LT || OP == LE || OP == GT || OP == GE) \ | |
4466 | ? CCFPEmode : CCFPmode) \ | |
4467 | : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \ | |
4468 | || GET_CODE (X) == NEG || GET_CODE (x) == ASHIFT) \ | |
4469 | ? CCNZmode : CCmode)) | |
4470 | @end smallexample | |
4471 | ||
4472 | Another reason to use modes is to retain information on which operands | |
4473 | were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in | |
4474 | this section. | |
4475 | ||
4476 | You should define this macro if and only if you define extra CC modes | |
4477 | in @file{@var{machine}-modes.def}. | |
4478 | @end defmac | |
4479 | ||
4480 | @hook TARGET_CANONICALIZE_COMPARISON | |
4481 | ||
4482 | @defmac REVERSIBLE_CC_MODE (@var{mode}) | |
4483 | A C expression whose value is one if it is always safe to reverse a | |
4484 | comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE} | |
4485 | can ever return @var{mode} for a floating-point inequality comparison, | |
4486 | then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero. | |
4487 | ||
4488 | You need not define this macro if it would always returns zero or if the | |
4489 | floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}. | |
4490 | For example, here is the definition used on the SPARC, where floating-point | |
4491 | inequality comparisons are given either @code{CCFPEmode} or @code{CCFPmode}: | |
4492 | ||
4493 | @smallexample | |
4494 | #define REVERSIBLE_CC_MODE(MODE) \ | |
4495 | ((MODE) != CCFPEmode && (MODE) != CCFPmode) | |
4496 | @end smallexample | |
4497 | @end defmac | |
4498 | ||
4499 | @defmac REVERSE_CONDITION (@var{code}, @var{mode}) | |
4500 | A C expression whose value is reversed condition code of the @var{code} for | |
4501 | comparison done in CC_MODE @var{mode}. The macro is used only in case | |
4502 | @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case | |
4503 | machine has some non-standard way how to reverse certain conditionals. For | |
4504 | instance in case all floating point conditions are non-trapping, compiler may | |
4505 | freely convert unordered compares to ordered ones. Then definition may look | |
4506 | like: | |
4507 | ||
4508 | @smallexample | |
4509 | #define REVERSE_CONDITION(CODE, MODE) \ | |
4510 | ((MODE) != CCFPmode ? reverse_condition (CODE) \ | |
4511 | : reverse_condition_maybe_unordered (CODE)) | |
4512 | @end smallexample | |
4513 | @end defmac | |
4514 | ||
4515 | @hook TARGET_FIXED_CONDITION_CODE_REGS | |
4516 | ||
4517 | @hook TARGET_CC_MODES_COMPATIBLE | |
4518 | ||
4519 | @hook TARGET_FLAGS_REGNUM | |
4520 | ||
4521 | @node Costs | |
4522 | @section Describing Relative Costs of Operations | |
4523 | @cindex costs of instructions | |
4524 | @cindex relative costs | |
4525 | @cindex speed of instructions | |
4526 | ||
4527 | These macros let you describe the relative speed of various operations | |
4528 | on the target machine. | |
4529 | ||
4530 | @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to}) | |
4531 | A C expression for the cost of moving data of mode @var{mode} from a | |
4532 | register in class @var{from} to one in class @var{to}. The classes are | |
4533 | expressed using the enumeration values such as @code{GENERAL_REGS}. A | |
4534 | value of 2 is the default; other values are interpreted relative to | |
4535 | that. | |
4536 | ||
4537 | It is not required that the cost always equal 2 when @var{from} is the | |
4538 | same as @var{to}; on some machines it is expensive to move between | |
4539 | registers if they are not general registers. | |
4540 | ||
4541 | If reload sees an insn consisting of a single @code{set} between two | |
4542 | hard registers, and if @code{REGISTER_MOVE_COST} applied to their | |
4543 | classes returns a value of 2, reload does not check to ensure that the | |
4544 | constraints of the insn are met. Setting a cost of other than 2 will | |
4545 | allow reload to verify that the constraints are met. You should do this | |
4546 | if the @samp{mov@var{m}} pattern's constraints do not allow such copying. | |
4547 | ||
4548 | These macros are obsolete, new ports should use the target hook | |
4549 | @code{TARGET_REGISTER_MOVE_COST} instead. | |
4550 | @end defmac | |
4551 | ||
4552 | @hook TARGET_REGISTER_MOVE_COST | |
4553 | ||
4554 | @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in}) | |
4555 | A C expression for the cost of moving data of mode @var{mode} between a | |
4556 | register of class @var{class} and memory; @var{in} is zero if the value | |
4557 | is to be written to memory, nonzero if it is to be read in. This cost | |
4558 | is relative to those in @code{REGISTER_MOVE_COST}. If moving between | |
4559 | registers and memory is more expensive than between two registers, you | |
4560 | should define this macro to express the relative cost. | |
4561 | ||
4562 | If you do not define this macro, GCC uses a default cost of 4 plus | |
4563 | the cost of copying via a secondary reload register, if one is | |
4564 | needed. If your machine requires a secondary reload register to copy | |
4565 | between memory and a register of @var{class} but the reload mechanism is | |
4566 | more complex than copying via an intermediate, define this macro to | |
4567 | reflect the actual cost of the move. | |
4568 | ||
4569 | GCC defines the function @code{memory_move_secondary_cost} if | |
4570 | secondary reloads are needed. It computes the costs due to copying via | |
4571 | a secondary register. If your machine copies from memory using a | |
4572 | secondary register in the conventional way but the default base value of | |
4573 | 4 is not correct for your machine, define this macro to add some other | |
4574 | value to the result of that function. The arguments to that function | |
4575 | are the same as to this macro. | |
4576 | ||
4577 | These macros are obsolete, new ports should use the target hook | |
4578 | @code{TARGET_MEMORY_MOVE_COST} instead. | |
4579 | @end defmac | |
4580 | ||
4581 | @hook TARGET_MEMORY_MOVE_COST | |
4582 | ||
4583 | @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p}) | |
4584 | A C expression for the cost of a branch instruction. A value of 1 is | |
4585 | the default; other values are interpreted relative to that. Parameter | |
4586 | @var{speed_p} is true when the branch in question should be optimized | |
4587 | for speed. When it is false, @code{BRANCH_COST} should return a value | |
4588 | optimal for code size rather than performance. @var{predictable_p} is | |
4589 | true for well-predicted branches. On many architectures the | |
4590 | @code{BRANCH_COST} can be reduced then. | |
4591 | @end defmac | |
4592 | ||
4593 | Here are additional macros which do not specify precise relative costs, | |
4594 | but only that certain actions are more expensive than GCC would | |
4595 | ordinarily expect. | |
4596 | ||
4597 | @defmac SLOW_BYTE_ACCESS | |
4598 | Define this macro as a C expression which is nonzero if accessing less | |
4599 | than a word of memory (i.e.@: a @code{char} or a @code{short}) is no | |
4600 | faster than accessing a word of memory, i.e., if such access | |
4601 | require more than one instruction or if there is no difference in cost | |
4602 | between byte and (aligned) word loads. | |
4603 | ||
4604 | When this macro is not defined, the compiler will access a field by | |
4605 | finding the smallest containing object; when it is defined, a fullword | |
4606 | load will be used if alignment permits. Unless bytes accesses are | |
4607 | faster than word accesses, using word accesses is preferable since it | |
4608 | may eliminate subsequent memory access if subsequent accesses occur to | |
4609 | other fields in the same word of the structure, but to different bytes. | |
4610 | @end defmac | |
4611 | ||
4612 | @hook TARGET_SLOW_UNALIGNED_ACCESS | |
4613 | ||
4614 | @defmac MOVE_RATIO (@var{speed}) | |
4615 | The threshold of number of scalar memory-to-memory move insns, @emph{below} | |
4616 | which a sequence of insns should be generated instead of a | |
4617 | string move insn or a library call. Increasing the value will always | |
4618 | make code faster, but eventually incurs high cost in increased code size. | |
4619 | ||
4620 | Note that on machines where the corresponding move insn is a | |
4621 | @code{define_expand} that emits a sequence of insns, this macro counts | |
4622 | the number of such sequences. | |
4623 | ||
4624 | The parameter @var{speed} is true if the code is currently being | |
4625 | optimized for speed rather than size. | |
4626 | ||
4627 | If you don't define this, a reasonable default is used. | |
4628 | @end defmac | |
4629 | ||
4630 | @hook TARGET_USE_BY_PIECES_INFRASTRUCTURE_P | |
4631 | ||
4632 | @hook TARGET_OVERLAP_OP_BY_PIECES_P | |
4633 | ||
4634 | @hook TARGET_COMPARE_BY_PIECES_BRANCH_RATIO | |
4635 | ||
4636 | @defmac MOVE_MAX_PIECES | |
4637 | A C expression used by @code{move_by_pieces} to determine the largest unit | |
4638 | a load or store used to copy memory is. Defaults to @code{MOVE_MAX}. | |
4639 | @end defmac | |
4640 | ||
4641 | @defmac STORE_MAX_PIECES | |
4642 | A C expression used by @code{store_by_pieces} to determine the largest unit | |
4643 | a store used to memory is. Defaults to @code{MOVE_MAX_PIECES}, or two times | |
4644 | the size of @code{HOST_WIDE_INT}, whichever is smaller. | |
4645 | @end defmac | |
4646 | ||
4647 | @defmac COMPARE_MAX_PIECES | |
4648 | A C expression used by @code{compare_by_pieces} to determine the largest unit | |
4649 | a load or store used to compare memory is. Defaults to | |
4650 | @code{MOVE_MAX_PIECES}. | |
4651 | @end defmac | |
4652 | ||
4653 | @defmac CLEAR_RATIO (@var{speed}) | |
4654 | The threshold of number of scalar move insns, @emph{below} which a sequence | |
4655 | of insns should be generated to clear memory instead of a string clear insn | |
4656 | or a library call. Increasing the value will always make code faster, but | |
4657 | eventually incurs high cost in increased code size. | |
4658 | ||
4659 | The parameter @var{speed} is true if the code is currently being | |
4660 | optimized for speed rather than size. | |
4661 | ||
4662 | If you don't define this, a reasonable default is used. | |
4663 | @end defmac | |
4664 | ||
4665 | @defmac SET_RATIO (@var{speed}) | |
4666 | The threshold of number of scalar move insns, @emph{below} which a sequence | |
4667 | of insns should be generated to set memory to a constant value, instead of | |
4668 | a block set insn or a library call. | |
4669 | Increasing the value will always make code faster, but | |
4670 | eventually incurs high cost in increased code size. | |
4671 | ||
4672 | The parameter @var{speed} is true if the code is currently being | |
4673 | optimized for speed rather than size. | |
4674 | ||
4675 | If you don't define this, it defaults to the value of @code{MOVE_RATIO}. | |
4676 | @end defmac | |
4677 | ||
4678 | @defmac USE_LOAD_POST_INCREMENT (@var{mode}) | |
4679 | A C expression used to determine whether a load postincrement is a good | |
4680 | thing to use for a given mode. Defaults to the value of | |
4681 | @code{HAVE_POST_INCREMENT}. | |
4682 | @end defmac | |
4683 | ||
4684 | @defmac USE_LOAD_POST_DECREMENT (@var{mode}) | |
4685 | A C expression used to determine whether a load postdecrement is a good | |
4686 | thing to use for a given mode. Defaults to the value of | |
4687 | @code{HAVE_POST_DECREMENT}. | |
4688 | @end defmac | |
4689 | ||
4690 | @defmac USE_LOAD_PRE_INCREMENT (@var{mode}) | |
4691 | A C expression used to determine whether a load preincrement is a good | |
4692 | thing to use for a given mode. Defaults to the value of | |
4693 | @code{HAVE_PRE_INCREMENT}. | |
4694 | @end defmac | |
4695 | ||
4696 | @defmac USE_LOAD_PRE_DECREMENT (@var{mode}) | |
4697 | A C expression used to determine whether a load predecrement is a good | |
4698 | thing to use for a given mode. Defaults to the value of | |
4699 | @code{HAVE_PRE_DECREMENT}. | |
4700 | @end defmac | |
4701 | ||
4702 | @defmac USE_STORE_POST_INCREMENT (@var{mode}) | |
4703 | A C expression used to determine whether a store postincrement is a good | |
4704 | thing to use for a given mode. Defaults to the value of | |
4705 | @code{HAVE_POST_INCREMENT}. | |
4706 | @end defmac | |
4707 | ||
4708 | @defmac USE_STORE_POST_DECREMENT (@var{mode}) | |
4709 | A C expression used to determine whether a store postdecrement is a good | |
4710 | thing to use for a given mode. Defaults to the value of | |
4711 | @code{HAVE_POST_DECREMENT}. | |
4712 | @end defmac | |
4713 | ||
4714 | @defmac USE_STORE_PRE_INCREMENT (@var{mode}) | |
4715 | This macro is used to determine whether a store preincrement is a good | |
4716 | thing to use for a given mode. Defaults to the value of | |
4717 | @code{HAVE_PRE_INCREMENT}. | |
4718 | @end defmac | |
4719 | ||
4720 | @defmac USE_STORE_PRE_DECREMENT (@var{mode}) | |
4721 | This macro is used to determine whether a store predecrement is a good | |
4722 | thing to use for a given mode. Defaults to the value of | |
4723 | @code{HAVE_PRE_DECREMENT}. | |
4724 | @end defmac | |
4725 | ||
4726 | @defmac NO_FUNCTION_CSE | |
4727 | Define this macro to be true if it is as good or better to call a constant | |
4728 | function address than to call an address kept in a register. | |
4729 | @end defmac | |
4730 | ||
4731 | @defmac LOGICAL_OP_NON_SHORT_CIRCUIT | |
4732 | Define this macro if a non-short-circuit operation produced by | |
4733 | @samp{fold_range_test ()} is optimal. This macro defaults to true if | |
4734 | @code{BRANCH_COST} is greater than or equal to the value 2. | |
4735 | @end defmac | |
4736 | ||
4737 | @hook TARGET_OPTAB_SUPPORTED_P | |
4738 | ||
4739 | @hook TARGET_RTX_COSTS | |
4740 | ||
4741 | @hook TARGET_ADDRESS_COST | |
4742 | ||
4743 | @hook TARGET_INSN_COST | |
4744 | ||
4745 | @hook TARGET_MAX_NOCE_IFCVT_SEQ_COST | |
4746 | ||
4747 | @hook TARGET_NOCE_CONVERSION_PROFITABLE_P | |
4748 | ||
4749 | @hook TARGET_NEW_ADDRESS_PROFITABLE_P | |
4750 | ||
4751 | @hook TARGET_NO_SPECULATION_IN_DELAY_SLOTS_P | |
4752 | ||
4753 | @hook TARGET_ESTIMATED_POLY_VALUE | |
4754 | ||
4755 | @node Scheduling | |
4756 | @section Adjusting the Instruction Scheduler | |
4757 | ||
4758 | The instruction scheduler may need a fair amount of machine-specific | |
4759 | adjustment in order to produce good code. GCC provides several target | |
4760 | hooks for this purpose. It is usually enough to define just a few of | |
4761 | them: try the first ones in this list first. | |
4762 | ||
4763 | @hook TARGET_SCHED_ISSUE_RATE | |
4764 | ||
4765 | @hook TARGET_SCHED_VARIABLE_ISSUE | |
4766 | ||
4767 | @hook TARGET_SCHED_ADJUST_COST | |
4768 | ||
4769 | @hook TARGET_SCHED_ADJUST_PRIORITY | |
4770 | ||
4771 | @hook TARGET_SCHED_REORDER | |
4772 | ||
4773 | @hook TARGET_SCHED_REORDER2 | |
4774 | ||
4775 | @hook TARGET_SCHED_MACRO_FUSION_P | |
4776 | ||
4777 | @hook TARGET_SCHED_MACRO_FUSION_PAIR_P | |
4778 | ||
4779 | @hook TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK | |
4780 | ||
4781 | @hook TARGET_SCHED_INIT | |
4782 | ||
4783 | @hook TARGET_SCHED_FINISH | |
4784 | ||
4785 | @hook TARGET_SCHED_INIT_GLOBAL | |
4786 | ||
4787 | @hook TARGET_SCHED_FINISH_GLOBAL | |
4788 | ||
4789 | @hook TARGET_SCHED_DFA_PRE_CYCLE_INSN | |
4790 | ||
4791 | @hook TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN | |
4792 | ||
4793 | @hook TARGET_SCHED_DFA_POST_CYCLE_INSN | |
4794 | ||
4795 | @hook TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN | |
4796 | ||
4797 | @hook TARGET_SCHED_DFA_PRE_ADVANCE_CYCLE | |
4798 | ||
4799 | @hook TARGET_SCHED_DFA_POST_ADVANCE_CYCLE | |
4800 | ||
4801 | @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD | |
4802 | ||
4803 | @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD | |
4804 | ||
4805 | @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BEGIN | |
4806 | ||
4807 | @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_ISSUE | |
4808 | ||
4809 | @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_BACKTRACK | |
4810 | ||
4811 | @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_END | |
4812 | ||
4813 | @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_INIT | |
4814 | ||
4815 | @hook TARGET_SCHED_FIRST_CYCLE_MULTIPASS_FINI | |
4816 | ||
4817 | @hook TARGET_SCHED_DFA_NEW_CYCLE | |
4818 | ||
4819 | @hook TARGET_SCHED_IS_COSTLY_DEPENDENCE | |
4820 | ||
4821 | @hook TARGET_SCHED_H_I_D_EXTENDED | |
4822 | ||
4823 | @hook TARGET_SCHED_ALLOC_SCHED_CONTEXT | |
4824 | ||
4825 | @hook TARGET_SCHED_INIT_SCHED_CONTEXT | |
4826 | ||
4827 | @hook TARGET_SCHED_SET_SCHED_CONTEXT | |
4828 | ||
4829 | @hook TARGET_SCHED_CLEAR_SCHED_CONTEXT | |
4830 | ||
4831 | @hook TARGET_SCHED_FREE_SCHED_CONTEXT | |
4832 | ||
4833 | @hook TARGET_SCHED_SPECULATE_INSN | |
4834 | ||
4835 | @hook TARGET_SCHED_NEEDS_BLOCK_P | |
4836 | ||
4837 | @hook TARGET_SCHED_GEN_SPEC_CHECK | |
4838 | ||
4839 | @hook TARGET_SCHED_SET_SCHED_FLAGS | |
4840 | ||
4841 | @hook TARGET_SCHED_CAN_SPECULATE_INSN | |
4842 | ||
4843 | @hook TARGET_SCHED_SMS_RES_MII | |
4844 | ||
4845 | @hook TARGET_SCHED_DISPATCH | |
4846 | ||
4847 | @hook TARGET_SCHED_DISPATCH_DO | |
4848 | ||
4849 | @hook TARGET_SCHED_EXPOSED_PIPELINE | |
4850 | ||
4851 | @hook TARGET_SCHED_REASSOCIATION_WIDTH | |
4852 | ||
4853 | @hook TARGET_SCHED_FUSION_PRIORITY | |
4854 | ||
4855 | @hook TARGET_EXPAND_DIVMOD_LIBFUNC | |
4856 | ||
4857 | @node Sections | |
4858 | @section Dividing the Output into Sections (Texts, Data, @dots{}) | |
4859 | @c the above section title is WAY too long. maybe cut the part between | |
4860 | @c the (...)? --mew 10feb93 | |
4861 | ||
4862 | An object file is divided into sections containing different types of | |
4863 | data. In the most common case, there are three sections: the @dfn{text | |
4864 | section}, which holds instructions and read-only data; the @dfn{data | |
4865 | section}, which holds initialized writable data; and the @dfn{bss | |
4866 | section}, which holds uninitialized data. Some systems have other kinds | |
4867 | of sections. | |
4868 | ||
4869 | @file{varasm.cc} provides several well-known sections, such as | |
4870 | @code{text_section}, @code{data_section} and @code{bss_section}. | |
4871 | The normal way of controlling a @code{@var{foo}_section} variable | |
4872 | is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro, | |
4873 | as described below. The macros are only read once, when @file{varasm.cc} | |
4874 | initializes itself, so their values must be run-time constants. | |
4875 | They may however depend on command-line flags. | |
4876 | ||
4877 | @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make | |
4878 | use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them | |
4879 | to be string literals. | |
4880 | ||
4881 | Some assemblers require a different string to be written every time a | |
4882 | section is selected. If your assembler falls into this category, you | |
4883 | should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use | |
4884 | @code{get_unnamed_section} to set up the sections. | |
4885 | ||
4886 | You must always create a @code{text_section}, either by defining | |
4887 | @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section} | |
4888 | in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of | |
4889 | @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not | |
4890 | create a distinct @code{readonly_data_section}, the default is to | |
4891 | reuse @code{text_section}. | |
4892 | ||
4893 | All the other @file{varasm.cc} sections are optional, and are null | |
4894 | if the target does not provide them. | |
4895 | ||
4896 | @defmac TEXT_SECTION_ASM_OP | |
4897 | A C expression whose value is a string, including spacing, containing the | |
4898 | assembler operation that should precede instructions and read-only data. | |
4899 | Normally @code{"\t.text"} is right. | |
4900 | @end defmac | |
4901 | ||
4902 | @defmac HOT_TEXT_SECTION_NAME | |
4903 | If defined, a C string constant for the name of the section containing most | |
4904 | frequently executed functions of the program. If not defined, GCC will provide | |
4905 | a default definition if the target supports named sections. | |
4906 | @end defmac | |
4907 | ||
4908 | @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME | |
4909 | If defined, a C string constant for the name of the section containing unlikely | |
4910 | executed functions in the program. | |
4911 | @end defmac | |
4912 | ||
4913 | @defmac DATA_SECTION_ASM_OP | |
4914 | A C expression whose value is a string, including spacing, containing the | |
4915 | assembler operation to identify the following data as writable initialized | |
4916 | data. Normally @code{"\t.data"} is right. | |
4917 | @end defmac | |
4918 | ||
4919 | @defmac SDATA_SECTION_ASM_OP | |
4920 | If defined, a C expression whose value is a string, including spacing, | |
4921 | containing the assembler operation to identify the following data as | |
4922 | initialized, writable small data. | |
4923 | @end defmac | |
4924 | ||
4925 | @defmac READONLY_DATA_SECTION_ASM_OP | |
4926 | A C expression whose value is a string, including spacing, containing the | |
4927 | assembler operation to identify the following data as read-only initialized | |
4928 | data. | |
4929 | @end defmac | |
4930 | ||
4931 | @defmac BSS_SECTION_ASM_OP | |
4932 | If defined, a C expression whose value is a string, including spacing, | |
4933 | containing the assembler operation to identify the following data as | |
4934 | uninitialized global data. If not defined, and | |
4935 | @code{ASM_OUTPUT_ALIGNED_BSS} not defined, | |
4936 | uninitialized global data will be output in the data section if | |
4937 | @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be | |
4938 | used. | |
4939 | @end defmac | |
4940 | ||
4941 | @defmac SBSS_SECTION_ASM_OP | |
4942 | If defined, a C expression whose value is a string, including spacing, | |
4943 | containing the assembler operation to identify the following data as | |
4944 | uninitialized, writable small data. | |
4945 | @end defmac | |
4946 | ||
4947 | @defmac TLS_COMMON_ASM_OP | |
4948 | If defined, a C expression whose value is a string containing the | |
4949 | assembler operation to identify the following data as thread-local | |
4950 | common data. The default is @code{".tls_common"}. | |
4951 | @end defmac | |
4952 | ||
4953 | @defmac TLS_SECTION_ASM_FLAG | |
4954 | If defined, a C expression whose value is a character constant | |
4955 | containing the flag used to mark a section as a TLS section. The | |
4956 | default is @code{'T'}. | |
4957 | @end defmac | |
4958 | ||
4959 | @defmac INIT_SECTION_ASM_OP | |
4960 | If defined, a C expression whose value is a string, including spacing, | |
4961 | containing the assembler operation to identify the following data as | |
4962 | initialization code. If not defined, GCC will assume such a section does | |
4963 | not exist. This section has no corresponding @code{init_section} | |
4964 | variable; it is used entirely in runtime code. | |
4965 | @end defmac | |
4966 | ||
4967 | @defmac FINI_SECTION_ASM_OP | |
4968 | If defined, a C expression whose value is a string, including spacing, | |
4969 | containing the assembler operation to identify the following data as | |
4970 | finalization code. If not defined, GCC will assume such a section does | |
4971 | not exist. This section has no corresponding @code{fini_section} | |
4972 | variable; it is used entirely in runtime code. | |
4973 | @end defmac | |
4974 | ||
4975 | @defmac INIT_ARRAY_SECTION_ASM_OP | |
4976 | If defined, a C expression whose value is a string, including spacing, | |
4977 | containing the assembler operation to identify the following data as | |
4978 | part of the @code{.init_array} (or equivalent) section. If not | |
4979 | defined, GCC will assume such a section does not exist. Do not define | |
4980 | both this macro and @code{INIT_SECTION_ASM_OP}. | |
4981 | @end defmac | |
4982 | ||
4983 | @defmac FINI_ARRAY_SECTION_ASM_OP | |
4984 | If defined, a C expression whose value is a string, including spacing, | |
4985 | containing the assembler operation to identify the following data as | |
4986 | part of the @code{.fini_array} (or equivalent) section. If not | |
4987 | defined, GCC will assume such a section does not exist. Do not define | |
4988 | both this macro and @code{FINI_SECTION_ASM_OP}. | |
4989 | @end defmac | |
4990 | ||
4991 | @defmac MACH_DEP_SECTION_ASM_FLAG | |
4992 | If defined, a C expression whose value is a character constant | |
4993 | containing the flag used to mark a machine-dependent section. This | |
4994 | corresponds to the @code{SECTION_MACH_DEP} section flag. | |
4995 | @end defmac | |
4996 | ||
4997 | @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function}) | |
4998 | If defined, an ASM statement that switches to a different section | |
4999 | via @var{section_op}, calls @var{function}, and switches back to | |
5000 | the text section. This is used in @file{crtstuff.c} if | |
5001 | @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls | |
5002 | to initialization and finalization functions from the init and fini | |
5003 | sections. By default, this macro uses a simple function call. Some | |
5004 | ports need hand-crafted assembly code to avoid dependencies on | |
5005 | registers initialized in the function prologue or to ensure that | |
5006 | constant pools don't end up too far way in the text section. | |
5007 | @end defmac | |
5008 | ||
5009 | @defmac TARGET_LIBGCC_SDATA_SECTION | |
5010 | If defined, a string which names the section into which small | |
5011 | variables defined in crtstuff and libgcc should go. This is useful | |
5012 | when the target has options for optimizing access to small data, and | |
5013 | you want the crtstuff and libgcc routines to be conservative in what | |
5014 | they expect of your application yet liberal in what your application | |
5015 | expects. For example, for targets with a @code{.sdata} section (like | |
5016 | MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't | |
5017 | require small data support from your application, but use this macro | |
5018 | to put small data into @code{.sdata} so that your application can | |
5019 | access these variables whether it uses small data or not. | |
5020 | @end defmac | |
5021 | ||
5022 | @defmac FORCE_CODE_SECTION_ALIGN | |
5023 | If defined, an ASM statement that aligns a code section to some | |
5024 | arbitrary boundary. This is used to force all fragments of the | |
5025 | @code{.init} and @code{.fini} sections to have to same alignment | |
5026 | and thus prevent the linker from having to add any padding. | |
5027 | @end defmac | |
5028 | ||
5029 | @defmac JUMP_TABLES_IN_TEXT_SECTION | |
5030 | Define this macro to be an expression with a nonzero value if jump | |
5031 | tables (for @code{tablejump} insns) should be output in the text | |
5032 | section, along with the assembler instructions. Otherwise, the | |
5033 | readonly data section is used. | |
5034 | ||
5035 | This macro is irrelevant if there is no separate readonly data section. | |
5036 | @end defmac | |
5037 | ||
5038 | @hook TARGET_ASM_INIT_SECTIONS | |
5039 | ||
5040 | @hook TARGET_ASM_RELOC_RW_MASK | |
5041 | ||
5042 | @hook TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC | |
5043 | ||
5044 | @hook TARGET_ASM_SELECT_SECTION | |
5045 | ||
5046 | @defmac USE_SELECT_SECTION_FOR_FUNCTIONS | |
5047 | Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called | |
5048 | for @code{FUNCTION_DECL}s as well as for variables and constants. | |
5049 | ||
5050 | In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the | |
5051 | function has been determined to be likely to be called, and nonzero if | |
5052 | it is unlikely to be called. | |
5053 | @end defmac | |
5054 | ||
5055 | @hook TARGET_ASM_UNIQUE_SECTION | |
5056 | ||
5057 | @hook TARGET_ASM_FUNCTION_RODATA_SECTION | |
5058 | ||
5059 | @hook TARGET_ASM_MERGEABLE_RODATA_PREFIX | |
5060 | ||
5061 | @hook TARGET_ASM_TM_CLONE_TABLE_SECTION | |
5062 | ||
5063 | @hook TARGET_ASM_SELECT_RTX_SECTION | |
5064 | ||
5065 | @hook TARGET_MANGLE_DECL_ASSEMBLER_NAME | |
5066 | ||
5067 | @hook TARGET_ENCODE_SECTION_INFO | |
5068 | ||
5069 | @hook TARGET_STRIP_NAME_ENCODING | |
5070 | ||
5071 | @hook TARGET_IN_SMALL_DATA_P | |
5072 | ||
5073 | @hook TARGET_HAVE_SRODATA_SECTION | |
5074 | ||
5075 | @hook TARGET_PROFILE_BEFORE_PROLOGUE | |
5076 | ||
5077 | @hook TARGET_BINDS_LOCAL_P | |
5078 | ||
5079 | @hook TARGET_HAVE_TLS | |
5080 | ||
5081 | ||
5082 | @node PIC | |
5083 | @section Position Independent Code | |
5084 | @cindex position independent code | |
5085 | @cindex PIC | |
5086 | ||
5087 | This section describes macros that help implement generation of position | |
5088 | independent code. Simply defining these macros is not enough to | |
5089 | generate valid PIC; you must also add support to the hook | |
5090 | @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro | |
5091 | @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You | |
5092 | must modify the definition of @samp{movsi} to do something appropriate | |
5093 | when the source operand contains a symbolic address. You may also | |
5094 | need to alter the handling of switch statements so that they use | |
5095 | relative addresses. | |
5096 | @c i rearranged the order of the macros above to try to force one of | |
5097 | @c them to the next line, to eliminate an overfull hbox. --mew 10feb93 | |
5098 | ||
5099 | @defmac PIC_OFFSET_TABLE_REGNUM | |
5100 | The register number of the register used to address a table of static | |
5101 | data addresses in memory. In some cases this register is defined by a | |
5102 | processor's ``application binary interface'' (ABI)@. When this macro | |
5103 | is defined, RTL is generated for this register once, as with the stack | |
5104 | pointer and frame pointer registers. If this macro is not defined, it | |
5105 | is up to the machine-dependent files to allocate such a register (if | |
5106 | necessary). Note that this register must be fixed when in use (e.g.@: | |
5107 | when @code{flag_pic} is true). | |
5108 | @end defmac | |
5109 | ||
5110 | @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED | |
5111 | A C expression that is nonzero if the register defined by | |
5112 | @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. If not defined, | |
5113 | the default is zero. Do not define | |
5114 | this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined. | |
5115 | @end defmac | |
5116 | ||
5117 | @defmac LEGITIMATE_PIC_OPERAND_P (@var{x}) | |
5118 | A C expression that is nonzero if @var{x} is a legitimate immediate | |
5119 | operand on the target machine when generating position independent code. | |
5120 | You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not | |
5121 | check this. You can also assume @var{flag_pic} is true, so you need not | |
5122 | check it either. You need not define this macro if all constants | |
5123 | (including @code{SYMBOL_REF}) can be immediate operands when generating | |
5124 | position independent code. | |
5125 | @end defmac | |
5126 | ||
5127 | @node Assembler Format | |
5128 | @section Defining the Output Assembler Language | |
5129 | ||
5130 | This section describes macros whose principal purpose is to describe how | |
5131 | to write instructions in assembler language---rather than what the | |
5132 | instructions do. | |
5133 | ||
5134 | @menu | |
5135 | * File Framework:: Structural information for the assembler file. | |
5136 | * Data Output:: Output of constants (numbers, strings, addresses). | |
5137 | * Uninitialized Data:: Output of uninitialized variables. | |
5138 | * Label Output:: Output and generation of labels. | |
5139 | * Initialization:: General principles of initialization | |
5140 | and termination routines. | |
5141 | * Macros for Initialization:: | |
5142 | Specific macros that control the handling of | |
5143 | initialization and termination routines. | |
5144 | * Instruction Output:: Output of actual instructions. | |
5145 | * Dispatch Tables:: Output of jump tables. | |
5146 | * Exception Region Output:: Output of exception region code. | |
5147 | * Alignment Output:: Pseudo ops for alignment and skipping data. | |
5148 | @end menu | |
5149 | ||
5150 | @node File Framework | |
5151 | @subsection The Overall Framework of an Assembler File | |
5152 | @cindex assembler format | |
5153 | @cindex output of assembler code | |
5154 | ||
5155 | @c prevent bad page break with this line | |
5156 | This describes the overall framework of an assembly file. | |
5157 | ||
5158 | @findex default_file_start | |
5159 | @hook TARGET_ASM_FILE_START | |
5160 | ||
5161 | @hook TARGET_ASM_FILE_START_APP_OFF | |
5162 | ||
5163 | @hook TARGET_ASM_FILE_START_FILE_DIRECTIVE | |
5164 | ||
5165 | @hook TARGET_ASM_FILE_END | |
5166 | ||
5167 | @deftypefun void file_end_indicate_exec_stack () | |
5168 | Some systems use a common convention, the @samp{.note.GNU-stack} | |
5169 | special section, to indicate whether or not an object file relies on | |
5170 | the stack being executable. If your system uses this convention, you | |
5171 | should define @code{TARGET_ASM_FILE_END} to this function. If you | |
5172 | need to do other things in that hook, have your hook function call | |
5173 | this function. | |
5174 | @end deftypefun | |
5175 | ||
5176 | @hook TARGET_ASM_LTO_START | |
5177 | ||
5178 | @hook TARGET_ASM_LTO_END | |
5179 | ||
5180 | @hook TARGET_ASM_CODE_END | |
5181 | ||
5182 | @defmac ASM_COMMENT_START | |
5183 | A C string constant describing how to begin a comment in the target | |
5184 | assembler language. The compiler assumes that the comment will end at | |
5185 | the end of the line. | |
5186 | @end defmac | |
5187 | ||
5188 | @defmac ASM_APP_ON | |
5189 | A C string constant for text to be output before each @code{asm} | |
5190 | statement or group of consecutive ones. Normally this is | |
5191 | @code{"#APP"}, which is a comment that has no effect on most | |
5192 | assemblers but tells the GNU assembler that it must check the lines | |
5193 | that follow for all valid assembler constructs. | |
5194 | @end defmac | |
5195 | ||
5196 | @defmac ASM_APP_OFF | |
5197 | A C string constant for text to be output after each @code{asm} | |
5198 | statement or group of consecutive ones. Normally this is | |
5199 | @code{"#NO_APP"}, which tells the GNU assembler to resume making the | |
5200 | time-saving assumptions that are valid for ordinary compiler output. | |
5201 | @end defmac | |
5202 | ||
5203 | @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name}) | |
5204 | A C statement to output COFF information or DWARF debugging information | |
5205 | which indicates that filename @var{name} is the current source file to | |
5206 | the stdio stream @var{stream}. | |
5207 | ||
5208 | This macro need not be defined if the standard form of output | |
5209 | for the file format in use is appropriate. | |
5210 | @end defmac | |
5211 | ||
5212 | @hook TARGET_ASM_OUTPUT_SOURCE_FILENAME | |
5213 | ||
5214 | @hook TARGET_ASM_OUTPUT_IDENT | |
5215 | ||
5216 | @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string}) | |
5217 | A C statement to output the string @var{string} to the stdio stream | |
5218 | @var{stream}. If you do not call the function @code{output_quoted_string} | |
5219 | in your config files, GCC will only call it to output filenames to | |
5220 | the assembler source. So you can use it to canonicalize the format | |
5221 | of the filename using this macro. | |
5222 | @end defmac | |
5223 | ||
5224 | @hook TARGET_ASM_NAMED_SECTION | |
5225 | ||
5226 | @hook TARGET_ASM_ELF_FLAGS_NUMERIC | |
5227 | ||
5228 | @hook TARGET_ASM_FUNCTION_SECTION | |
5229 | ||
5230 | @hook TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS | |
5231 | ||
5232 | @hook TARGET_HAVE_NAMED_SECTIONS | |
5233 | This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}. | |
5234 | It must not be modified by command-line option processing. | |
5235 | @end deftypevr | |
5236 | ||
5237 | @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS} | |
5238 | @hook TARGET_HAVE_SWITCHABLE_BSS_SECTIONS | |
5239 | ||
5240 | @hook TARGET_SECTION_TYPE_FLAGS | |
5241 | ||
5242 | @hook TARGET_ASM_RECORD_GCC_SWITCHES | |
5243 | ||
5244 | @hook TARGET_ASM_RECORD_GCC_SWITCHES_SECTION | |
5245 | ||
5246 | @need 2000 | |
5247 | @node Data Output | |
5248 | @subsection Output of Data | |
5249 | ||
5250 | ||
5251 | @hook TARGET_ASM_BYTE_OP | |
5252 | ||
5253 | @hook TARGET_ASM_INTEGER | |
5254 | ||
5255 | @hook TARGET_ASM_DECL_END | |
5256 | ||
5257 | @hook TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA | |
5258 | ||
5259 | @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len}) | |
5260 | A C statement to output to the stdio stream @var{stream} an assembler | |
5261 | instruction to assemble a string constant containing the @var{len} | |
5262 | bytes at @var{ptr}. @var{ptr} will be a C expression of type | |
5263 | @code{char *} and @var{len} a C expression of type @code{int}. | |
5264 | ||
5265 | If the assembler has a @code{.ascii} pseudo-op as found in the | |
5266 | Berkeley Unix assembler, do not define the macro | |
5267 | @code{ASM_OUTPUT_ASCII}. | |
5268 | @end defmac | |
5269 | ||
5270 | @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n}) | |
5271 | A C statement to output word @var{n} of a function descriptor for | |
5272 | @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS} | |
5273 | is defined, and is otherwise unused. | |
5274 | @end defmac | |
5275 | ||
5276 | @defmac CONSTANT_POOL_BEFORE_FUNCTION | |
5277 | You may define this macro as a C expression. You should define the | |
5278 | expression to have a nonzero value if GCC should output the constant | |
5279 | pool for a function before the code for the function, or a zero value if | |
5280 | GCC should output the constant pool after the function. If you do | |
5281 | not define this macro, the usual case, GCC will output the constant | |
5282 | pool before the function. | |
5283 | @end defmac | |
5284 | ||
5285 | @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size}) | |
5286 | A C statement to output assembler commands to define the start of the | |
5287 | constant pool for a function. @var{funname} is a string giving | |
5288 | the name of the function. Should the return type of the function | |
5289 | be required, it can be obtained via @var{fundecl}. @var{size} | |
5290 | is the size, in bytes, of the constant pool that will be written | |
5291 | immediately after this call. | |
5292 | ||
5293 | If no constant-pool prefix is required, the usual case, this macro need | |
5294 | not be defined. | |
5295 | @end defmac | |
5296 | ||
5297 | @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto}) | |
5298 | A C statement (with or without semicolon) to output a constant in the | |
5299 | constant pool, if it needs special treatment. (This macro need not do | |
5300 | anything for RTL expressions that can be output normally.) | |
5301 | ||
5302 | The argument @var{file} is the standard I/O stream to output the | |
5303 | assembler code on. @var{x} is the RTL expression for the constant to | |
5304 | output, and @var{mode} is the machine mode (in case @var{x} is a | |
5305 | @samp{const_int}). @var{align} is the required alignment for the value | |
5306 | @var{x}; you should output an assembler directive to force this much | |
5307 | alignment. | |
5308 | ||
5309 | The argument @var{labelno} is a number to use in an internal label for | |
5310 | the address of this pool entry. The definition of this macro is | |
5311 | responsible for outputting the label definition at the proper place. | |
5312 | Here is how to do this: | |
5313 | ||
5314 | @smallexample | |
5315 | @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno}); | |
5316 | @end smallexample | |
5317 | ||
5318 | When you output a pool entry specially, you should end with a | |
5319 | @code{goto} to the label @var{jumpto}. This will prevent the same pool | |
5320 | entry from being output a second time in the usual manner. | |
5321 | ||
5322 | You need not define this macro if it would do nothing. | |
5323 | @end defmac | |
5324 | ||
5325 | @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size}) | |
5326 | A C statement to output assembler commands to at the end of the constant | |
5327 | pool for a function. @var{funname} is a string giving the name of the | |
5328 | function. Should the return type of the function be required, you can | |
5329 | obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the | |
5330 | constant pool that GCC wrote immediately before this call. | |
5331 | ||
5332 | If no constant-pool epilogue is required, the usual case, you need not | |
5333 | define this macro. | |
5334 | @end defmac | |
5335 | ||
5336 | @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR}) | |
5337 | Define this macro as a C expression which is nonzero if @var{C} is | |
5338 | used as a logical line separator by the assembler. @var{STR} points | |
5339 | to the position in the string where @var{C} was found; this can be used if | |
5340 | a line separator uses multiple characters. | |
5341 | ||
5342 | If you do not define this macro, the default is that only | |
5343 | the character @samp{;} is treated as a logical line separator. | |
5344 | @end defmac | |
5345 | ||
5346 | @hook TARGET_ASM_OPEN_PAREN | |
5347 | ||
5348 | These macros are provided by @file{real.h} for writing the definitions | |
5349 | of @code{ASM_OUTPUT_DOUBLE} and the like: | |
5350 | ||
5351 | @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l}) | |
5352 | @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l}) | |
5353 | @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l}) | |
5354 | @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l}) | |
5355 | @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l}) | |
5356 | @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l}) | |
5357 | These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the | |
5358 | target's floating point representation, and store its bit pattern in | |
5359 | the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and | |
5360 | @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a | |
5361 | simple @code{long int}. For the others, it should be an array of | |
5362 | @code{long int}. The number of elements in this array is determined | |
5363 | by the size of the desired target floating point data type: 32 bits of | |
5364 | it go in each @code{long int} array element. Each array element holds | |
5365 | 32 bits of the result, even if @code{long int} is wider than 32 bits | |
5366 | on the host machine. | |
5367 | ||
5368 | The array element values are designed so that you can print them out | |
5369 | using @code{fprintf} in the order they should appear in the target | |
5370 | machine's memory. | |
5371 | @end defmac | |
5372 | ||
5373 | @node Uninitialized Data | |
5374 | @subsection Output of Uninitialized Variables | |
5375 | ||
5376 | Each of the macros in this section is used to do the whole job of | |
5377 | outputting a single uninitialized variable. | |
5378 | ||
5379 | @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
5380 | A C statement (sans semicolon) to output to the stdio stream | |
5381 | @var{stream} the assembler definition of a common-label named | |
5382 | @var{name} whose size is @var{size} bytes. The variable @var{rounded} | |
5383 | is the size rounded up to whatever alignment the caller wants. It is | |
5384 | possible that @var{size} may be zero, for instance if a struct with no | |
5385 | other member than a zero-length array is defined. In this case, the | |
5386 | backend must output a symbol definition that allocates at least one | |
5387 | byte, both so that the address of the resulting object does not compare | |
5388 | equal to any other, and because some object formats cannot even express | |
5389 | the concept of a zero-sized common symbol, as that is how they represent | |
5390 | an ordinary undefined external. | |
5391 | ||
5392 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
5393 | output the name itself; before and after that, output the additional | |
5394 | assembler syntax for defining the name, and a newline. | |
5395 | ||
5396 | This macro controls how the assembler definitions of uninitialized | |
5397 | common global variables are output. | |
5398 | @end defmac | |
5399 | ||
5400 | @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment}) | |
5401 | Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a | |
5402 | separate, explicit argument. If you define this macro, it is used in | |
5403 | place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in | |
5404 | handling the required alignment of the variable. The alignment is specified | |
5405 | as the number of bits. | |
5406 | @end defmac | |
5407 | ||
5408 | @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) | |
5409 | Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the | |
5410 | variable to be output, if there is one, or @code{NULL_TREE} if there | |
5411 | is no corresponding variable. If you define this macro, GCC will use it | |
5412 | in place of both @code{ASM_OUTPUT_COMMON} and | |
5413 | @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see | |
5414 | the variable's decl in order to chose what to output. | |
5415 | @end defmac | |
5416 | ||
5417 | @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) | |
5418 | A C statement (sans semicolon) to output to the stdio stream | |
5419 | @var{stream} the assembler definition of uninitialized global @var{decl} named | |
5420 | @var{name} whose size is @var{size} bytes. The variable @var{alignment} | |
5421 | is the alignment specified as the number of bits. | |
5422 | ||
5423 | Try to use function @code{asm_output_aligned_bss} defined in file | |
5424 | @file{varasm.cc} when defining this macro. If unable, use the expression | |
5425 | @code{assemble_name (@var{stream}, @var{name})} to output the name itself; | |
5426 | before and after that, output the additional assembler syntax for defining | |
5427 | the name, and a newline. | |
5428 | ||
5429 | There are two ways of handling global BSS@. One is to define this macro. | |
5430 | The other is to have @code{TARGET_ASM_SELECT_SECTION} return a | |
5431 | switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}). | |
5432 | You do not need to do both. | |
5433 | ||
5434 | Some languages do not have @code{common} data, and require a | |
5435 | non-common form of global BSS in order to handle uninitialized globals | |
5436 | efficiently. C++ is one example of this. However, if the target does | |
5437 | not support global BSS, the front end may choose to make globals | |
5438 | common in order to save space in the object file. | |
5439 | @end defmac | |
5440 | ||
5441 | @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
5442 | A C statement (sans semicolon) to output to the stdio stream | |
5443 | @var{stream} the assembler definition of a local-common-label named | |
5444 | @var{name} whose size is @var{size} bytes. The variable @var{rounded} | |
5445 | is the size rounded up to whatever alignment the caller wants. | |
5446 | ||
5447 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
5448 | output the name itself; before and after that, output the additional | |
5449 | assembler syntax for defining the name, and a newline. | |
5450 | ||
5451 | This macro controls how the assembler definitions of uninitialized | |
5452 | static variables are output. | |
5453 | @end defmac | |
5454 | ||
5455 | @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment}) | |
5456 | Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a | |
5457 | separate, explicit argument. If you define this macro, it is used in | |
5458 | place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in | |
5459 | handling the required alignment of the variable. The alignment is specified | |
5460 | as the number of bits. | |
5461 | @end defmac | |
5462 | ||
5463 | @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) | |
5464 | Like @code{ASM_OUTPUT_ALIGNED_LOCAL} except that @var{decl} of the | |
5465 | variable to be output, if there is one, or @code{NULL_TREE} if there | |
5466 | is no corresponding variable. If you define this macro, GCC will use it | |
5467 | in place of both @code{ASM_OUTPUT_LOCAL} and | |
5468 | @code{ASM_OUTPUT_ALIGNED_LOCAL}. Define this macro when you need to see | |
5469 | the variable's decl in order to chose what to output. | |
5470 | @end defmac | |
5471 | ||
5472 | @node Label Output | |
5473 | @subsection Output and Generation of Labels | |
5474 | ||
5475 | @c prevent bad page break with this line | |
5476 | This is about outputting labels. | |
5477 | ||
5478 | @findex assemble_name | |
5479 | @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name}) | |
5480 | A C statement (sans semicolon) to output to the stdio stream | |
5481 | @var{stream} the assembler definition of a label named @var{name}. | |
5482 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
5483 | output the name itself; before and after that, output the additional | |
5484 | assembler syntax for defining the name, and a newline. A default | |
5485 | definition of this macro is provided which is correct for most systems. | |
5486 | @end defmac | |
5487 | ||
5488 | @defmac ASM_OUTPUT_FUNCTION_LABEL (@var{stream}, @var{name}, @var{decl}) | |
5489 | A C statement (sans semicolon) to output to the stdio stream | |
5490 | @var{stream} the assembler definition of a label named @var{name} of | |
5491 | a function. | |
5492 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
5493 | output the name itself; before and after that, output the additional | |
5494 | assembler syntax for defining the name, and a newline. A default | |
5495 | definition of this macro is provided which is correct for most systems. | |
5496 | ||
5497 | If this macro is not defined, then the function name is defined in the | |
5498 | usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). | |
5499 | @end defmac | |
5500 | ||
5501 | @findex assemble_name_raw | |
5502 | @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name}) | |
5503 | Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known | |
5504 | to refer to a compiler-generated label. The default definition uses | |
5505 | @code{assemble_name_raw}, which is like @code{assemble_name} except | |
5506 | that it is more efficient. | |
5507 | @end defmac | |
5508 | ||
5509 | @defmac SIZE_ASM_OP | |
5510 | A C string containing the appropriate assembler directive to specify the | |
5511 | size of a symbol, without any arguments. On systems that use ELF, the | |
5512 | default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other | |
5513 | systems, the default is not to define this macro. | |
5514 | ||
5515 | Define this macro only if it is correct to use the default definitions | |
5516 | of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE} | |
5517 | for your system. If you need your own custom definitions of those | |
5518 | macros, or if you do not need explicit symbol sizes at all, do not | |
5519 | define this macro. | |
5520 | @end defmac | |
5521 | ||
5522 | @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size}) | |
5523 | A C statement (sans semicolon) to output to the stdio stream | |
5524 | @var{stream} a directive telling the assembler that the size of the | |
5525 | symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}. | |
5526 | If you define @code{SIZE_ASM_OP}, a default definition of this macro is | |
5527 | provided. | |
5528 | @end defmac | |
5529 | ||
5530 | @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name}) | |
5531 | A C statement (sans semicolon) to output to the stdio stream | |
5532 | @var{stream} a directive telling the assembler to calculate the size of | |
5533 | the symbol @var{name} by subtracting its address from the current | |
5534 | address. | |
5535 | ||
5536 | If you define @code{SIZE_ASM_OP}, a default definition of this macro is | |
5537 | provided. The default assumes that the assembler recognizes a special | |
5538 | @samp{.} symbol as referring to the current address, and can calculate | |
5539 | the difference between this and another symbol. If your assembler does | |
5540 | not recognize @samp{.} or cannot do calculations with it, you will need | |
5541 | to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique. | |
5542 | @end defmac | |
5543 | ||
5544 | @defmac NO_DOLLAR_IN_LABEL | |
5545 | Define this macro if the assembler does not accept the character | |
5546 | @samp{$} in label names. By default constructors and destructors in | |
5547 | G++ have @samp{$} in the identifiers. If this macro is defined, | |
5548 | @samp{.} is used instead. | |
5549 | @end defmac | |
5550 | ||
5551 | @defmac NO_DOT_IN_LABEL | |
5552 | Define this macro if the assembler does not accept the character | |
5553 | @samp{.} in label names. By default constructors and destructors in G++ | |
5554 | have names that use @samp{.}. If this macro is defined, these names | |
5555 | are rewritten to avoid @samp{.}. | |
5556 | @end defmac | |
5557 | ||
5558 | @defmac TYPE_ASM_OP | |
5559 | A C string containing the appropriate assembler directive to specify the | |
5560 | type of a symbol, without any arguments. On systems that use ELF, the | |
5561 | default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other | |
5562 | systems, the default is not to define this macro. | |
5563 | ||
5564 | Define this macro only if it is correct to use the default definition of | |
5565 | @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own | |
5566 | custom definition of this macro, or if you do not need explicit symbol | |
5567 | types at all, do not define this macro. | |
5568 | @end defmac | |
5569 | ||
5570 | @defmac TYPE_OPERAND_FMT | |
5571 | A C string which specifies (using @code{printf} syntax) the format of | |
5572 | the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the | |
5573 | default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems, | |
5574 | the default is not to define this macro. | |
5575 | ||
5576 | Define this macro only if it is correct to use the default definition of | |
5577 | @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own | |
5578 | custom definition of this macro, or if you do not need explicit symbol | |
5579 | types at all, do not define this macro. | |
5580 | @end defmac | |
5581 | ||
5582 | @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type}) | |
5583 | A C statement (sans semicolon) to output to the stdio stream | |
5584 | @var{stream} a directive telling the assembler that the type of the | |
5585 | symbol @var{name} is @var{type}. @var{type} is a C string; currently, | |
5586 | that string is always either @samp{"function"} or @samp{"object"}, but | |
5587 | you should not count on this. | |
5588 | ||
5589 | If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default | |
5590 | definition of this macro is provided. | |
5591 | @end defmac | |
5592 | ||
5593 | @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl}) | |
5594 | A C statement (sans semicolon) to output to the stdio stream | |
5595 | @var{stream} any text necessary for declaring the name @var{name} of a | |
5596 | function which is being defined. This macro is responsible for | |
5597 | outputting the label definition (perhaps using | |
5598 | @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the | |
5599 | @code{FUNCTION_DECL} tree node representing the function. | |
5600 | ||
5601 | If this macro is not defined, then the function name is defined in the | |
5602 | usual manner as a label (by means of @code{ASM_OUTPUT_FUNCTION_LABEL}). | |
5603 | ||
5604 | You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition | |
5605 | of this macro. | |
5606 | @end defmac | |
5607 | ||
5608 | @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl}) | |
5609 | A C statement (sans semicolon) to output to the stdio stream | |
5610 | @var{stream} any text necessary for declaring the size of a function | |
5611 | which is being defined. The argument @var{name} is the name of the | |
5612 | function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node | |
5613 | representing the function. | |
5614 | ||
5615 | If this macro is not defined, then the function size is not defined. | |
5616 | ||
5617 | You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition | |
5618 | of this macro. | |
5619 | @end defmac | |
5620 | ||
5621 | @defmac ASM_DECLARE_COLD_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl}) | |
5622 | A C statement (sans semicolon) to output to the stdio stream | |
5623 | @var{stream} any text necessary for declaring the name @var{name} of a | |
5624 | cold function partition which is being defined. This macro is responsible | |
5625 | for outputting the label definition (perhaps using | |
5626 | @code{ASM_OUTPUT_FUNCTION_LABEL}). The argument @var{decl} is the | |
5627 | @code{FUNCTION_DECL} tree node representing the function. | |
5628 | ||
5629 | If this macro is not defined, then the cold partition name is defined in the | |
5630 | usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). | |
5631 | ||
5632 | You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition | |
5633 | of this macro. | |
5634 | @end defmac | |
5635 | ||
5636 | @defmac ASM_DECLARE_COLD_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl}) | |
5637 | A C statement (sans semicolon) to output to the stdio stream | |
5638 | @var{stream} any text necessary for declaring the size of a cold function | |
5639 | partition which is being defined. The argument @var{name} is the name of the | |
5640 | cold partition of the function. The argument @var{decl} is the | |
5641 | @code{FUNCTION_DECL} tree node representing the function. | |
5642 | ||
5643 | If this macro is not defined, then the partition size is not defined. | |
5644 | ||
5645 | You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition | |
5646 | of this macro. | |
5647 | @end defmac | |
5648 | ||
5649 | @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl}) | |
5650 | A C statement (sans semicolon) to output to the stdio stream | |
5651 | @var{stream} any text necessary for declaring the name @var{name} of an | |
5652 | initialized variable which is being defined. This macro must output the | |
5653 | label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument | |
5654 | @var{decl} is the @code{VAR_DECL} tree node representing the variable. | |
5655 | ||
5656 | If this macro is not defined, then the variable name is defined in the | |
5657 | usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). | |
5658 | ||
5659 | You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or | |
5660 | @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro. | |
5661 | @end defmac | |
5662 | ||
5663 | @hook TARGET_ASM_DECLARE_CONSTANT_NAME | |
5664 | ||
5665 | @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name}) | |
5666 | A C statement (sans semicolon) to output to the stdio stream | |
5667 | @var{stream} any text necessary for claiming a register @var{regno} | |
5668 | for a global variable @var{decl} with name @var{name}. | |
5669 | ||
5670 | If you don't define this macro, that is equivalent to defining it to do | |
5671 | nothing. | |
5672 | @end defmac | |
5673 | ||
5674 | @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend}) | |
5675 | A C statement (sans semicolon) to finish up declaring a variable name | |
5676 | once the compiler has processed its initializer fully and thus has had a | |
5677 | chance to determine the size of an array when controlled by an | |
5678 | initializer. This is used on systems where it's necessary to declare | |
5679 | something about the size of the object. | |
5680 | ||
5681 | If you don't define this macro, that is equivalent to defining it to do | |
5682 | nothing. | |
5683 | ||
5684 | You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or | |
5685 | @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro. | |
5686 | @end defmac | |
5687 | ||
5688 | @hook TARGET_ASM_GLOBALIZE_LABEL | |
5689 | ||
5690 | @hook TARGET_ASM_GLOBALIZE_DECL_NAME | |
5691 | ||
5692 | @hook TARGET_ASM_ASSEMBLE_UNDEFINED_DECL | |
5693 | ||
5694 | @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name}) | |
5695 | A C statement (sans semicolon) to output to the stdio stream | |
5696 | @var{stream} some commands that will make the label @var{name} weak; | |
5697 | that is, available for reference from other files but only used if | |
5698 | no other definition is available. Use the expression | |
5699 | @code{assemble_name (@var{stream}, @var{name})} to output the name | |
5700 | itself; before and after that, output the additional assembler syntax | |
5701 | for making that name weak, and a newline. | |
5702 | ||
5703 | If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not | |
5704 | support weak symbols and you should not define the @code{SUPPORTS_WEAK} | |
5705 | macro. | |
5706 | @end defmac | |
5707 | ||
5708 | @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value}) | |
5709 | Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and | |
5710 | @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function | |
5711 | or variable decl. If @var{value} is not @code{NULL}, this C statement | |
5712 | should output to the stdio stream @var{stream} assembler code which | |
5713 | defines (equates) the weak symbol @var{name} to have the value | |
5714 | @var{value}. If @var{value} is @code{NULL}, it should output commands | |
5715 | to make @var{name} weak. | |
5716 | @end defmac | |
5717 | ||
5718 | @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value}) | |
5719 | Outputs a directive that enables @var{name} to be used to refer to | |
5720 | symbol @var{value} with weak-symbol semantics. @code{decl} is the | |
5721 | declaration of @code{name}. | |
5722 | @end defmac | |
5723 | ||
5724 | @defmac SUPPORTS_WEAK | |
5725 | A preprocessor constant expression which evaluates to true if the target | |
5726 | supports weak symbols. | |
5727 | ||
5728 | If you don't define this macro, @file{defaults.h} provides a default | |
5729 | definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL} | |
5730 | is defined, the default definition is @samp{1}; otherwise, it is @samp{0}. | |
5731 | @end defmac | |
5732 | ||
5733 | @defmac TARGET_SUPPORTS_WEAK | |
5734 | A C expression which evaluates to true if the target supports weak symbols. | |
5735 | ||
5736 | If you don't define this macro, @file{defaults.h} provides a default | |
5737 | definition. The default definition is @samp{(SUPPORTS_WEAK)}. Define | |
5738 | this macro if you want to control weak symbol support with a compiler | |
5739 | flag such as @option{-melf}. | |
5740 | @end defmac | |
5741 | ||
5742 | @defmac MAKE_DECL_ONE_ONLY (@var{decl}) | |
5743 | A C statement (sans semicolon) to mark @var{decl} to be emitted as a | |
5744 | public symbol such that extra copies in multiple translation units will | |
5745 | be discarded by the linker. Define this macro if your object file | |
5746 | format provides support for this concept, such as the @samp{COMDAT} | |
5747 | section flags in the Microsoft Windows PE/COFF format, and this support | |
5748 | requires changes to @var{decl}, such as putting it in a separate section. | |
5749 | @end defmac | |
5750 | ||
5751 | @defmac SUPPORTS_ONE_ONLY | |
5752 | A C expression which evaluates to true if the target supports one-only | |
5753 | semantics. | |
5754 | ||
5755 | If you don't define this macro, @file{varasm.cc} provides a default | |
5756 | definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default | |
5757 | definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if | |
5758 | you want to control one-only symbol support with a compiler flag, or if | |
5759 | setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to | |
5760 | be emitted as one-only. | |
5761 | @end defmac | |
5762 | ||
5763 | @hook TARGET_ASM_ASSEMBLE_VISIBILITY | |
5764 | ||
5765 | @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC | |
5766 | A C expression that evaluates to true if the target's linker expects | |
5767 | that weak symbols do not appear in a static archive's table of contents. | |
5768 | The default is @code{0}. | |
5769 | ||
5770 | Leaving weak symbols out of an archive's table of contents means that, | |
5771 | if a symbol will only have a definition in one translation unit and | |
5772 | will have undefined references from other translation units, that | |
5773 | symbol should not be weak. Defining this macro to be nonzero will | |
5774 | thus have the effect that certain symbols that would normally be weak | |
5775 | (explicit template instantiations, and vtables for polymorphic classes | |
5776 | with noninline key methods) will instead be nonweak. | |
5777 | ||
5778 | The C++ ABI requires this macro to be zero. Define this macro for | |
5779 | targets where full C++ ABI compliance is impossible and where linker | |
5780 | restrictions require weak symbols to be left out of a static archive's | |
5781 | table of contents. | |
5782 | @end defmac | |
5783 | ||
5784 | @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name}) | |
5785 | A C statement (sans semicolon) to output to the stdio stream | |
5786 | @var{stream} any text necessary for declaring the name of an external | |
5787 | symbol named @var{name} which is referenced in this compilation but | |
5788 | not defined. The value of @var{decl} is the tree node for the | |
5789 | declaration. | |
5790 | ||
5791 | This macro need not be defined if it does not need to output anything. | |
5792 | The GNU assembler and most Unix assemblers don't require anything. | |
5793 | @end defmac | |
5794 | ||
5795 | @hook TARGET_ASM_EXTERNAL_LIBCALL | |
5796 | ||
5797 | @hook TARGET_ASM_MARK_DECL_PRESERVED | |
5798 | ||
5799 | @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name}) | |
5800 | A C statement (sans semicolon) to output to the stdio stream | |
5801 | @var{stream} a reference in assembler syntax to a label named | |
5802 | @var{name}. This should add @samp{_} to the front of the name, if that | |
5803 | is customary on your operating system, as it is in most Berkeley Unix | |
5804 | systems. This macro is used in @code{assemble_name}. | |
5805 | @end defmac | |
5806 | ||
5807 | @hook TARGET_MANGLE_ASSEMBLER_NAME | |
5808 | ||
5809 | @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym}) | |
5810 | A C statement (sans semicolon) to output a reference to | |
5811 | @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name} | |
5812 | will be used to output the name of the symbol. This macro may be used | |
5813 | to modify the way a symbol is referenced depending on information | |
5814 | encoded by @code{TARGET_ENCODE_SECTION_INFO}. | |
5815 | @end defmac | |
5816 | ||
5817 | @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf}) | |
5818 | A C statement (sans semicolon) to output a reference to @var{buf}, the | |
5819 | result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined, | |
5820 | @code{assemble_name} will be used to output the name of the symbol. | |
5821 | This macro is not used by @code{output_asm_label}, or the @code{%l} | |
5822 | specifier that calls it; the intention is that this macro should be set | |
5823 | when it is necessary to output a label differently when its address is | |
5824 | being taken. | |
5825 | @end defmac | |
5826 | ||
5827 | @hook TARGET_ASM_INTERNAL_LABEL | |
5828 | ||
5829 | @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num}) | |
5830 | A C statement to output to the stdio stream @var{stream} a debug info | |
5831 | label whose name is made from the string @var{prefix} and the number | |
5832 | @var{num}. This is useful for VLIW targets, where debug info labels | |
5833 | may need to be treated differently than branch target labels. On some | |
5834 | systems, branch target labels must be at the beginning of instruction | |
5835 | bundles, but debug info labels can occur in the middle of instruction | |
5836 | bundles. | |
5837 | ||
5838 | If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be | |
5839 | used. | |
5840 | @end defmac | |
5841 | ||
5842 | @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num}) | |
5843 | A C statement to store into the string @var{string} a label whose name | |
5844 | is made from the string @var{prefix} and the number @var{num}. | |
5845 | ||
5846 | This string, when output subsequently by @code{assemble_name}, should | |
5847 | produce the output that @code{(*targetm.asm_out.internal_label)} would produce | |
5848 | with the same @var{prefix} and @var{num}. | |
5849 | ||
5850 | If the string begins with @samp{*}, then @code{assemble_name} will | |
5851 | output the rest of the string unchanged. It is often convenient for | |
5852 | @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the | |
5853 | string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets | |
5854 | to output the string, and may change it. (Of course, | |
5855 | @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so | |
5856 | you should know what it does on your machine.) | |
5857 | @end defmac | |
5858 | ||
5859 | @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number}) | |
5860 | A C expression to assign to @var{outvar} (which is a variable of type | |
5861 | @code{char *}) a newly allocated string made from the string | |
5862 | @var{name} and the number @var{number}, with some suitable punctuation | |
5863 | added. Use @code{alloca} to get space for the string. | |
5864 | ||
5865 | The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to | |
5866 | produce an assembler label for an internal static variable whose name is | |
5867 | @var{name}. Therefore, the string must be such as to result in valid | |
5868 | assembler code. The argument @var{number} is different each time this | |
5869 | macro is executed; it prevents conflicts between similarly-named | |
5870 | internal static variables in different scopes. | |
5871 | ||
5872 | Ideally this string should not be a valid C identifier, to prevent any | |
5873 | conflict with the user's own symbols. Most assemblers allow periods | |
5874 | or percent signs in assembler symbols; putting at least one of these | |
5875 | between the name and the number will suffice. | |
5876 | ||
5877 | If this macro is not defined, a default definition will be provided | |
5878 | which is correct for most systems. | |
5879 | @end defmac | |
5880 | ||
5881 | @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value}) | |
5882 | A C statement to output to the stdio stream @var{stream} assembler code | |
5883 | which defines (equates) the symbol @var{name} to have the value @var{value}. | |
5884 | ||
5885 | @findex SET_ASM_OP | |
5886 | If @code{SET_ASM_OP} is defined, a default definition is provided which is | |
5887 | correct for most systems. | |
5888 | @end defmac | |
5889 | ||
5890 | @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value}) | |
5891 | A C statement to output to the stdio stream @var{stream} assembler code | |
5892 | which defines (equates) the symbol whose tree node is @var{decl_of_name} | |
5893 | to have the value of the tree node @var{decl_of_value}. This macro will | |
5894 | be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if | |
5895 | the tree nodes are available. | |
5896 | ||
5897 | @findex SET_ASM_OP | |
5898 | If @code{SET_ASM_OP} is defined, a default definition is provided which is | |
5899 | correct for most systems. | |
5900 | @end defmac | |
5901 | ||
5902 | @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value}) | |
5903 | A C statement that evaluates to true if the assembler code which defines | |
5904 | (equates) the symbol whose tree node is @var{decl_of_name} to have the value | |
5905 | of the tree node @var{decl_of_value} should be emitted near the end of the | |
5906 | current compilation unit. The default is to not defer output of defines. | |
5907 | This macro affects defines output by @samp{ASM_OUTPUT_DEF} and | |
5908 | @samp{ASM_OUTPUT_DEF_FROM_DECLS}. | |
5909 | @end defmac | |
5910 | ||
5911 | @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value}) | |
5912 | A C statement to output to the stdio stream @var{stream} assembler code | |
5913 | which defines (equates) the weak symbol @var{name} to have the value | |
5914 | @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as | |
5915 | an undefined weak symbol. | |
5916 | ||
5917 | Define this macro if the target only supports weak aliases; define | |
5918 | @code{ASM_OUTPUT_DEF} instead if possible. | |
5919 | @end defmac | |
5920 | ||
5921 | @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name}) | |
5922 | Define this macro to override the default assembler names used for | |
5923 | Objective-C methods. | |
5924 | ||
5925 | The default name is a unique method number followed by the name of the | |
5926 | class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of | |
5927 | the category is also included in the assembler name (e.g.@: | |
5928 | @samp{_1_Foo_Bar}). | |
5929 | ||
5930 | These names are safe on most systems, but make debugging difficult since | |
5931 | the method's selector is not present in the name. Therefore, particular | |
5932 | systems define other ways of computing names. | |
5933 | ||
5934 | @var{buf} is an expression of type @code{char *} which gives you a | |
5935 | buffer in which to store the name; its length is as long as | |
5936 | @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus | |
5937 | 50 characters extra. | |
5938 | ||
5939 | The argument @var{is_inst} specifies whether the method is an instance | |
5940 | method or a class method; @var{class_name} is the name of the class; | |
5941 | @var{cat_name} is the name of the category (or @code{NULL} if the method is not | |
5942 | in a category); and @var{sel_name} is the name of the selector. | |
5943 | ||
5944 | On systems where the assembler can handle quoted names, you can use this | |
5945 | macro to provide more human-readable names. | |
5946 | @end defmac | |
5947 | ||
5948 | @node Initialization | |
5949 | @subsection How Initialization Functions Are Handled | |
5950 | @cindex initialization routines | |
5951 | @cindex termination routines | |
5952 | @cindex constructors, output of | |
5953 | @cindex destructors, output of | |
5954 | ||
5955 | The compiled code for certain languages includes @dfn{constructors} | |
5956 | (also called @dfn{initialization routines})---functions to initialize | |
5957 | data in the program when the program is started. These functions need | |
5958 | to be called before the program is ``started''---that is to say, before | |
5959 | @code{main} is called. | |
5960 | ||
5961 | Compiling some languages generates @dfn{destructors} (also called | |
5962 | @dfn{termination routines}) that should be called when the program | |
5963 | terminates. | |
5964 | ||
5965 | To make the initialization and termination functions work, the compiler | |
5966 | must output something in the assembler code to cause those functions to | |
5967 | be called at the appropriate time. When you port the compiler to a new | |
5968 | system, you need to specify how to do this. | |
5969 | ||
5970 | There are two major ways that GCC currently supports the execution of | |
5971 | initialization and termination functions. Each way has two variants. | |
5972 | Much of the structure is common to all four variations. | |
5973 | ||
5974 | @findex __CTOR_LIST__ | |
5975 | @findex __DTOR_LIST__ | |
5976 | The linker must build two lists of these functions---a list of | |
5977 | initialization functions, called @code{__CTOR_LIST__}, and a list of | |
5978 | termination functions, called @code{__DTOR_LIST__}. | |
5979 | ||
5980 | Each list always begins with an ignored function pointer (which may hold | |
5981 | 0, @minus{}1, or a count of the function pointers after it, depending on | |
5982 | the environment). This is followed by a series of zero or more function | |
5983 | pointers to constructors (or destructors), followed by a function | |
5984 | pointer containing zero. | |
5985 | ||
5986 | Depending on the operating system and its executable file format, either | |
5987 | @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup | |
5988 | time and exit time. Constructors are called in reverse order of the | |
5989 | list; destructors in forward order. | |
5990 | ||
5991 | The best way to handle static constructors works only for object file | |
5992 | formats which provide arbitrarily-named sections. A section is set | |
5993 | aside for a list of constructors, and another for a list of destructors. | |
5994 | Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each | |
5995 | object file that defines an initialization function also puts a word in | |
5996 | the constructor section to point to that function. The linker | |
5997 | accumulates all these words into one contiguous @samp{.ctors} section. | |
5998 | Termination functions are handled similarly. | |
5999 | ||
6000 | This method will be chosen as the default by @file{target-def.h} if | |
6001 | @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not | |
6002 | support arbitrary sections, but does support special designated | |
6003 | constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP} | |
6004 | and @code{DTORS_SECTION_ASM_OP} to achieve the same effect. | |
6005 | ||
6006 | When arbitrary sections are available, there are two variants, depending | |
6007 | upon how the code in @file{crtstuff.c} is called. On systems that | |
6008 | support a @dfn{.init} section which is executed at program startup, | |
6009 | parts of @file{crtstuff.c} are compiled into that section. The | |
6010 | program is linked by the @command{gcc} driver like this: | |
6011 | ||
6012 | @smallexample | |
6013 | ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o | |
6014 | @end smallexample | |
6015 | ||
6016 | The prologue of a function (@code{__init}) appears in the @code{.init} | |
6017 | section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise | |
6018 | for the function @code{__fini} in the @dfn{.fini} section. Normally these | |
6019 | files are provided by the operating system or by the GNU C library, but | |
6020 | are provided by GCC for a few targets. | |
6021 | ||
6022 | The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets) | |
6023 | compiled from @file{crtstuff.c}. They contain, among other things, code | |
6024 | fragments within the @code{.init} and @code{.fini} sections that branch | |
6025 | to routines in the @code{.text} section. The linker will pull all parts | |
6026 | of a section together, which results in a complete @code{__init} function | |
6027 | that invokes the routines we need at startup. | |
6028 | ||
6029 | To use this variant, you must define the @code{INIT_SECTION_ASM_OP} | |
6030 | macro properly. | |
6031 | ||
6032 | If no init section is available, when GCC compiles any function called | |
6033 | @code{main} (or more accurately, any function designated as a program | |
6034 | entry point by the language front end calling @code{expand_main_function}), | |
6035 | it inserts a procedure call to @code{__main} as the first executable code | |
6036 | after the function prologue. The @code{__main} function is defined | |
6037 | in @file{libgcc2.c} and runs the global constructors. | |
6038 | ||
6039 | In file formats that don't support arbitrary sections, there are again | |
6040 | two variants. In the simplest variant, the GNU linker (GNU @code{ld}) | |
6041 | and an `a.out' format must be used. In this case, | |
6042 | @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs} | |
6043 | entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__}, | |
6044 | and with the address of the void function containing the initialization | |
6045 | code as its value. The GNU linker recognizes this as a request to add | |
6046 | the value to a @dfn{set}; the values are accumulated, and are eventually | |
6047 | placed in the executable as a vector in the format described above, with | |
6048 | a leading (ignored) count and a trailing zero element. | |
6049 | @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init | |
6050 | section is available, the absence of @code{INIT_SECTION_ASM_OP} causes | |
6051 | the compilation of @code{main} to call @code{__main} as above, starting | |
6052 | the initialization process. | |
6053 | ||
6054 | The last variant uses neither arbitrary sections nor the GNU linker. | |
6055 | This is preferable when you want to do dynamic linking and when using | |
6056 | file formats which the GNU linker does not support, such as `ECOFF'@. In | |
6057 | this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and | |
6058 | termination functions are recognized simply by their names. This requires | |
6059 | an extra program in the linkage step, called @command{collect2}. This program | |
6060 | pretends to be the linker, for use with GCC; it does its job by running | |
6061 | the ordinary linker, but also arranges to include the vectors of | |
6062 | initialization and termination functions. These functions are called | |
6063 | via @code{__main} as described above. In order to use this method, | |
6064 | @code{use_collect2} must be defined in the target in @file{config.gcc}. | |
6065 | ||
6066 | @ifinfo | |
6067 | The following section describes the specific macros that control and | |
6068 | customize the handling of initialization and termination functions. | |
6069 | @end ifinfo | |
6070 | ||
6071 | @node Macros for Initialization | |
6072 | @subsection Macros Controlling Initialization Routines | |
6073 | ||
6074 | Here are the macros that control how the compiler handles initialization | |
6075 | and termination functions: | |
6076 | ||
6077 | @defmac INIT_SECTION_ASM_OP | |
6078 | If defined, a C string constant, including spacing, for the assembler | |
6079 | operation to identify the following data as initialization code. If not | |
6080 | defined, GCC will assume such a section does not exist. When you are | |
6081 | using special sections for initialization and termination functions, this | |
6082 | macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to | |
6083 | run the initialization functions. | |
6084 | @end defmac | |
6085 | ||
6086 | @defmac HAS_INIT_SECTION | |
6087 | If defined, @code{main} will not call @code{__main} as described above. | |
6088 | This macro should be defined for systems that control start-up code | |
6089 | on a symbol-by-symbol basis, such as OSF/1, and should not | |
6090 | be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}. | |
6091 | @end defmac | |
6092 | ||
6093 | @defmac LD_INIT_SWITCH | |
6094 | If defined, a C string constant for a switch that tells the linker that | |
6095 | the following symbol is an initialization routine. | |
6096 | @end defmac | |
6097 | ||
6098 | @defmac LD_FINI_SWITCH | |
6099 | If defined, a C string constant for a switch that tells the linker that | |
6100 | the following symbol is a finalization routine. | |
6101 | @end defmac | |
6102 | ||
6103 | @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func}) | |
6104 | If defined, a C statement that will write a function that can be | |
6105 | automatically called when a shared library is loaded. The function | |
6106 | should call @var{func}, which takes no arguments. If not defined, and | |
6107 | the object format requires an explicit initialization function, then a | |
6108 | function called @code{_GLOBAL__DI} will be generated. | |
6109 | ||
6110 | This function and the following one are used by collect2 when linking a | |
6111 | shared library that needs constructors or destructors, or has DWARF2 | |
6112 | exception tables embedded in the code. | |
6113 | @end defmac | |
6114 | ||
6115 | @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func}) | |
6116 | If defined, a C statement that will write a function that can be | |
6117 | automatically called when a shared library is unloaded. The function | |
6118 | should call @var{func}, which takes no arguments. If not defined, and | |
6119 | the object format requires an explicit finalization function, then a | |
6120 | function called @code{_GLOBAL__DD} will be generated. | |
6121 | @end defmac | |
6122 | ||
6123 | @defmac INVOKE__main | |
6124 | If defined, @code{main} will call @code{__main} despite the presence of | |
6125 | @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems | |
6126 | where the init section is not actually run automatically, but is still | |
6127 | useful for collecting the lists of constructors and destructors. | |
6128 | @end defmac | |
6129 | ||
6130 | @defmac SUPPORTS_INIT_PRIORITY | |
6131 | If nonzero, the C++ @code{init_priority} attribute is supported and the | |
6132 | compiler should emit instructions to control the order of initialization | |
6133 | of objects. If zero, the compiler will issue an error message upon | |
6134 | encountering an @code{init_priority} attribute. | |
6135 | @end defmac | |
6136 | ||
6137 | @hook TARGET_HAVE_CTORS_DTORS | |
6138 | ||
6139 | @hook TARGET_DTORS_FROM_CXA_ATEXIT | |
6140 | ||
6141 | @hook TARGET_ASM_CONSTRUCTOR | |
6142 | ||
6143 | @hook TARGET_ASM_DESTRUCTOR | |
6144 | ||
6145 | If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine | |
6146 | generated for the generated object file will have static linkage. | |
6147 | ||
6148 | If your system uses @command{collect2} as the means of processing | |
6149 | constructors, then that program normally uses @command{nm} to scan | |
6150 | an object file for constructor functions to be called. | |
6151 | ||
6152 | On certain kinds of systems, you can define this macro to make | |
6153 | @command{collect2} work faster (and, in some cases, make it work at all): | |
6154 | ||
6155 | @defmac OBJECT_FORMAT_COFF | |
6156 | Define this macro if the system uses COFF (Common Object File Format) | |
6157 | object files, so that @command{collect2} can assume this format and scan | |
6158 | object files directly for dynamic constructor/destructor functions. | |
6159 | ||
6160 | This macro is effective only in a native compiler; @command{collect2} as | |
6161 | part of a cross compiler always uses @command{nm} for the target machine. | |
6162 | @end defmac | |
6163 | ||
6164 | @defmac REAL_NM_FILE_NAME | |
6165 | Define this macro as a C string constant containing the file name to use | |
6166 | to execute @command{nm}. The default is to search the path normally for | |
6167 | @command{nm}. | |
6168 | @end defmac | |
6169 | ||
6170 | @defmac NM_FLAGS | |
6171 | @command{collect2} calls @command{nm} to scan object files for static | |
6172 | constructors and destructors and LTO info. By default, @option{-n} is | |
6173 | passed. Define @code{NM_FLAGS} to a C string constant if other options | |
6174 | are needed to get the same output format as GNU @command{nm -n} | |
6175 | produces. | |
6176 | @end defmac | |
6177 | ||
6178 | If your system supports shared libraries and has a program to list the | |
6179 | dynamic dependencies of a given library or executable, you can define | |
6180 | these macros to enable support for running initialization and | |
6181 | termination functions in shared libraries: | |
6182 | ||
6183 | @defmac LDD_SUFFIX | |
6184 | Define this macro to a C string constant containing the name of the program | |
6185 | which lists dynamic dependencies, like @command{ldd} under SunOS 4. | |
6186 | @end defmac | |
6187 | ||
6188 | @defmac PARSE_LDD_OUTPUT (@var{ptr}) | |
6189 | Define this macro to be C code that extracts filenames from the output | |
6190 | of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable | |
6191 | of type @code{char *} that points to the beginning of a line of output | |
6192 | from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the | |
6193 | code must advance @var{ptr} to the beginning of the filename on that | |
6194 | line. Otherwise, it must set @var{ptr} to @code{NULL}. | |
6195 | @end defmac | |
6196 | ||
6197 | @defmac SHLIB_SUFFIX | |
6198 | Define this macro to a C string constant containing the default shared | |
6199 | library extension of the target (e.g., @samp{".so"}). @command{collect2} | |
6200 | strips version information after this suffix when generating global | |
6201 | constructor and destructor names. This define is only needed on targets | |
6202 | that use @command{collect2} to process constructors and destructors. | |
6203 | @end defmac | |
6204 | ||
6205 | @node Instruction Output | |
6206 | @subsection Output of Assembler Instructions | |
6207 | ||
6208 | @c prevent bad page break with this line | |
6209 | This describes assembler instruction output. | |
6210 | ||
6211 | @defmac REGISTER_NAMES | |
6212 | A C initializer containing the assembler's names for the machine | |
6213 | registers, each one as a C string constant. This is what translates | |
6214 | register numbers in the compiler into assembler language. | |
6215 | @end defmac | |
6216 | ||
6217 | @defmac ADDITIONAL_REGISTER_NAMES | |
6218 | If defined, a C initializer for an array of structures containing a name | |
6219 | and a register number. This macro defines additional names for hard | |
6220 | registers, thus allowing the @code{asm} option in declarations to refer | |
6221 | to registers using alternate names. | |
6222 | @end defmac | |
6223 | ||
6224 | @defmac OVERLAPPING_REGISTER_NAMES | |
6225 | If defined, a C initializer for an array of structures containing a | |
6226 | name, a register number and a count of the number of consecutive | |
6227 | machine registers the name overlaps. This macro defines additional | |
6228 | names for hard registers, thus allowing the @code{asm} option in | |
6229 | declarations to refer to registers using alternate names. Unlike | |
6230 | @code{ADDITIONAL_REGISTER_NAMES}, this macro should be used when the | |
6231 | register name implies multiple underlying registers. | |
6232 | ||
6233 | This macro should be used when it is important that a clobber in an | |
6234 | @code{asm} statement clobbers all the underlying values implied by the | |
6235 | register name. For example, on ARM, clobbering the double-precision | |
6236 | VFP register ``d0'' implies clobbering both single-precision registers | |
6237 | ``s0'' and ``s1''. | |
6238 | @end defmac | |
6239 | ||
6240 | @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr}) | |
6241 | Define this macro if you are using an unusual assembler that | |
6242 | requires different names for the machine instructions. | |
6243 | ||
6244 | The definition is a C statement or statements which output an | |
6245 | assembler instruction opcode to the stdio stream @var{stream}. The | |
6246 | macro-operand @var{ptr} is a variable of type @code{char *} which | |
6247 | points to the opcode name in its ``internal'' form---the form that is | |
6248 | written in the machine description. The definition should output the | |
6249 | opcode name to @var{stream}, performing any translation you desire, and | |
6250 | increment the variable @var{ptr} to point at the end of the opcode | |
6251 | so that it will not be output twice. | |
6252 | ||
6253 | In fact, your macro definition may process less than the entire opcode | |
6254 | name, or more than the opcode name; but if you want to process text | |
6255 | that includes @samp{%}-sequences to substitute operands, you must take | |
6256 | care of the substitution yourself. Just be sure to increment | |
6257 | @var{ptr} over whatever text should not be output normally. | |
6258 | ||
6259 | @findex recog_data.operand | |
6260 | If you need to look at the operand values, they can be found as the | |
6261 | elements of @code{recog_data.operand}. | |
6262 | ||
6263 | If the macro definition does nothing, the instruction is output | |
6264 | in the usual way. | |
6265 | @end defmac | |
6266 | ||
6267 | @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands}) | |
6268 | If defined, a C statement to be executed just prior to the output of | |
6269 | assembler code for @var{insn}, to modify the extracted operands so | |
6270 | they will be output differently. | |
6271 | ||
6272 | Here the argument @var{opvec} is the vector containing the operands | |
6273 | extracted from @var{insn}, and @var{noperands} is the number of | |
6274 | elements of the vector which contain meaningful data for this insn. | |
6275 | The contents of this vector are what will be used to convert the insn | |
6276 | template into assembler code, so you can change the assembler output | |
6277 | by changing the contents of the vector. | |
6278 | ||
6279 | This macro is useful when various assembler syntaxes share a single | |
6280 | file of instruction patterns; by defining this macro differently, you | |
6281 | can cause a large class of instructions to be output differently (such | |
6282 | as with rearranged operands). Naturally, variations in assembler | |
6283 | syntax affecting individual insn patterns ought to be handled by | |
6284 | writing conditional output routines in those patterns. | |
6285 | ||
6286 | If this macro is not defined, it is equivalent to a null statement. | |
6287 | @end defmac | |
6288 | ||
6289 | @hook TARGET_ASM_FINAL_POSTSCAN_INSN | |
6290 | ||
6291 | @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code}) | |
6292 | A C compound statement to output to stdio stream @var{stream} the | |
6293 | assembler syntax for an instruction operand @var{x}. @var{x} is an | |
6294 | RTL expression. | |
6295 | ||
6296 | @var{code} is a value that can be used to specify one of several ways | |
6297 | of printing the operand. It is used when identical operands must be | |
6298 | printed differently depending on the context. @var{code} comes from | |
6299 | the @samp{%} specification that was used to request printing of the | |
6300 | operand. If the specification was just @samp{%@var{digit}} then | |
6301 | @var{code} is 0; if the specification was @samp{%@var{ltr} | |
6302 | @var{digit}} then @var{code} is the ASCII code for @var{ltr}. | |
6303 | ||
6304 | @findex reg_names | |
6305 | If @var{x} is a register, this macro should print the register's name. | |
6306 | The names can be found in an array @code{reg_names} whose type is | |
6307 | @code{char *[]}. @code{reg_names} is initialized from | |
6308 | @code{REGISTER_NAMES}. | |
6309 | ||
6310 | When the machine description has a specification @samp{%@var{punct}} | |
6311 | (a @samp{%} followed by a punctuation character), this macro is called | |
6312 | with a null pointer for @var{x} and the punctuation character for | |
6313 | @var{code}. | |
6314 | @end defmac | |
6315 | ||
6316 | @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code}) | |
6317 | A C expression which evaluates to true if @var{code} is a valid | |
6318 | punctuation character for use in the @code{PRINT_OPERAND} macro. If | |
6319 | @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no | |
6320 | punctuation characters (except for the standard one, @samp{%}) are used | |
6321 | in this way. | |
6322 | @end defmac | |
6323 | ||
6324 | @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x}) | |
6325 | A C compound statement to output to stdio stream @var{stream} the | |
6326 | assembler syntax for an instruction operand that is a memory reference | |
6327 | whose address is @var{x}. @var{x} is an RTL expression. | |
6328 | ||
6329 | @cindex @code{TARGET_ENCODE_SECTION_INFO} usage | |
6330 | On some machines, the syntax for a symbolic address depends on the | |
6331 | section that the address refers to. On these machines, define the hook | |
6332 | @code{TARGET_ENCODE_SECTION_INFO} to store the information into the | |
6333 | @code{symbol_ref}, and then check for it here. @xref{Assembler | |
6334 | Format}. | |
6335 | @end defmac | |
6336 | ||
6337 | @findex dbr_sequence_length | |
6338 | @defmac DBR_OUTPUT_SEQEND (@var{file}) | |
6339 | A C statement, to be executed after all slot-filler instructions have | |
6340 | been output. If necessary, call @code{dbr_sequence_length} to | |
6341 | determine the number of slots filled in a sequence (zero if not | |
6342 | currently outputting a sequence), to decide how many no-ops to output, | |
6343 | or whatever. | |
6344 | ||
6345 | Don't define this macro if it has nothing to do, but it is helpful in | |
6346 | reading assembly output if the extent of the delay sequence is made | |
6347 | explicit (e.g.@: with white space). | |
6348 | @end defmac | |
6349 | ||
6350 | @findex final_sequence | |
6351 | Note that output routines for instructions with delay slots must be | |
6352 | prepared to deal with not being output as part of a sequence | |
6353 | (i.e.@: when the scheduling pass is not run, or when no slot fillers could be | |
6354 | found.) The variable @code{final_sequence} is null when not | |
6355 | processing a sequence, otherwise it contains the @code{sequence} rtx | |
6356 | being output. | |
6357 | ||
6358 | @findex asm_fprintf | |
6359 | @defmac REGISTER_PREFIX | |
6360 | @defmacx LOCAL_LABEL_PREFIX | |
6361 | @defmacx USER_LABEL_PREFIX | |
6362 | @defmacx IMMEDIATE_PREFIX | |
6363 | If defined, C string expressions to be used for the @samp{%R}, @samp{%L}, | |
6364 | @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see | |
6365 | @file{final.cc}). These are useful when a single @file{md} file must | |
6366 | support multiple assembler formats. In that case, the various @file{tm.h} | |
6367 | files can define these macros differently. | |
6368 | @end defmac | |
6369 | ||
6370 | @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format}) | |
6371 | If defined this macro should expand to a series of @code{case} | |
6372 | statements which will be parsed inside the @code{switch} statement of | |
6373 | the @code{asm_fprintf} function. This allows targets to define extra | |
6374 | printf formats which may useful when generating their assembler | |
6375 | statements. Note that uppercase letters are reserved for future | |
6376 | generic extensions to asm_fprintf, and so are not available to target | |
6377 | specific code. The output file is given by the parameter @var{file}. | |
6378 | The varargs input pointer is @var{argptr} and the rest of the format | |
6379 | string, starting the character after the one that is being switched | |
6380 | upon, is pointed to by @var{format}. | |
6381 | @end defmac | |
6382 | ||
6383 | @defmac ASSEMBLER_DIALECT | |
6384 | If your target supports multiple dialects of assembler language (such as | |
6385 | different opcodes), define this macro as a C expression that gives the | |
6386 | numeric index of the assembler language dialect to use, with zero as the | |
6387 | first variant. | |
6388 | ||
6389 | If this macro is defined, you may use constructs of the form | |
6390 | @smallexample | |
6391 | @samp{@{option0|option1|option2@dots{}@}} | |
6392 | @end smallexample | |
6393 | @noindent | |
6394 | in the output templates of patterns (@pxref{Output Template}) or in the | |
6395 | first argument of @code{asm_fprintf}. This construct outputs | |
6396 | @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of | |
6397 | @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters | |
6398 | within these strings retain their usual meaning. If there are fewer | |
6399 | alternatives within the braces than the value of | |
6400 | @code{ASSEMBLER_DIALECT}, the construct outputs nothing. If it's needed | |
6401 | to print curly braces or @samp{|} character in assembler output directly, | |
6402 | @samp{%@{}, @samp{%@}} and @samp{%|} can be used. | |
6403 | ||
6404 | If you do not define this macro, the characters @samp{@{}, @samp{|} and | |
6405 | @samp{@}} do not have any special meaning when used in templates or | |
6406 | operands to @code{asm_fprintf}. | |
6407 | ||
6408 | Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX}, | |
6409 | @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express | |
6410 | the variations in assembler language syntax with that mechanism. Define | |
6411 | @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax | |
6412 | if the syntax variant are larger and involve such things as different | |
6413 | opcodes or operand order. | |
6414 | @end defmac | |
6415 | ||
6416 | @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno}) | |
6417 | A C expression to output to @var{stream} some assembler code | |
6418 | which will push hard register number @var{regno} onto the stack. | |
6419 | The code need not be optimal, since this macro is used only when | |
6420 | profiling. | |
6421 | @end defmac | |
6422 | ||
6423 | @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno}) | |
6424 | A C expression to output to @var{stream} some assembler code | |
6425 | which will pop hard register number @var{regno} off of the stack. | |
6426 | The code need not be optimal, since this macro is used only when | |
6427 | profiling. | |
6428 | @end defmac | |
6429 | ||
6430 | @node Dispatch Tables | |
6431 | @subsection Output of Dispatch Tables | |
6432 | ||
6433 | @c prevent bad page break with this line | |
6434 | This concerns dispatch tables. | |
6435 | ||
6436 | @cindex dispatch table | |
6437 | @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel}) | |
6438 | A C statement to output to the stdio stream @var{stream} an assembler | |
6439 | pseudo-instruction to generate a difference between two labels. | |
6440 | @var{value} and @var{rel} are the numbers of two internal labels. The | |
6441 | definitions of these labels are output using | |
6442 | @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same | |
6443 | way here. For example, | |
6444 | ||
6445 | @smallexample | |
6446 | fprintf (@var{stream}, "\t.word L%d-L%d\n", | |
6447 | @var{value}, @var{rel}) | |
6448 | @end smallexample | |
6449 | ||
6450 | You must provide this macro on machines where the addresses in a | |
6451 | dispatch table are relative to the table's own address. If defined, GCC | |
6452 | will also use this macro on all machines when producing PIC@. | |
6453 | @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the | |
6454 | mode and flags can be read. | |
6455 | @end defmac | |
6456 | ||
6457 | @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value}) | |
6458 | This macro should be provided on machines where the addresses | |
6459 | in a dispatch table are absolute. | |
6460 | ||
6461 | The definition should be a C statement to output to the stdio stream | |
6462 | @var{stream} an assembler pseudo-instruction to generate a reference to | |
6463 | a label. @var{value} is the number of an internal label whose | |
6464 | definition is output using @code{(*targetm.asm_out.internal_label)}. | |
6465 | For example, | |
6466 | ||
6467 | @smallexample | |
6468 | fprintf (@var{stream}, "\t.word L%d\n", @var{value}) | |
6469 | @end smallexample | |
6470 | @end defmac | |
6471 | ||
6472 | @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table}) | |
6473 | Define this if the label before a jump-table needs to be output | |
6474 | specially. The first three arguments are the same as for | |
6475 | @code{(*targetm.asm_out.internal_label)}; the fourth argument is the | |
6476 | jump-table which follows (a @code{jump_table_data} containing an | |
6477 | @code{addr_vec} or @code{addr_diff_vec}). | |
6478 | ||
6479 | This feature is used on system V to output a @code{swbeg} statement | |
6480 | for the table. | |
6481 | ||
6482 | If this macro is not defined, these labels are output with | |
6483 | @code{(*targetm.asm_out.internal_label)}. | |
6484 | @end defmac | |
6485 | ||
6486 | @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table}) | |
6487 | Define this if something special must be output at the end of a | |
6488 | jump-table. The definition should be a C statement to be executed | |
6489 | after the assembler code for the table is written. It should write | |
6490 | the appropriate code to stdio stream @var{stream}. The argument | |
6491 | @var{table} is the jump-table insn, and @var{num} is the label-number | |
6492 | of the preceding label. | |
6493 | ||
6494 | If this macro is not defined, nothing special is output at the end of | |
6495 | the jump-table. | |
6496 | @end defmac | |
6497 | ||
6498 | @hook TARGET_ASM_POST_CFI_STARTPROC | |
6499 | ||
6500 | @hook TARGET_ASM_EMIT_UNWIND_LABEL | |
6501 | ||
6502 | @hook TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL | |
6503 | ||
6504 | @hook TARGET_ASM_EMIT_EXCEPT_PERSONALITY | |
6505 | ||
6506 | @hook TARGET_ASM_UNWIND_EMIT | |
6507 | ||
6508 | @hook TARGET_ASM_MAKE_EH_SYMBOL_INDIRECT | |
6509 | ||
6510 | @hook TARGET_ASM_UNWIND_EMIT_BEFORE_INSN | |
6511 | ||
6512 | @hook TARGET_ASM_SHOULD_RESTORE_CFA_STATE | |
6513 | ||
6514 | @node Exception Region Output | |
6515 | @subsection Assembler Commands for Exception Regions | |
6516 | ||
6517 | @c prevent bad page break with this line | |
6518 | ||
6519 | This describes commands marking the start and the end of an exception | |
6520 | region. | |
6521 | ||
6522 | @defmac EH_FRAME_SECTION_NAME | |
6523 | If defined, a C string constant for the name of the section containing | |
6524 | exception handling frame unwind information. If not defined, GCC will | |
6525 | provide a default definition if the target supports named sections. | |
6526 | @file{crtstuff.c} uses this macro to switch to the appropriate section. | |
6527 | ||
6528 | You should define this symbol if your target supports DWARF 2 frame | |
6529 | unwind information and the default definition does not work. | |
6530 | @end defmac | |
6531 | ||
6532 | @defmac EH_FRAME_THROUGH_COLLECT2 | |
6533 | If defined, DWARF 2 frame unwind information will identified by | |
6534 | specially named labels. The collect2 process will locate these | |
6535 | labels and generate code to register the frames. | |
6536 | ||
6537 | This might be necessary, for instance, if the system linker will not | |
6538 | place the eh_frames in-between the sentinals from @file{crtstuff.c}, | |
6539 | or if the system linker does garbage collection and sections cannot | |
6540 | be marked as not to be collected. | |
6541 | @end defmac | |
6542 | ||
6543 | @defmac EH_TABLES_CAN_BE_READ_ONLY | |
6544 | Define this macro to 1 if your target is such that no frame unwind | |
6545 | information encoding used with non-PIC code will ever require a | |
6546 | runtime relocation, but the linker may not support merging read-only | |
6547 | and read-write sections into a single read-write section. | |
6548 | @end defmac | |
6549 | ||
6550 | @defmac MASK_RETURN_ADDR | |
6551 | An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so | |
6552 | that it does not contain any extraneous set bits in it. | |
6553 | @end defmac | |
6554 | ||
6555 | @defmac DWARF2_UNWIND_INFO | |
6556 | Define this macro to 0 if your target supports DWARF 2 frame unwind | |
6557 | information, but it does not yet work with exception handling. | |
6558 | Otherwise, if your target supports this information (if it defines | |
6559 | @code{INCOMING_RETURN_ADDR_RTX} and @code{OBJECT_FORMAT_ELF}), | |
6560 | GCC will provide a default definition of 1. | |
6561 | @end defmac | |
6562 | ||
6563 | @hook TARGET_EXCEPT_UNWIND_INFO | |
6564 | This hook defines the mechanism that will be used for exception handling | |
6565 | by the target. If the target has ABI specified unwind tables, the hook | |
6566 | should return @code{UI_TARGET}. If the target is to use the | |
6567 | @code{setjmp}/@code{longjmp}-based exception handling scheme, the hook | |
6568 | should return @code{UI_SJLJ}. If the target supports DWARF 2 frame unwind | |
6569 | information, the hook should return @code{UI_DWARF2}. | |
6570 | ||
6571 | A target may, if exceptions are disabled, choose to return @code{UI_NONE}. | |
6572 | This may end up simplifying other parts of target-specific code. The | |
6573 | default implementation of this hook never returns @code{UI_NONE}. | |
6574 | ||
6575 | Note that the value returned by this hook should be constant. It should | |
6576 | not depend on anything except the command-line switches described by | |
6577 | @var{opts}. In particular, the | |
6578 | setting @code{UI_SJLJ} must be fixed at compiler start-up as C pre-processor | |
6579 | macros and builtin functions related to exception handling are set up | |
6580 | depending on this setting. | |
6581 | ||
6582 | The default implementation of the hook first honors the | |
6583 | @option{--enable-sjlj-exceptions} configure option, then | |
6584 | @code{DWARF2_UNWIND_INFO}, and finally defaults to @code{UI_SJLJ}. If | |
6585 | @code{DWARF2_UNWIND_INFO} depends on command-line options, the target | |
6586 | must define this hook so that @var{opts} is used correctly. | |
6587 | @end deftypefn | |
6588 | ||
6589 | @hook TARGET_UNWIND_TABLES_DEFAULT | |
6590 | This variable should be set to @code{true} if the target ABI requires unwinding | |
6591 | tables even when exceptions are not used. It must not be modified by | |
6592 | command-line option processing. | |
6593 | @end deftypevr | |
6594 | ||
6595 | @defmac DONT_USE_BUILTIN_SETJMP | |
6596 | Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme | |
6597 | should use the @code{setjmp}/@code{longjmp} functions from the C library | |
6598 | instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery. | |
6599 | @end defmac | |
6600 | ||
6601 | @defmac JMP_BUF_SIZE | |
6602 | This macro has no effect unless @code{DONT_USE_BUILTIN_SETJMP} is also | |
6603 | defined. Define this macro if the default size of @code{jmp_buf} buffer | |
6604 | for the @code{setjmp}/@code{longjmp}-based exception handling mechanism | |
6605 | is not large enough, or if it is much too large. | |
6606 | The default size is @code{FIRST_PSEUDO_REGISTER * sizeof(void *)}. | |
6607 | @end defmac | |
6608 | ||
6609 | @defmac DWARF_CIE_DATA_ALIGNMENT | |
6610 | This macro need only be defined if the target might save registers in the | |
6611 | function prologue at an offset to the stack pointer that is not aligned to | |
6612 | @code{UNITS_PER_WORD}. The definition should be the negative minimum | |
6613 | alignment if @code{STACK_GROWS_DOWNWARD} is true, and the positive | |
6614 | minimum alignment otherwise. @xref{DWARF}. Only applicable if | |
6615 | the target supports DWARF 2 frame unwind information. | |
6616 | @end defmac | |
6617 | ||
6618 | @hook TARGET_TERMINATE_DW2_EH_FRAME_INFO | |
6619 | ||
6620 | @hook TARGET_DWARF_REGISTER_SPAN | |
6621 | ||
6622 | @hook TARGET_DWARF_FRAME_REG_MODE | |
6623 | ||
9e1c71ba ML |
6624 | @hook TARGET_OUTPUT_CFI_DIRECTIVE |
6625 | ||
6626 | @hook TARGET_DW_CFI_OPRND1_DESC | |
6627 | ||
d77de738 ML |
6628 | @hook TARGET_INIT_DWARF_REG_SIZES_EXTRA |
6629 | ||
6630 | @hook TARGET_ASM_TTYPE | |
6631 | ||
6632 | @hook TARGET_ARM_EABI_UNWINDER | |
6633 | ||
6634 | @node Alignment Output | |
6635 | @subsection Assembler Commands for Alignment | |
6636 | ||
6637 | @c prevent bad page break with this line | |
6638 | This describes commands for alignment. | |
6639 | ||
6640 | @defmac JUMP_ALIGN (@var{label}) | |
6641 | The alignment (log base 2) to put in front of @var{label}, which is | |
6642 | a common destination of jumps and has no fallthru incoming edge. | |
6643 | ||
6644 | This macro need not be defined if you don't want any special alignment | |
6645 | to be done at such a time. Most machine descriptions do not currently | |
6646 | define the macro. | |
6647 | ||
6648 | Unless it's necessary to inspect the @var{label} parameter, it is better | |
6649 | to set the variable @var{align_jumps} in the target's | |
6650 | @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's | |
6651 | selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation. | |
6652 | @end defmac | |
6653 | ||
6654 | @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label}) | |
6655 | The alignment (log base 2) to put in front of @var{label}, which follows | |
6656 | a @code{BARRIER}. | |
6657 | ||
6658 | This macro need not be defined if you don't want any special alignment | |
6659 | to be done at such a time. Most machine descriptions do not currently | |
6660 | define the macro. | |
6661 | @end defmac | |
6662 | ||
6663 | @defmac LOOP_ALIGN (@var{label}) | |
6664 | The alignment (log base 2) to put in front of @var{label} that heads | |
6665 | a frequently executed basic block (usually the header of a loop). | |
6666 | ||
6667 | This macro need not be defined if you don't want any special alignment | |
6668 | to be done at such a time. Most machine descriptions do not currently | |
6669 | define the macro. | |
6670 | ||
6671 | Unless it's necessary to inspect the @var{label} parameter, it is better | |
6672 | to set the variable @code{align_loops} in the target's | |
6673 | @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's | |
6674 | selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation. | |
6675 | @end defmac | |
6676 | ||
6677 | @defmac LABEL_ALIGN (@var{label}) | |
6678 | The alignment (log base 2) to put in front of @var{label}. | |
6679 | If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment, | |
6680 | the maximum of the specified values is used. | |
6681 | ||
6682 | Unless it's necessary to inspect the @var{label} parameter, it is better | |
6683 | to set the variable @code{align_labels} in the target's | |
6684 | @code{TARGET_OPTION_OVERRIDE}. Otherwise, you should try to honor the user's | |
6685 | selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation. | |
6686 | @end defmac | |
6687 | ||
6688 | @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes}) | |
6689 | A C statement to output to the stdio stream @var{stream} an assembler | |
6690 | instruction to advance the location counter by @var{nbytes} bytes. | |
6691 | Those bytes should be zero when loaded. @var{nbytes} will be a C | |
6692 | expression of type @code{unsigned HOST_WIDE_INT}. | |
6693 | @end defmac | |
6694 | ||
6695 | @defmac ASM_NO_SKIP_IN_TEXT | |
6696 | Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the | |
6697 | text section because it fails to put zeros in the bytes that are skipped. | |
6698 | This is true on many Unix systems, where the pseudo--op to skip bytes | |
6699 | produces no-op instructions rather than zeros when used in the text | |
6700 | section. | |
6701 | @end defmac | |
6702 | ||
6703 | @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power}) | |
6704 | A C statement to output to the stdio stream @var{stream} an assembler | |
6705 | command to advance the location counter to a multiple of 2 to the | |
6706 | @var{power} bytes. @var{power} will be a C expression of type @code{int}. | |
6707 | @end defmac | |
6708 | ||
6709 | @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power}) | |
6710 | Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used | |
6711 | for padding, if necessary. | |
6712 | @end defmac | |
6713 | ||
6714 | @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip}) | |
6715 | A C statement to output to the stdio stream @var{stream} an assembler | |
6716 | command to advance the location counter to a multiple of 2 to the | |
6717 | @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to | |
6718 | satisfy the alignment request. @var{power} and @var{max_skip} will be | |
6719 | a C expression of type @code{int}. | |
6720 | @end defmac | |
6721 | ||
6722 | @need 3000 | |
6723 | @node Debugging Info | |
6724 | @section Controlling Debugging Information Format | |
6725 | ||
6726 | @c prevent bad page break with this line | |
6727 | This describes how to specify debugging information. | |
6728 | ||
6729 | @menu | |
6730 | * All Debuggers:: Macros that affect all debugging formats uniformly. | |
6731 | * DWARF:: Macros for DWARF format. | |
6732 | * VMS Debug:: Macros for VMS debug format. | |
6733 | * CTF Debug:: Macros for CTF debug format. | |
6734 | * BTF Debug:: Macros for BTF debug format. | |
6735 | @end menu | |
6736 | ||
6737 | @node All Debuggers | |
6738 | @subsection Macros Affecting All Debugging Formats | |
6739 | ||
6740 | @c prevent bad page break with this line | |
6741 | These macros affect all debugging formats. | |
6742 | ||
6743 | @defmac DEBUGGER_REGNO (@var{regno}) | |
6744 | A C expression that returns the debugger register number for the compiler | |
6745 | register number @var{regno}. In the default macro provided, the value | |
6746 | of this expression will be @var{regno} itself. But sometimes there are | |
6747 | some registers that the compiler knows about and debugger does not, or vice | |
6748 | versa. In such cases, some register may need to have one number in the | |
6749 | compiler and another for debugger@. | |
6750 | ||
6751 | If two registers have consecutive numbers inside GCC, and they can be | |
6752 | used as a pair to hold a multiword value, then they @emph{must} have | |
6753 | consecutive numbers after renumbering with @code{DEBUGGER_REGNO}. | |
6754 | Otherwise, debuggers will be unable to access such a pair, because they | |
6755 | expect register pairs to be consecutive in their own numbering scheme. | |
6756 | ||
6757 | If you find yourself defining @code{DEBUGGER_REGNO} in way that | |
6758 | does not preserve register pairs, then what you must do instead is | |
6759 | redefine the actual register numbering scheme. | |
6760 | @end defmac | |
6761 | ||
6762 | @defmac DEBUGGER_AUTO_OFFSET (@var{x}) | |
6763 | A C expression that returns the integer offset value for an automatic | |
6764 | variable having address @var{x} (an RTL expression). The default | |
6765 | computation assumes that @var{x} is based on the frame-pointer and | |
6766 | gives the offset from the frame-pointer. This is required for targets | |
6767 | that produce debugging output for debugger and allow the frame-pointer to be | |
6768 | eliminated when the @option{-g} option is used. | |
6769 | @end defmac | |
6770 | ||
6771 | @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x}) | |
6772 | A C expression that returns the integer offset value for an argument | |
6773 | having address @var{x} (an RTL expression). The nominal offset is | |
6774 | @var{offset}. | |
6775 | @end defmac | |
6776 | ||
6777 | @defmac PREFERRED_DEBUGGING_TYPE | |
6778 | A C expression that returns the type of debugging output GCC should | |
6779 | produce when the user specifies just @option{-g}. Define | |
6780 | this if you have arranged for GCC to support more than one format of | |
6781 | debugging output. Currently, the allowable values are | |
6782 | @code{DWARF2_DEBUG}, @code{VMS_DEBUG}, | |
6783 | and @code{VMS_AND_DWARF2_DEBUG}. | |
6784 | ||
6785 | When the user specifies @option{-ggdb}, GCC normally also uses the | |
6786 | value of this macro to select the debugging output format, but with two | |
6787 | exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the | |
6788 | value @code{DWARF2_DEBUG}. | |
6789 | ||
6790 | The value of this macro only affects the default debugging output; the | |
6791 | user can always get a specific type of output by using @option{-gdwarf-2}, | |
6792 | or @option{-gvms}. | |
6793 | @end defmac | |
6794 | ||
6795 | @defmac DEFAULT_GDB_EXTENSIONS | |
6796 | Define this macro to control whether GCC should by default generate | |
6797 | GDB's extended version of debugging information. If you don't define the | |
6798 | macro, the default is 1: always generate the extended information | |
6799 | if there is any occasion to. | |
6800 | @end defmac | |
6801 | ||
6802 | @need 2000 | |
6803 | @node DWARF | |
6804 | @subsection Macros for DWARF Output | |
6805 | ||
6806 | @c prevent bad page break with this line | |
6807 | Here are macros for DWARF output. | |
6808 | ||
6809 | @defmac DWARF2_DEBUGGING_INFO | |
6810 | Define this macro if GCC should produce dwarf version 2 format | |
6811 | debugging output in response to the @option{-g} option. | |
6812 | ||
6813 | To support optional call frame debugging information, you must also | |
6814 | define @code{INCOMING_RETURN_ADDR_RTX} and either set | |
6815 | @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the | |
6816 | prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save} | |
6817 | as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't. | |
6818 | @end defmac | |
6819 | ||
6820 | @hook TARGET_DWARF_CALLING_CONVENTION | |
6821 | ||
6822 | @defmac DWARF2_FRAME_INFO | |
6823 | Define this macro to a nonzero value if GCC should always output | |
6824 | Dwarf 2 frame information. If @code{TARGET_EXCEPT_UNWIND_INFO} | |
6825 | (@pxref{Exception Region Output}) returns @code{UI_DWARF2}, and | |
6826 | exceptions are enabled, GCC will output this information not matter | |
6827 | how you define @code{DWARF2_FRAME_INFO}. | |
6828 | @end defmac | |
6829 | ||
6830 | @hook TARGET_DEBUG_UNWIND_INFO | |
6831 | ||
6832 | @defmac DWARF2_ASM_LINE_DEBUG_INFO | |
6833 | Define this macro to be a nonzero value if the assembler can generate Dwarf 2 | |
6834 | line debug info sections. This will result in much more compact line number | |
6835 | tables, and hence is desirable if it works. | |
6836 | @end defmac | |
6837 | ||
6838 | @defmac DWARF2_ASM_VIEW_DEBUG_INFO | |
6839 | Define this macro to be a nonzero value if the assembler supports view | |
6840 | assignment and verification in @code{.loc}. If it does not, but the | |
6841 | user enables location views, the compiler may have to fallback to | |
6842 | internal line number tables. | |
6843 | @end defmac | |
6844 | ||
6845 | @hook TARGET_RESET_LOCATION_VIEW | |
6846 | ||
6847 | @hook TARGET_WANT_DEBUG_PUB_SECTIONS | |
6848 | ||
6849 | @hook TARGET_DELAY_SCHED2 | |
6850 | ||
6851 | @hook TARGET_DELAY_VARTRACK | |
6852 | ||
6853 | @hook TARGET_NO_REGISTER_ALLOCATION | |
6854 | ||
6855 | @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2}) | |
6856 | A C statement to issue assembly directives that create a difference | |
6857 | @var{lab1} minus @var{lab2}, using an integer of the given @var{size}. | |
6858 | @end defmac | |
6859 | ||
6860 | @defmac ASM_OUTPUT_DWARF_VMS_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2}) | |
6861 | A C statement to issue assembly directives that create a difference | |
6862 | between the two given labels in system defined units, e.g.@: instruction | |
6863 | slots on IA64 VMS, using an integer of the given size. | |
6864 | @end defmac | |
6865 | ||
6866 | @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{offset}, @var{section}) | |
6867 | A C statement to issue assembly directives that create a | |
6868 | section-relative reference to the given @var{label} plus @var{offset}, using | |
6869 | an integer of the given @var{size}. The label is known to be defined in the | |
6870 | given @var{section}. | |
6871 | @end defmac | |
6872 | ||
6873 | @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label}) | |
6874 | A C statement to issue assembly directives that create a self-relative | |
6875 | reference to the given @var{label}, using an integer of the given @var{size}. | |
6876 | @end defmac | |
6877 | ||
6878 | @defmac ASM_OUTPUT_DWARF_DATAREL (@var{stream}, @var{size}, @var{label}) | |
6879 | A C statement to issue assembly directives that create a reference to the | |
6880 | given @var{label} relative to the dbase, using an integer of the given @var{size}. | |
6881 | @end defmac | |
6882 | ||
6883 | @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label}) | |
6884 | A C statement to issue assembly directives that create a reference to | |
6885 | the DWARF table identifier @var{label} from the current section. This | |
6886 | is used on some systems to avoid garbage collecting a DWARF table which | |
6887 | is referenced by a function. | |
6888 | @end defmac | |
6889 | ||
6890 | @hook TARGET_ASM_OUTPUT_DWARF_DTPREL | |
6891 | ||
6892 | @need 2000 | |
6893 | @node VMS Debug | |
6894 | @subsection Macros for VMS Debug Format | |
6895 | ||
6896 | @c prevent bad page break with this line | |
6897 | Here are macros for VMS debug format. | |
6898 | ||
6899 | @defmac VMS_DEBUGGING_INFO | |
6900 | Define this macro if GCC should produce debugging output for VMS | |
6901 | in response to the @option{-g} option. The default behavior for VMS | |
6902 | is to generate minimal debug info for a traceback in the absence of | |
6903 | @option{-g} unless explicitly overridden with @option{-g0}. This | |
6904 | behavior is controlled by @code{TARGET_OPTION_OPTIMIZATION} and | |
6905 | @code{TARGET_OPTION_OVERRIDE}. | |
6906 | @end defmac | |
6907 | ||
6908 | @need 2000 | |
6909 | @node CTF Debug | |
6910 | @subsection Macros for CTF Debug Format | |
6911 | ||
6912 | @c prevent bad page break with this line | |
6913 | Here are macros for CTF debug format. | |
6914 | ||
6915 | @defmac CTF_DEBUGGING_INFO | |
6916 | Define this macro if GCC should produce debugging output in CTF debug | |
6917 | format in response to the @option{-gctf} option. | |
6918 | @end defmac | |
6919 | ||
6920 | @need 2000 | |
6921 | @node BTF Debug | |
6922 | @subsection Macros for BTF Debug Format | |
6923 | ||
6924 | @c prevent bad page break with this line | |
6925 | Here are macros for BTF debug format. | |
6926 | ||
6927 | @defmac BTF_DEBUGGING_INFO | |
6928 | Define this macro if GCC should produce debugging output in BTF debug | |
6929 | format in response to the @option{-gbtf} option. | |
6930 | @end defmac | |
6931 | ||
6932 | @node Floating Point | |
6933 | @section Cross Compilation and Floating Point | |
6934 | @cindex cross compilation and floating point | |
6935 | @cindex floating point and cross compilation | |
6936 | ||
6937 | While all modern machines use twos-complement representation for integers, | |
6938 | there are a variety of representations for floating point numbers. This | |
6939 | means that in a cross-compiler the representation of floating point numbers | |
6940 | in the compiled program may be different from that used in the machine | |
6941 | doing the compilation. | |
6942 | ||
6943 | Because different representation systems may offer different amounts of | |
6944 | range and precision, all floating point constants must be represented in | |
6945 | the target machine's format. Therefore, the cross compiler cannot | |
6946 | safely use the host machine's floating point arithmetic; it must emulate | |
6947 | the target's arithmetic. To ensure consistency, GCC always uses | |
6948 | emulation to work with floating point values, even when the host and | |
6949 | target floating point formats are identical. | |
6950 | ||
6951 | The following macros are provided by @file{real.h} for the compiler to | |
6952 | use. All parts of the compiler which generate or optimize | |
6953 | floating-point calculations must use these macros. They may evaluate | |
6954 | their operands more than once, so operands must not have side effects. | |
6955 | ||
6956 | @defmac REAL_VALUE_TYPE | |
6957 | The C data type to be used to hold a floating point value in the target | |
6958 | machine's format. Typically this is a @code{struct} containing an | |
6959 | array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque | |
6960 | quantity. | |
6961 | @end defmac | |
6962 | ||
6963 | @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x}) | |
6964 | Truncates @var{x} to a signed integer, rounding toward zero. | |
6965 | @end deftypefn | |
6966 | ||
6967 | @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x}) | |
6968 | Truncates @var{x} to an unsigned integer, rounding toward zero. If | |
6969 | @var{x} is negative, returns zero. | |
6970 | @end deftypefn | |
6971 | ||
6972 | @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, machine_mode @var{mode}) | |
6973 | Converts @var{string} into a floating point number in the target machine's | |
6974 | representation for mode @var{mode}. This routine can handle both | |
6975 | decimal and hexadecimal floating point constants, using the syntax | |
6976 | defined by the C language for both. | |
6977 | @end deftypefn | |
6978 | ||
6979 | @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x}) | |
6980 | Returns 1 if @var{x} is negative (including negative zero), 0 otherwise. | |
6981 | @end deftypefn | |
6982 | ||
6983 | @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x}) | |
6984 | Determines whether @var{x} represents infinity (positive or negative). | |
6985 | @end deftypefn | |
6986 | ||
6987 | @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x}) | |
6988 | Determines whether @var{x} represents a ``NaN'' (not-a-number). | |
6989 | @end deftypefn | |
6990 | ||
6991 | @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x}) | |
6992 | Returns the negative of the floating point value @var{x}. | |
6993 | @end deftypefn | |
6994 | ||
6995 | @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x}) | |
6996 | Returns the absolute value of @var{x}. | |
6997 | @end deftypefn | |
6998 | ||
6999 | @node Mode Switching | |
7000 | @section Mode Switching Instructions | |
7001 | @cindex mode switching | |
7002 | The following macros control mode switching optimizations: | |
7003 | ||
7004 | @defmac OPTIMIZE_MODE_SWITCHING (@var{entity}) | |
7005 | Define this macro if the port needs extra instructions inserted for mode | |
8479a375 | 7006 | switching. |
d77de738 ML |
7007 | |
7008 | For an example, the SH4 can perform both single and double precision | |
7009 | floating point operations, but to perform a single precision operation, | |
7010 | the FPSCR PR bit has to be cleared, while for a double precision | |
7011 | operation, this bit has to be set. Changing the PR bit requires a general | |
7012 | purpose register as a scratch register, hence these FPSCR sets have to | |
7013 | be inserted before reload, i.e.@: you cannot put this into instruction emitting | |
7014 | or @code{TARGET_MACHINE_DEPENDENT_REORG}. | |
7015 | ||
8479a375 RS |
7016 | You can have multiple entities that are mode-switched, some of which might |
7017 | only be needed conditionally. The entities are identified by their index | |
7018 | into the @code{NUM_MODES_FOR_MODE_SWITCHING} initializer, with the length | |
7019 | of the initializer determining the number of entities. | |
7020 | ||
7021 | @code{OPTIMIZE_MODE_SWITCHING} should return nonzero for any @var{entity} | |
7022 | that needs mode-switching. | |
7023 | ||
d77de738 ML |
7024 | If you define this macro, you also have to define |
7025 | @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{TARGET_MODE_NEEDED}, | |
7026 | @code{TARGET_MODE_PRIORITY} and @code{TARGET_MODE_EMIT}. | |
8479a375 | 7027 | The other macros in this section are optional. |
d77de738 ML |
7028 | @end defmac |
7029 | ||
7030 | @defmac NUM_MODES_FOR_MODE_SWITCHING | |
7031 | If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as | |
7032 | initializer for an array of integers. Each initializer element | |
7033 | N refers to an entity that needs mode switching, and specifies the number | |
8479a375 RS |
7034 | of different modes that are defined for that entity. |
7035 | The position of the element in the initializer---starting counting at | |
d77de738 ML |
7036 | zero---determines the integer that is used to refer to the mode-switched |
7037 | entity in question. | |
8479a375 RS |
7038 | Modes are represented as numbers 0 @dots{} N @minus{} 1. |
7039 | In mode arguments and return values, N either represents an unknown | |
7040 | mode or ``no mode'', depending on context. | |
d77de738 ML |
7041 | @end defmac |
7042 | ||
7043 | @hook TARGET_MODE_EMIT | |
7044 | ||
7045 | @hook TARGET_MODE_NEEDED | |
7046 | ||
7047 | @hook TARGET_MODE_AFTER | |
7048 | ||
493b0038 RS |
7049 | @hook TARGET_MODE_CONFLUENCE |
7050 | ||
fc8458e2 RS |
7051 | @hook TARGET_MODE_BACKPROP |
7052 | ||
d77de738 ML |
7053 | @hook TARGET_MODE_ENTRY |
7054 | ||
7055 | @hook TARGET_MODE_EXIT | |
7056 | ||
4b803fbf RS |
7057 | @hook TARGET_MODE_EH_HANDLER |
7058 | ||
d77de738 ML |
7059 | @hook TARGET_MODE_PRIORITY |
7060 | ||
7061 | @node Target Attributes | |
7062 | @section Defining target-specific uses of @code{__attribute__} | |
7063 | @cindex target attributes | |
7064 | @cindex machine attributes | |
7065 | @cindex attributes, target-specific | |
7066 | ||
7067 | Target-specific attributes may be defined for functions, data and types. | |
7068 | These are described using the following target hooks; they also need to | |
7069 | be documented in @file{extend.texi}. | |
7070 | ||
7071 | @hook TARGET_ATTRIBUTE_TABLE | |
7072 | ||
7073 | @hook TARGET_ATTRIBUTE_TAKES_IDENTIFIER_P | |
7074 | ||
7075 | @hook TARGET_COMP_TYPE_ATTRIBUTES | |
7076 | ||
7077 | @hook TARGET_SET_DEFAULT_TYPE_ATTRIBUTES | |
7078 | ||
7079 | @hook TARGET_MERGE_TYPE_ATTRIBUTES | |
7080 | ||
7081 | @hook TARGET_MERGE_DECL_ATTRIBUTES | |
7082 | ||
7083 | @hook TARGET_VALID_DLLIMPORT_ATTRIBUTE_P | |
7084 | ||
7085 | @defmac TARGET_DECLSPEC | |
7086 | Define this macro to a nonzero value if you want to treat | |
7087 | @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By | |
7088 | default, this behavior is enabled only for targets that define | |
7089 | @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation | |
7090 | of @code{__declspec} is via a built-in macro, but you should not rely | |
7091 | on this implementation detail. | |
7092 | @end defmac | |
7093 | ||
7094 | @hook TARGET_INSERT_ATTRIBUTES | |
7095 | ||
7096 | @hook TARGET_HANDLE_GENERIC_ATTRIBUTE | |
7097 | ||
7098 | @hook TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P | |
7099 | ||
7100 | @hook TARGET_OPTION_VALID_ATTRIBUTE_P | |
7101 | ||
79891c4c AC |
7102 | @hook TARGET_OPTION_VALID_VERSION_ATTRIBUTE_P |
7103 | ||
d77de738 ML |
7104 | @hook TARGET_OPTION_SAVE |
7105 | ||
7106 | @hook TARGET_OPTION_RESTORE | |
7107 | ||
7108 | @hook TARGET_OPTION_POST_STREAM_IN | |
7109 | ||
7110 | @hook TARGET_OPTION_PRINT | |
7111 | ||
7112 | @hook TARGET_OPTION_PRAGMA_PARSE | |
7113 | ||
7114 | @hook TARGET_OPTION_OVERRIDE | |
7115 | ||
7116 | @hook TARGET_OPTION_FUNCTION_VERSIONS | |
7117 | ||
7118 | @hook TARGET_CAN_INLINE_P | |
7119 | ||
7120 | @hook TARGET_UPDATE_IPA_FN_TARGET_INFO | |
7121 | ||
7122 | @hook TARGET_NEED_IPA_FN_TARGET_INFO | |
7123 | ||
7124 | @hook TARGET_RELAYOUT_FUNCTION | |
7125 | ||
7126 | @node Emulated TLS | |
7127 | @section Emulating TLS | |
7128 | @cindex Emulated TLS | |
7129 | ||
7130 | For targets whose psABI does not provide Thread Local Storage via | |
7131 | specific relocations and instruction sequences, an emulation layer is | |
7132 | used. A set of target hooks allows this emulation layer to be | |
7133 | configured for the requirements of a particular target. For instance | |
7134 | the psABI may in fact specify TLS support in terms of an emulation | |
7135 | layer. | |
7136 | ||
7137 | The emulation layer works by creating a control object for every TLS | |
7138 | object. To access the TLS object, a lookup function is provided | |
7139 | which, when given the address of the control object, will return the | |
7140 | address of the current thread's instance of the TLS object. | |
7141 | ||
7142 | @hook TARGET_EMUTLS_GET_ADDRESS | |
7143 | ||
7144 | @hook TARGET_EMUTLS_REGISTER_COMMON | |
7145 | ||
7146 | @hook TARGET_EMUTLS_VAR_SECTION | |
7147 | ||
7148 | @hook TARGET_EMUTLS_TMPL_SECTION | |
7149 | ||
7150 | @hook TARGET_EMUTLS_VAR_PREFIX | |
7151 | ||
7152 | @hook TARGET_EMUTLS_TMPL_PREFIX | |
7153 | ||
7154 | @hook TARGET_EMUTLS_VAR_FIELDS | |
7155 | ||
7156 | @hook TARGET_EMUTLS_VAR_INIT | |
7157 | ||
7158 | @hook TARGET_EMUTLS_VAR_ALIGN_FIXED | |
7159 | ||
7160 | @hook TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS | |
7161 | ||
7162 | @node MIPS Coprocessors | |
7163 | @section Defining coprocessor specifics for MIPS targets. | |
7164 | @cindex MIPS coprocessor-definition macros | |
7165 | ||
7166 | The MIPS specification allows MIPS implementations to have as many as 4 | |
7167 | coprocessors, each with as many as 32 private registers. GCC supports | |
7168 | accessing these registers and transferring values between the registers | |
7169 | and memory using asm-ized variables. For example: | |
7170 | ||
7171 | @smallexample | |
7172 | register unsigned int cp0count asm ("c0r1"); | |
7173 | unsigned int d; | |
7174 | ||
7175 | d = cp0count + 3; | |
7176 | @end smallexample | |
7177 | ||
7178 | (``c0r1'' is the default name of register 1 in coprocessor 0; alternate | |
7179 | names may be added as described below, or the default names may be | |
7180 | overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.) | |
7181 | ||
7182 | Coprocessor registers are assumed to be epilogue-used; sets to them will | |
7183 | be preserved even if it does not appear that the register is used again | |
7184 | later in the function. | |
7185 | ||
7186 | Another note: according to the MIPS spec, coprocessor 1 (if present) is | |
7187 | the FPU@. One accesses COP1 registers through standard mips | |
7188 | floating-point support; they are not included in this mechanism. | |
7189 | ||
7190 | @node PCH Target | |
7191 | @section Parameters for Precompiled Header Validity Checking | |
7192 | @cindex parameters, precompiled headers | |
7193 | ||
7194 | @hook TARGET_GET_PCH_VALIDITY | |
7195 | ||
7196 | @hook TARGET_PCH_VALID_P | |
7197 | ||
7198 | @hook TARGET_CHECK_PCH_TARGET_FLAGS | |
7199 | ||
7200 | @hook TARGET_PREPARE_PCH_SAVE | |
7201 | ||
7202 | @node C++ ABI | |
7203 | @section C++ ABI parameters | |
7204 | @cindex parameters, c++ abi | |
7205 | ||
7206 | @hook TARGET_CXX_GUARD_TYPE | |
7207 | ||
7208 | @hook TARGET_CXX_GUARD_MASK_BIT | |
7209 | ||
7210 | @hook TARGET_CXX_GET_COOKIE_SIZE | |
7211 | ||
7212 | @hook TARGET_CXX_COOKIE_HAS_SIZE | |
7213 | ||
7214 | @hook TARGET_CXX_IMPORT_EXPORT_CLASS | |
7215 | ||
7216 | @hook TARGET_CXX_CDTOR_RETURNS_THIS | |
7217 | ||
7218 | @hook TARGET_CXX_KEY_METHOD_MAY_BE_INLINE | |
7219 | ||
7220 | @hook TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY | |
7221 | ||
7222 | @hook TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT | |
7223 | ||
7224 | @hook TARGET_CXX_LIBRARY_RTTI_COMDAT | |
7225 | ||
7226 | @hook TARGET_CXX_USE_AEABI_ATEXIT | |
7227 | ||
7228 | @hook TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT | |
7229 | ||
e5d8fd9c JJ |
7230 | @hook TARGET_CXX_ADJUST_CDTOR_CALLABI_FNTYPE |
7231 | ||
d77de738 ML |
7232 | @hook TARGET_CXX_ADJUST_CLASS_AT_DEFINITION |
7233 | ||
7234 | @hook TARGET_CXX_DECL_MANGLING_CONTEXT | |
7235 | ||
7236 | @node D Language and ABI | |
7237 | @section D ABI parameters | |
7238 | @cindex parameters, d abi | |
7239 | ||
7240 | @hook TARGET_D_CPU_VERSIONS | |
7241 | ||
7242 | @hook TARGET_D_OS_VERSIONS | |
7243 | ||
7244 | @hook TARGET_D_REGISTER_CPU_TARGET_INFO | |
7245 | ||
7246 | @hook TARGET_D_REGISTER_OS_TARGET_INFO | |
7247 | ||
7248 | @hook TARGET_D_MINFO_SECTION | |
7249 | ||
7250 | @hook TARGET_D_MINFO_SECTION_START | |
7251 | ||
7252 | @hook TARGET_D_MINFO_SECTION_END | |
7253 | ||
7254 | @hook TARGET_D_HAS_STDCALL_CONVENTION | |
7255 | ||
7256 | @hook TARGET_D_TEMPLATES_ALWAYS_COMDAT | |
7257 | ||
a7ea90e4 IB |
7258 | @node Rust Language and ABI |
7259 | @section Rust ABI parameters | |
7260 | @cindex parameters, rust abi | |
7261 | ||
7262 | @hook TARGET_RUST_CPU_INFO | |
7263 | ||
335da4e9 IB |
7264 | @hook TARGET_RUST_OS_INFO |
7265 | ||
d77de738 ML |
7266 | @node Named Address Spaces |
7267 | @section Adding support for named address spaces | |
7268 | @cindex named address spaces | |
7269 | ||
7270 | The draft technical report of the ISO/IEC JTC1 S22 WG14 N1275 | |
7271 | standards committee, @cite{Programming Languages - C - Extensions to | |
7272 | support embedded processors}, specifies a syntax for embedded | |
7273 | processors to specify alternate address spaces. You can configure a | |
7274 | GCC port to support section 5.1 of the draft report to add support for | |
7275 | address spaces other than the default address space. These address | |
7276 | spaces are new keywords that are similar to the @code{volatile} and | |
7277 | @code{const} type attributes. | |
7278 | ||
7279 | Pointers to named address spaces can have a different size than | |
7280 | pointers to the generic address space. | |
7281 | ||
7282 | For example, the SPU port uses the @code{__ea} address space to refer | |
7283 | to memory in the host processor, rather than memory local to the SPU | |
7284 | processor. Access to memory in the @code{__ea} address space involves | |
7285 | issuing DMA operations to move data between the host processor and the | |
7286 | local processor memory address space. Pointers in the @code{__ea} | |
7287 | address space are either 32 bits or 64 bits based on the | |
7288 | @option{-mea32} or @option{-mea64} switches (native SPU pointers are | |
7289 | always 32 bits). | |
7290 | ||
7291 | Internally, address spaces are represented as a small integer in the | |
7292 | range 0 to 15 with address space 0 being reserved for the generic | |
7293 | address space. | |
7294 | ||
7295 | To register a named address space qualifier keyword with the C front end, | |
7296 | the target may call the @code{c_register_addr_space} routine. For example, | |
7297 | the SPU port uses the following to declare @code{__ea} as the keyword for | |
7298 | named address space #1: | |
7299 | @smallexample | |
7300 | #define ADDR_SPACE_EA 1 | |
7301 | c_register_addr_space ("__ea", ADDR_SPACE_EA); | |
7302 | @end smallexample | |
7303 | ||
7304 | @hook TARGET_ADDR_SPACE_POINTER_MODE | |
7305 | ||
7306 | @hook TARGET_ADDR_SPACE_ADDRESS_MODE | |
7307 | ||
7308 | @hook TARGET_ADDR_SPACE_VALID_POINTER_MODE | |
7309 | ||
7310 | @hook TARGET_ADDR_SPACE_LEGITIMATE_ADDRESS_P | |
7311 | ||
7312 | @hook TARGET_ADDR_SPACE_LEGITIMIZE_ADDRESS | |
7313 | ||
7314 | @hook TARGET_ADDR_SPACE_SUBSET_P | |
7315 | ||
7316 | @hook TARGET_ADDR_SPACE_ZERO_ADDRESS_VALID | |
7317 | ||
7318 | @hook TARGET_ADDR_SPACE_CONVERT | |
7319 | ||
7320 | @hook TARGET_ADDR_SPACE_DEBUG | |
7321 | ||
7322 | @hook TARGET_ADDR_SPACE_DIAGNOSE_USAGE | |
7323 | ||
7324 | @node Misc | |
7325 | @section Miscellaneous Parameters | |
7326 | @cindex parameters, miscellaneous | |
7327 | ||
7328 | @c prevent bad page break with this line | |
7329 | Here are several miscellaneous parameters. | |
7330 | ||
7331 | @defmac HAS_LONG_COND_BRANCH | |
7332 | Define this boolean macro to indicate whether or not your architecture | |
7333 | has conditional branches that can span all of memory. It is used in | |
7334 | conjunction with an optimization that partitions hot and cold basic | |
7335 | blocks into separate sections of the executable. If this macro is | |
7336 | set to false, gcc will convert any conditional branches that attempt | |
7337 | to cross between sections into unconditional branches or indirect jumps. | |
7338 | @end defmac | |
7339 | ||
7340 | @defmac HAS_LONG_UNCOND_BRANCH | |
7341 | Define this boolean macro to indicate whether or not your architecture | |
7342 | has unconditional branches that can span all of memory. It is used in | |
7343 | conjunction with an optimization that partitions hot and cold basic | |
7344 | blocks into separate sections of the executable. If this macro is | |
7345 | set to false, gcc will convert any unconditional branches that attempt | |
7346 | to cross between sections into indirect jumps. | |
7347 | @end defmac | |
7348 | ||
7349 | @defmac CASE_VECTOR_MODE | |
7350 | An alias for a machine mode name. This is the machine mode that | |
7351 | elements of a jump-table should have. | |
7352 | @end defmac | |
7353 | ||
7354 | @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body}) | |
7355 | Optional: return the preferred mode for an @code{addr_diff_vec} | |
7356 | when the minimum and maximum offset are known. If you define this, | |
7357 | it enables extra code in branch shortening to deal with @code{addr_diff_vec}. | |
7358 | To make this work, you also have to define @code{INSN_ALIGN} and | |
7359 | make the alignment for @code{addr_diff_vec} explicit. | |
7360 | The @var{body} argument is provided so that the offset_unsigned and scale | |
7361 | flags can be updated. | |
7362 | @end defmac | |
7363 | ||
7364 | @defmac CASE_VECTOR_PC_RELATIVE | |
7365 | Define this macro to be a C expression to indicate when jump-tables | |
7366 | should contain relative addresses. You need not define this macro if | |
7367 | jump-tables never contain relative addresses, or jump-tables should | |
7368 | contain relative addresses only when @option{-fPIC} or @option{-fPIC} | |
7369 | is in effect. | |
7370 | @end defmac | |
7371 | ||
7372 | @hook TARGET_CASE_VALUES_THRESHOLD | |
7373 | ||
7374 | @defmac WORD_REGISTER_OPERATIONS | |
7375 | Define this macro to 1 if operations between registers with integral mode | |
7376 | smaller than a word are always performed on the entire register. To be | |
7377 | more explicit, if you start with a pair of @code{word_mode} registers with | |
7378 | known values and you do a subword, for example @code{QImode}, addition on | |
7379 | the low part of the registers, then the compiler may consider that the | |
7380 | result has a known value in @code{word_mode} too if the macro is defined | |
7381 | to 1. Most RISC machines have this property and most CISC machines do not. | |
7382 | @end defmac | |
7383 | ||
7384 | @hook TARGET_MIN_ARITHMETIC_PRECISION | |
7385 | ||
7386 | @defmac LOAD_EXTEND_OP (@var{mem_mode}) | |
7387 | Define this macro to be a C expression indicating when insns that read | |
7388 | memory in @var{mem_mode}, an integral mode narrower than a word, set the | |
7389 | bits outside of @var{mem_mode} to be either the sign-extension or the | |
7390 | zero-extension of the data read. Return @code{SIGN_EXTEND} for values | |
7391 | of @var{mem_mode} for which the | |
7392 | insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and | |
7393 | @code{UNKNOWN} for other modes. | |
7394 | ||
7395 | This macro is not called with @var{mem_mode} non-integral or with a width | |
7396 | greater than or equal to @code{BITS_PER_WORD}, so you may return any | |
7397 | value in this case. Do not define this macro if it would always return | |
7398 | @code{UNKNOWN}. On machines where this macro is defined, you will normally | |
7399 | define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}. | |
7400 | ||
7401 | You may return a non-@code{UNKNOWN} value even if for some hard registers | |
7402 | the sign extension is not performed, if for the @code{REGNO_REG_CLASS} | |
7403 | of these hard registers @code{TARGET_CAN_CHANGE_MODE_CLASS} returns false | |
7404 | when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any | |
7405 | integral mode larger than this but not larger than @code{word_mode}. | |
7406 | ||
7407 | You must return @code{UNKNOWN} if for some hard registers that allow this | |
7408 | mode, @code{TARGET_CAN_CHANGE_MODE_CLASS} says that they cannot change to | |
7409 | @code{word_mode}, but that they can change to another integral mode that | |
7410 | is larger then @var{mem_mode} but still smaller than @code{word_mode}. | |
7411 | @end defmac | |
7412 | ||
7413 | @defmac SHORT_IMMEDIATES_SIGN_EXTEND | |
7414 | Define this macro to 1 if loading short immediate values into registers sign | |
7415 | extends. | |
7416 | @end defmac | |
7417 | ||
7418 | @hook TARGET_MIN_DIVISIONS_FOR_RECIP_MUL | |
7419 | ||
7420 | @defmac MOVE_MAX | |
7421 | The maximum number of bytes that a single instruction can move quickly | |
7422 | between memory and registers or between two memory locations. | |
7423 | @end defmac | |
7424 | ||
7425 | @defmac MAX_MOVE_MAX | |
7426 | The maximum number of bytes that a single instruction can move quickly | |
7427 | between memory and registers or between two memory locations. If this | |
7428 | is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the | |
7429 | constant value that is the largest value that @code{MOVE_MAX} can have | |
7430 | at run-time. | |
7431 | @end defmac | |
7432 | ||
7433 | @defmac SHIFT_COUNT_TRUNCATED | |
7434 | A C expression that is nonzero if on this machine the number of bits | |
7435 | actually used for the count of a shift operation is equal to the number | |
7436 | of bits needed to represent the size of the object being shifted. When | |
7437 | this macro is nonzero, the compiler will assume that it is safe to omit | |
7438 | a sign-extend, zero-extend, and certain bitwise `and' instructions that | |
7439 | truncates the count of a shift operation. On machines that have | |
7440 | instructions that act on bit-fields at variable positions, which may | |
7441 | include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED} | |
7442 | also enables deletion of truncations of the values that serve as | |
7443 | arguments to bit-field instructions. | |
7444 | ||
7445 | If both types of instructions truncate the count (for shifts) and | |
7446 | position (for bit-field operations), or if no variable-position bit-field | |
7447 | instructions exist, you should define this macro. | |
7448 | ||
7449 | However, on some machines, such as the 80386 and the 680x0, truncation | |
7450 | only applies to shift operations and not the (real or pretended) | |
7451 | bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on | |
7452 | such machines. Instead, add patterns to the @file{md} file that include | |
7453 | the implied truncation of the shift instructions. | |
7454 | ||
7455 | You need not define this macro if it would always have the value of zero. | |
7456 | @end defmac | |
7457 | ||
7458 | @anchor{TARGET_SHIFT_TRUNCATION_MASK} | |
7459 | @hook TARGET_SHIFT_TRUNCATION_MASK | |
7460 | ||
7461 | @hook TARGET_TRULY_NOOP_TRUNCATION | |
7462 | ||
7463 | @hook TARGET_MODE_REP_EXTENDED | |
7464 | ||
7465 | @hook TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P | |
7466 | ||
7467 | @defmac STORE_FLAG_VALUE | |
7468 | A C expression describing the value returned by a comparison operator | |
7469 | with an integral mode and stored by a store-flag instruction | |
7470 | (@samp{cstore@var{mode}4}) when the condition is true. This description must | |
7471 | apply to @emph{all} the @samp{cstore@var{mode}4} patterns and all the | |
7472 | comparison operators whose results have a @code{MODE_INT} mode. | |
7473 | ||
7474 | A value of 1 or @minus{}1 means that the instruction implementing the | |
7475 | comparison operator returns exactly 1 or @minus{}1 when the comparison is true | |
7476 | and 0 when the comparison is false. Otherwise, the value indicates | |
7477 | which bits of the result are guaranteed to be 1 when the comparison is | |
7478 | true. This value is interpreted in the mode of the comparison | |
7479 | operation, which is given by the mode of the first operand in the | |
7480 | @samp{cstore@var{mode}4} pattern. Either the low bit or the sign bit of | |
7481 | @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by | |
7482 | the compiler. | |
7483 | ||
7484 | If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will | |
7485 | generate code that depends only on the specified bits. It can also | |
7486 | replace comparison operators with equivalent operations if they cause | |
7487 | the required bits to be set, even if the remaining bits are undefined. | |
7488 | For example, on a machine whose comparison operators return an | |
7489 | @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as | |
7490 | @samp{0x80000000}, saying that just the sign bit is relevant, the | |
7491 | expression | |
7492 | ||
7493 | @smallexample | |
7494 | (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0)) | |
7495 | @end smallexample | |
7496 | ||
7497 | @noindent | |
7498 | can be converted to | |
7499 | ||
7500 | @smallexample | |
7501 | (ashift:SI @var{x} (const_int @var{n})) | |
7502 | @end smallexample | |
7503 | ||
7504 | @noindent | |
7505 | where @var{n} is the appropriate shift count to move the bit being | |
7506 | tested into the sign bit. | |
7507 | ||
7508 | There is no way to describe a machine that always sets the low-order bit | |
7509 | for a true value, but does not guarantee the value of any other bits, | |
7510 | but we do not know of any machine that has such an instruction. If you | |
7511 | are trying to port GCC to such a machine, include an instruction to | |
7512 | perform a logical-and of the result with 1 in the pattern for the | |
7513 | comparison operators and let us know at @email{gcc@@gcc.gnu.org}. | |
7514 | ||
7515 | Often, a machine will have multiple instructions that obtain a value | |
7516 | from a comparison (or the condition codes). Here are rules to guide the | |
7517 | choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions | |
7518 | to be used: | |
7519 | ||
7520 | @itemize @bullet | |
7521 | @item | |
7522 | Use the shortest sequence that yields a valid definition for | |
7523 | @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to | |
7524 | ``normalize'' the value (convert it to, e.g., 1 or 0) than for the | |
7525 | comparison operators to do so because there may be opportunities to | |
7526 | combine the normalization with other operations. | |
7527 | ||
7528 | @item | |
7529 | For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being | |
7530 | slightly preferred on machines with expensive jumps and 1 preferred on | |
7531 | other machines. | |
7532 | ||
7533 | @item | |
7534 | As a second choice, choose a value of @samp{0x80000001} if instructions | |
7535 | exist that set both the sign and low-order bits but do not define the | |
7536 | others. | |
7537 | ||
7538 | @item | |
7539 | Otherwise, use a value of @samp{0x80000000}. | |
7540 | @end itemize | |
7541 | ||
7542 | Many machines can produce both the value chosen for | |
7543 | @code{STORE_FLAG_VALUE} and its negation in the same number of | |
7544 | instructions. On those machines, you should also define a pattern for | |
7545 | those cases, e.g., one matching | |
7546 | ||
7547 | @smallexample | |
7548 | (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C}))) | |
7549 | @end smallexample | |
7550 | ||
7551 | Some machines can also perform @code{and} or @code{plus} operations on | |
7552 | condition code values with less instructions than the corresponding | |
7553 | @samp{cstore@var{mode}4} insn followed by @code{and} or @code{plus}. On those | |
7554 | machines, define the appropriate patterns. Use the names @code{incscc} | |
7555 | and @code{decscc}, respectively, for the patterns which perform | |
7556 | @code{plus} or @code{minus} operations on condition code values. See | |
7557 | @file{rs6000.md} for some examples. The GNU Superoptimizer can be used to | |
7558 | find such instruction sequences on other machines. | |
7559 | ||
7560 | If this macro is not defined, the default value, 1, is used. You need | |
7561 | not define @code{STORE_FLAG_VALUE} if the machine has no store-flag | |
7562 | instructions, or if the value generated by these instructions is 1. | |
7563 | @end defmac | |
7564 | ||
7565 | @defmac FLOAT_STORE_FLAG_VALUE (@var{mode}) | |
7566 | A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is | |
7567 | returned when comparison operators with floating-point results are true. | |
7568 | Define this macro on machines that have comparison operations that return | |
7569 | floating-point values. If there are no such operations, do not define | |
7570 | this macro. | |
7571 | @end defmac | |
7572 | ||
7573 | @defmac VECTOR_STORE_FLAG_VALUE (@var{mode}) | |
7574 | A C expression that gives an rtx representing the nonzero true element | |
7575 | for vector comparisons. The returned rtx should be valid for the inner | |
7576 | mode of @var{mode} which is guaranteed to be a vector mode. Define | |
7577 | this macro on machines that have vector comparison operations that | |
7578 | return a vector result. If there are no such operations, do not define | |
7579 | this macro. Typically, this macro is defined as @code{const1_rtx} or | |
7580 | @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent | |
7581 | the compiler optimizing such vector comparison operations for the | |
7582 | given mode. | |
7583 | @end defmac | |
7584 | ||
7585 | @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value}) | |
7586 | @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value}) | |
7587 | A C expression that indicates whether the architecture defines a value | |
7588 | for @code{clz} or @code{ctz} with a zero operand. | |
7589 | A result of @code{0} indicates the value is undefined. | |
7590 | If the value is defined for only the RTL expression, the macro should | |
7591 | evaluate to @code{1}; if the value applies also to the corresponding optab | |
7592 | entry (which is normally the case if it expands directly into | |
7593 | the corresponding RTL), then the macro should evaluate to @code{2}. | |
7594 | In the cases where the value is defined, @var{value} should be set to | |
7595 | this value. | |
7596 | ||
7597 | If this macro is not defined, the value of @code{clz} or | |
7598 | @code{ctz} at zero is assumed to be undefined. | |
7599 | ||
7600 | This macro must be defined if the target's expansion for @code{ffs} | |
7601 | relies on a particular value to get correct results. Otherwise it | |
7602 | is not necessary, though it may be used to optimize some corner cases, and | |
7603 | to provide a default expansion for the @code{ffs} optab. | |
7604 | ||
7605 | Note that regardless of this macro the ``definedness'' of @code{clz} | |
7606 | and @code{ctz} at zero do @emph{not} extend to the builtin functions | |
7607 | visible to the user. Thus one may be free to adjust the value at will | |
7608 | to match the target expansion of these operations without fear of | |
7609 | breaking the API@. | |
7610 | @end defmac | |
7611 | ||
7612 | @defmac Pmode | |
7613 | An alias for the machine mode for pointers. On most machines, define | |
7614 | this to be the integer mode corresponding to the width of a hardware | |
7615 | pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines. | |
7616 | On some machines you must define this to be one of the partial integer | |
7617 | modes, such as @code{PSImode}. | |
7618 | ||
7619 | The width of @code{Pmode} must be at least as large as the value of | |
7620 | @code{POINTER_SIZE}. If it is not equal, you must define the macro | |
7621 | @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended | |
7622 | to @code{Pmode}. | |
7623 | @end defmac | |
7624 | ||
7625 | @defmac FUNCTION_MODE | |
7626 | An alias for the machine mode used for memory references to functions | |
7627 | being called, in @code{call} RTL expressions. On most CISC machines, | |
7628 | where an instruction can begin at any byte address, this should be | |
7629 | @code{QImode}. On most RISC machines, where all instructions have fixed | |
7630 | size and alignment, this should be a mode with the same size and alignment | |
7631 | as the machine instruction words - typically @code{SImode} or @code{HImode}. | |
7632 | @end defmac | |
7633 | ||
7634 | @defmac STDC_0_IN_SYSTEM_HEADERS | |
7635 | In normal operation, the preprocessor expands @code{__STDC__} to the | |
7636 | constant 1, to signify that GCC conforms to ISO Standard C@. On some | |
7637 | hosts, like Solaris, the system compiler uses a different convention, | |
7638 | where @code{__STDC__} is normally 0, but is 1 if the user specifies | |
7639 | strict conformance to the C Standard. | |
7640 | ||
7641 | Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host | |
7642 | convention when processing system header files, but when processing user | |
7643 | files @code{__STDC__} will always expand to 1. | |
7644 | @end defmac | |
7645 | ||
7646 | @hook TARGET_C_PREINCLUDE | |
7647 | ||
7648 | @hook TARGET_CXX_IMPLICIT_EXTERN_C | |
7649 | ||
7650 | @defmac SYSTEM_IMPLICIT_EXTERN_C | |
7651 | Define this macro if the system header files do not support C++@. | |
7652 | This macro handles system header files by pretending that system | |
7653 | header files are enclosed in @samp{extern "C" @{@dots{}@}}. | |
7654 | @end defmac | |
7655 | ||
7656 | @findex #pragma | |
7657 | @findex pragma | |
7658 | @defmac REGISTER_TARGET_PRAGMAS () | |
7659 | Define this macro if you want to implement any target-specific pragmas. | |
7660 | If defined, it is a C expression which makes a series of calls to | |
7661 | @code{c_register_pragma} or @code{c_register_pragma_with_expansion} | |
7662 | for each pragma. The macro may also do any | |
7663 | setup required for the pragmas. | |
7664 | ||
7665 | The primary reason to define this macro is to provide compatibility with | |
7666 | other compilers for the same target. In general, we discourage | |
7667 | definition of target-specific pragmas for GCC@. | |
7668 | ||
7669 | If the pragma can be implemented by attributes then you should consider | |
7670 | defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well. | |
7671 | ||
7672 | Preprocessor macros that appear on pragma lines are not expanded. All | |
7673 | @samp{#pragma} directives that do not match any registered pragma are | |
7674 | silently ignored, unless the user specifies @option{-Wunknown-pragmas}. | |
7675 | @end defmac | |
7676 | ||
7677 | @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *)) | |
7678 | @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *)) | |
7679 | ||
7680 | Each call to @code{c_register_pragma} or | |
7681 | @code{c_register_pragma_with_expansion} establishes one pragma. The | |
7682 | @var{callback} routine will be called when the preprocessor encounters a | |
7683 | pragma of the form | |
7684 | ||
7685 | @smallexample | |
7686 | #pragma [@var{space}] @var{name} @dots{} | |
7687 | @end smallexample | |
7688 | ||
7689 | @var{space} is the case-sensitive namespace of the pragma, or | |
7690 | @code{NULL} to put the pragma in the global namespace. The callback | |
7691 | routine receives @var{pfile} as its first argument, which can be passed | |
7692 | on to cpplib's functions if necessary. You can lex tokens after the | |
7693 | @var{name} by calling @code{pragma_lex}. Tokens that are not read by the | |
7694 | callback will be silently ignored. The end of the line is indicated by | |
7695 | a token of type @code{CPP_EOF}. Macro expansion occurs on the | |
7696 | arguments of pragmas registered with | |
7697 | @code{c_register_pragma_with_expansion} but not on the arguments of | |
7698 | pragmas registered with @code{c_register_pragma}. | |
7699 | ||
7700 | Note that the use of @code{pragma_lex} is specific to the C and C++ | |
7701 | compilers. It will not work in the Java or Fortran compilers, or any | |
7702 | other language compilers for that matter. Thus if @code{pragma_lex} is going | |
7703 | to be called from target-specific code, it must only be done so when | |
7704 | building the C and C++ compilers. This can be done by defining the | |
7705 | variables @code{c_target_objs} and @code{cxx_target_objs} in the | |
7706 | target entry in the @file{config.gcc} file. These variables should name | |
7707 | the target-specific, language-specific object file which contains the | |
7708 | code that uses @code{pragma_lex}. Note it will also be necessary to add a | |
7709 | rule to the makefile fragment pointed to by @code{tmake_file} that shows | |
7710 | how to build this object file. | |
7711 | @end deftypefun | |
7712 | ||
7713 | @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION | |
7714 | Define this macro if macros should be expanded in the | |
7715 | arguments of @samp{#pragma pack}. | |
7716 | @end defmac | |
7717 | ||
7718 | @defmac TARGET_DEFAULT_PACK_STRUCT | |
7719 | If your target requires a structure packing default other than 0 (meaning | |
7720 | the machine default), define this macro to the necessary value (in bytes). | |
7721 | This must be a value that would also be valid to use with | |
7722 | @samp{#pragma pack()} (that is, a small power of two). | |
7723 | @end defmac | |
7724 | ||
7725 | @defmac DOLLARS_IN_IDENTIFIERS | |
7726 | Define this macro to control use of the character @samp{$} in | |
7727 | identifier names for the C family of languages. 0 means @samp{$} is | |
7728 | not allowed by default; 1 means it is allowed. 1 is the default; | |
7729 | there is no need to define this macro in that case. | |
7730 | @end defmac | |
7731 | ||
7732 | @defmac INSN_SETS_ARE_DELAYED (@var{insn}) | |
7733 | Define this macro as a C expression that is nonzero if it is safe for the | |
7734 | delay slot scheduler to place instructions in the delay slot of @var{insn}, | |
7735 | even if they appear to use a resource set or clobbered in @var{insn}. | |
7736 | @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that | |
7737 | every @code{call_insn} has this behavior. On machines where some @code{insn} | |
7738 | or @code{jump_insn} is really a function call and hence has this behavior, | |
7739 | you should define this macro. | |
7740 | ||
7741 | You need not define this macro if it would always return zero. | |
7742 | @end defmac | |
7743 | ||
7744 | @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn}) | |
7745 | Define this macro as a C expression that is nonzero if it is safe for the | |
7746 | delay slot scheduler to place instructions in the delay slot of @var{insn}, | |
7747 | even if they appear to set or clobber a resource referenced in @var{insn}. | |
7748 | @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where | |
7749 | some @code{insn} or @code{jump_insn} is really a function call and its operands | |
7750 | are registers whose use is actually in the subroutine it calls, you should | |
7751 | define this macro. Doing so allows the delay slot scheduler to move | |
7752 | instructions which copy arguments into the argument registers into the delay | |
7753 | slot of @var{insn}. | |
7754 | ||
7755 | You need not define this macro if it would always return zero. | |
7756 | @end defmac | |
7757 | ||
7758 | @defmac MULTIPLE_SYMBOL_SPACES | |
7759 | Define this macro as a C expression that is nonzero if, in some cases, | |
7760 | global symbols from one translation unit may not be bound to undefined | |
7761 | symbols in another translation unit without user intervention. For | |
7762 | instance, under Microsoft Windows symbols must be explicitly imported | |
7763 | from shared libraries (DLLs). | |
7764 | ||
7765 | You need not define this macro if it would always evaluate to zero. | |
7766 | @end defmac | |
7767 | ||
7768 | @hook TARGET_MD_ASM_ADJUST | |
7769 | ||
7770 | @defmac MATH_LIBRARY | |
7771 | Define this macro as a C string constant for the linker argument to link | |
7772 | in the system math library, minus the initial @samp{"-l"}, or | |
7773 | @samp{""} if the target does not have a | |
7774 | separate math library. | |
7775 | ||
7776 | You need only define this macro if the default of @samp{"m"} is wrong. | |
7777 | @end defmac | |
7778 | ||
7779 | @defmac LIBRARY_PATH_ENV | |
7780 | Define this macro as a C string constant for the environment variable that | |
7781 | specifies where the linker should look for libraries. | |
7782 | ||
7783 | You need only define this macro if the default of @samp{"LIBRARY_PATH"} | |
7784 | is wrong. | |
7785 | @end defmac | |
7786 | ||
7787 | @defmac TARGET_POSIX_IO | |
7788 | Define this macro if the target supports the following POSIX@ file | |
7789 | functions, access, mkdir and file locking with fcntl / F_SETLKW@. | |
7790 | Defining @code{TARGET_POSIX_IO} will enable the test coverage code | |
7791 | to use file locking when exiting a program, which avoids race conditions | |
7792 | if the program has forked. It will also create directories at run-time | |
7793 | for cross-profiling. | |
7794 | @end defmac | |
7795 | ||
7796 | @defmac MAX_CONDITIONAL_EXECUTE | |
7797 | ||
7798 | A C expression for the maximum number of instructions to execute via | |
7799 | conditional execution instructions instead of a branch. A value of | |
7800 | @code{BRANCH_COST}+1 is the default. | |
7801 | @end defmac | |
7802 | ||
7803 | @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr}) | |
7804 | Used if the target needs to perform machine-dependent modifications on the | |
7805 | conditionals used for turning basic blocks into conditionally executed code. | |
7806 | @var{ce_info} points to a data structure, @code{struct ce_if_block}, which | |
7807 | contains information about the currently processed blocks. @var{true_expr} | |
7808 | and @var{false_expr} are the tests that are used for converting the | |
7809 | then-block and the else-block, respectively. Set either @var{true_expr} or | |
7810 | @var{false_expr} to a null pointer if the tests cannot be converted. | |
7811 | @end defmac | |
7812 | ||
7813 | @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr}) | |
7814 | Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated | |
7815 | if-statements into conditions combined by @code{and} and @code{or} operations. | |
7816 | @var{bb} contains the basic block that contains the test that is currently | |
7817 | being processed and about to be turned into a condition. | |
7818 | @end defmac | |
7819 | ||
7820 | @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn}) | |
7821 | A C expression to modify the @var{PATTERN} of an @var{INSN} that is to | |
7822 | be converted to conditional execution format. @var{ce_info} points to | |
7823 | a data structure, @code{struct ce_if_block}, which contains information | |
7824 | about the currently processed blocks. | |
7825 | @end defmac | |
7826 | ||
7827 | @defmac IFCVT_MODIFY_FINAL (@var{ce_info}) | |
7828 | A C expression to perform any final machine dependent modifications in | |
7829 | converting code to conditional execution. The involved basic blocks | |
7830 | can be found in the @code{struct ce_if_block} structure that is pointed | |
7831 | to by @var{ce_info}. | |
7832 | @end defmac | |
7833 | ||
7834 | @defmac IFCVT_MODIFY_CANCEL (@var{ce_info}) | |
7835 | A C expression to cancel any machine dependent modifications in | |
7836 | converting code to conditional execution. The involved basic blocks | |
7837 | can be found in the @code{struct ce_if_block} structure that is pointed | |
7838 | to by @var{ce_info}. | |
7839 | @end defmac | |
7840 | ||
7841 | @defmac IFCVT_MACHDEP_INIT (@var{ce_info}) | |
7842 | A C expression to initialize any machine specific data for if-conversion | |
7843 | of the if-block in the @code{struct ce_if_block} structure that is pointed | |
7844 | to by @var{ce_info}. | |
7845 | @end defmac | |
7846 | ||
e9d2ae6b RS |
7847 | @hook TARGET_USE_LATE_PROLOGUE_EPILOGUE |
7848 | ||
2e0aefa7 RS |
7849 | @hook TARGET_EMIT_EPILOGUE_FOR_SIBCALL |
7850 | ||
d77de738 ML |
7851 | @hook TARGET_MACHINE_DEPENDENT_REORG |
7852 | ||
7853 | @hook TARGET_INIT_BUILTINS | |
7854 | ||
7855 | @hook TARGET_BUILTIN_DECL | |
7856 | ||
7857 | @hook TARGET_EXPAND_BUILTIN | |
7858 | ||
7859 | @hook TARGET_RESOLVE_OVERLOADED_BUILTIN | |
7860 | ||
7861 | @hook TARGET_CHECK_BUILTIN_CALL | |
7862 | ||
7863 | @hook TARGET_FOLD_BUILTIN | |
7864 | ||
7865 | @hook TARGET_GIMPLE_FOLD_BUILTIN | |
7866 | ||
7867 | @hook TARGET_COMPARE_VERSION_PRIORITY | |
7868 | ||
7869 | @hook TARGET_GET_FUNCTION_VERSIONS_DISPATCHER | |
7870 | ||
7871 | @hook TARGET_GENERATE_VERSION_DISPATCHER_BODY | |
7872 | ||
7873 | @hook TARGET_PREDICT_DOLOOP_P | |
7874 | ||
7875 | @hook TARGET_HAVE_COUNT_REG_DECR_P | |
7876 | ||
7877 | @hook TARGET_DOLOOP_COST_FOR_GENERIC | |
7878 | ||
7879 | @hook TARGET_DOLOOP_COST_FOR_ADDRESS | |
7880 | ||
7881 | @hook TARGET_CAN_USE_DOLOOP_P | |
7882 | ||
7883 | @hook TARGET_INVALID_WITHIN_DOLOOP | |
7884 | ||
7885 | @hook TARGET_PREFERRED_DOLOOP_MODE | |
7886 | ||
7887 | @hook TARGET_LEGITIMATE_COMBINED_INSN | |
7888 | ||
7889 | @hook TARGET_CAN_FOLLOW_JUMP | |
7890 | ||
7891 | @hook TARGET_COMMUTATIVE_P | |
7892 | ||
7893 | @hook TARGET_ALLOCATE_INITIAL_VALUE | |
7894 | ||
7895 | @hook TARGET_UNSPEC_MAY_TRAP_P | |
7896 | ||
7897 | @hook TARGET_SET_CURRENT_FUNCTION | |
7898 | ||
7899 | @defmac TARGET_OBJECT_SUFFIX | |
7900 | Define this macro to be a C string representing the suffix for object | |
7901 | files on your target machine. If you do not define this macro, GCC will | |
7902 | use @samp{.o} as the suffix for object files. | |
7903 | @end defmac | |
7904 | ||
7905 | @defmac TARGET_EXECUTABLE_SUFFIX | |
7906 | Define this macro to be a C string representing the suffix to be | |
7907 | automatically added to executable files on your target machine. If you | |
7908 | do not define this macro, GCC will use the null string as the suffix for | |
7909 | executable files. | |
7910 | @end defmac | |
7911 | ||
7912 | @defmac COLLECT_EXPORT_LIST | |
7913 | If defined, @code{collect2} will scan the individual object files | |
7914 | specified on its command line and create an export list for the linker. | |
7915 | Define this macro for systems like AIX, where the linker discards | |
7916 | object files that are not referenced from @code{main} and uses export | |
7917 | lists. | |
7918 | @end defmac | |
7919 | ||
7920 | @hook TARGET_CANNOT_MODIFY_JUMPS_P | |
7921 | ||
7922 | @hook TARGET_HAVE_CONDITIONAL_EXECUTION | |
7923 | ||
7924 | @hook TARGET_GEN_CCMP_FIRST | |
7925 | ||
7926 | @hook TARGET_GEN_CCMP_NEXT | |
7927 | ||
83a76576 HW |
7928 | @hook TARGET_HAVE_CCMP |
7929 | ||
d77de738 ML |
7930 | @hook TARGET_LOOP_UNROLL_ADJUST |
7931 | ||
7932 | @defmac POWI_MAX_MULTS | |
7933 | If defined, this macro is interpreted as a signed integer C expression | |
7934 | that specifies the maximum number of floating point multiplications | |
7935 | that should be emitted when expanding exponentiation by an integer | |
7936 | constant inline. When this value is defined, exponentiation requiring | |
7937 | more than this number of multiplications is implemented by calling the | |
7938 | system library's @code{pow}, @code{powf} or @code{powl} routines. | |
7939 | The default value places no upper bound on the multiplication count. | |
7940 | @end defmac | |
7941 | ||
7942 | @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc}) | |
7943 | This target hook should register any extra include files for the | |
7944 | target. The parameter @var{stdinc} indicates if normal include files | |
7945 | are present. The parameter @var{sysroot} is the system root directory. | |
7946 | The parameter @var{iprefix} is the prefix for the gcc directory. | |
7947 | @end deftypefn | |
7948 | ||
7949 | @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc}) | |
7950 | This target hook should register any extra include files for the | |
7951 | target before any standard headers. The parameter @var{stdinc} | |
7952 | indicates if normal include files are present. The parameter | |
7953 | @var{sysroot} is the system root directory. The parameter | |
7954 | @var{iprefix} is the prefix for the gcc directory. | |
7955 | @end deftypefn | |
7956 | ||
7957 | @deftypefn Macro void TARGET_OPTF (char *@var{path}) | |
7958 | This target hook should register special include paths for the target. | |
7959 | The parameter @var{path} is the include to register. On Darwin | |
7960 | systems, this is used for Framework includes, which have semantics | |
7961 | that are different from @option{-I}. | |
7962 | @end deftypefn | |
7963 | ||
7964 | @defmac bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl}) | |
7965 | This target macro returns @code{true} if it is safe to use a local alias | |
7966 | for a virtual function @var{fndecl} when constructing thunks, | |
7967 | @code{false} otherwise. By default, the macro returns @code{true} for all | |
7968 | functions, if a target supports aliases (i.e.@: defines | |
7969 | @code{ASM_OUTPUT_DEF}), @code{false} otherwise, | |
7970 | @end defmac | |
7971 | ||
7972 | @defmac TARGET_FORMAT_TYPES | |
7973 | If defined, this macro is the name of a global variable containing | |
7974 | target-specific format checking information for the @option{-Wformat} | |
7975 | option. The default is to have no target-specific format checks. | |
7976 | @end defmac | |
7977 | ||
7978 | @defmac TARGET_N_FORMAT_TYPES | |
7979 | If defined, this macro is the number of entries in | |
7980 | @code{TARGET_FORMAT_TYPES}. | |
7981 | @end defmac | |
7982 | ||
7983 | @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES | |
7984 | If defined, this macro is the name of a global variable containing | |
7985 | target-specific format overrides for the @option{-Wformat} option. The | |
7986 | default is to have no target-specific format overrides. If defined, | |
7987 | @code{TARGET_FORMAT_TYPES} and @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT} | |
7988 | must be defined, too. | |
7989 | @end defmac | |
7990 | ||
7991 | @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT | |
7992 | If defined, this macro specifies the number of entries in | |
7993 | @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}. | |
7994 | @end defmac | |
7995 | ||
7996 | @defmac TARGET_OVERRIDES_FORMAT_INIT | |
7997 | If defined, this macro specifies the optional initialization | |
7998 | routine for target specific customizations of the system printf | |
7999 | and scanf formatter settings. | |
8000 | @end defmac | |
8001 | ||
8002 | @hook TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN | |
8003 | ||
8004 | @hook TARGET_INVALID_CONVERSION | |
8005 | ||
8006 | @hook TARGET_INVALID_UNARY_OP | |
8007 | ||
8008 | @hook TARGET_INVALID_BINARY_OP | |
8009 | ||
8010 | @hook TARGET_PROMOTED_TYPE | |
8011 | ||
8012 | @hook TARGET_CONVERT_TO_TYPE | |
8013 | ||
8014 | @hook TARGET_VERIFY_TYPE_CONTEXT | |
8015 | ||
8016 | @defmac OBJC_JBLEN | |
8017 | This macro determines the size of the objective C jump buffer for the | |
8018 | NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value. | |
8019 | @end defmac | |
8020 | ||
8021 | @defmac LIBGCC2_UNWIND_ATTRIBUTE | |
8022 | Define this macro if any target-specific attributes need to be attached | |
8023 | to the functions in @file{libgcc} that provide low-level support for | |
8024 | call stack unwinding. It is used in declarations in @file{unwind-generic.h} | |
8025 | and the associated definitions of those functions. | |
8026 | @end defmac | |
8027 | ||
8028 | @hook TARGET_UPDATE_STACK_BOUNDARY | |
8029 | ||
8030 | @hook TARGET_GET_DRAP_RTX | |
8031 | ||
8032 | @hook TARGET_ZERO_CALL_USED_REGS | |
8033 | ||
8034 | @hook TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS | |
8035 | ||
8036 | @hook TARGET_CONST_ANCHOR | |
8037 | ||
8038 | @hook TARGET_ASAN_SHADOW_OFFSET | |
8039 | ||
8040 | @hook TARGET_MEMMODEL_CHECK | |
8041 | ||
8042 | @hook TARGET_ATOMIC_TEST_AND_SET_TRUEVAL | |
8043 | ||
8044 | @hook TARGET_HAS_IFUNC_P | |
8045 | ||
8046 | @hook TARGET_IFUNC_REF_LOCAL_OK | |
8047 | ||
8048 | @hook TARGET_ATOMIC_ALIGN_FOR_MODE | |
8049 | ||
8050 | @hook TARGET_ATOMIC_ASSIGN_EXPAND_FENV | |
8051 | ||
8052 | @hook TARGET_RECORD_OFFLOAD_SYMBOL | |
8053 | ||
8054 | @hook TARGET_OFFLOAD_OPTIONS | |
8055 | ||
8056 | @defmac TARGET_SUPPORTS_WIDE_INT | |
8057 | ||
8058 | On older ports, large integers are stored in @code{CONST_DOUBLE} rtl | |
8059 | objects. Newer ports define @code{TARGET_SUPPORTS_WIDE_INT} to be nonzero | |
8060 | to indicate that large integers are stored in | |
8061 | @code{CONST_WIDE_INT} rtl objects. The @code{CONST_WIDE_INT} allows | |
8062 | very large integer constants to be represented. @code{CONST_DOUBLE} | |
8063 | is limited to twice the size of the host's @code{HOST_WIDE_INT} | |
8064 | representation. | |
8065 | ||
8066 | Converting a port mostly requires looking for the places where | |
8067 | @code{CONST_DOUBLE}s are used with @code{VOIDmode} and replacing that | |
8068 | code with code that accesses @code{CONST_WIDE_INT}s. @samp{"grep -i | |
8069 | const_double"} at the port level gets you to 95% of the changes that | |
8070 | need to be made. There are a few places that require a deeper look. | |
8071 | ||
8072 | @itemize @bullet | |
8073 | @item | |
8074 | There is no equivalent to @code{hval} and @code{lval} for | |
8075 | @code{CONST_WIDE_INT}s. This would be difficult to express in the md | |
8076 | language since there are a variable number of elements. | |
8077 | ||
8078 | Most ports only check that @code{hval} is either 0 or -1 to see if the | |
8079 | value is small. As mentioned above, this will no longer be necessary | |
8080 | since small constants are always @code{CONST_INT}. Of course there | |
8081 | are still a few exceptions, the alpha's constraint used by the zap | |
8082 | instruction certainly requires careful examination by C code. | |
8083 | However, all the current code does is pass the hval and lval to C | |
8084 | code, so evolving the c code to look at the @code{CONST_WIDE_INT} is | |
8085 | not really a large change. | |
8086 | ||
8087 | @item | |
8088 | Because there is no standard template that ports use to materialize | |
8089 | constants, there is likely to be some futzing that is unique to each | |
8090 | port in this code. | |
8091 | ||
8092 | @item | |
8093 | The rtx costs may have to be adjusted to properly account for larger | |
8094 | constants that are represented as @code{CONST_WIDE_INT}. | |
8095 | @end itemize | |
8096 | ||
8097 | All and all it does not take long to convert ports that the | |
8098 | maintainer is familiar with. | |
8099 | ||
8100 | @end defmac | |
8101 | ||
8102 | @hook TARGET_HAVE_SPECULATION_SAFE_VALUE | |
8103 | ||
8104 | @hook TARGET_SPECULATION_SAFE_VALUE | |
8105 | ||
8106 | @hook TARGET_RUN_TARGET_SELFTESTS | |
8107 | ||
8108 | @hook TARGET_MEMTAG_CAN_TAG_ADDRESSES | |
8109 | ||
8110 | @hook TARGET_MEMTAG_TAG_SIZE | |
8111 | ||
8112 | @hook TARGET_MEMTAG_GRANULE_SIZE | |
8113 | ||
8114 | @hook TARGET_MEMTAG_INSERT_RANDOM_TAG | |
8115 | ||
8116 | @hook TARGET_MEMTAG_ADD_TAG | |
8117 | ||
8118 | @hook TARGET_MEMTAG_SET_TAG | |
8119 | ||
8120 | @hook TARGET_MEMTAG_EXTRACT_TAG | |
8121 | ||
8122 | @hook TARGET_MEMTAG_UNTAGGED_POINTER | |
8123 | ||
d77de738 | 8124 | @hook TARGET_HAVE_SHADOW_CALL_STACK |
3ef8882a SH |
8125 | |
8126 | @hook TARGET_HAVE_LIBATOMIC |