]>
Commit | Line | Data |
---|---|---|
feca2ed3 JW |
1 | @c Copyright (C) 1988,89,92,93,94,96 Free Software Foundation, Inc. |
2 | @c This is part of the GCC manual. | |
3 | @c For copying conditions, see the file gcc.texi. | |
4 | ||
5 | @node Target Macros | |
6 | @chapter Target Description Macros | |
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}. This header file defines numerous macros | |
15 | that convey the information about the target machine that does not fit | |
16 | into the scheme of the @file{.md} file. The file @file{tm.h} should be | |
17 | a link to @file{@var{machine}.h}. The header file @file{config.h} | |
18 | includes @file{tm.h} and most compiler source files include | |
19 | @file{config.h}. | |
20 | ||
21 | @menu | |
22 | * Driver:: Controlling how the driver runs the compilation passes. | |
23 | * Run-time Target:: Defining @samp{-m} options like @samp{-m68000} and @samp{-m68020}. | |
24 | * Storage Layout:: Defining sizes and alignments of data. | |
25 | * Type Layout:: Defining sizes and properties of basic user data types. | |
26 | * Registers:: Naming and describing the hardware registers. | |
27 | * Register Classes:: Defining the classes of hardware registers. | |
28 | * Stack and Calling:: Defining which way the stack grows and by how much. | |
29 | * Varargs:: Defining the varargs macros. | |
30 | * Trampolines:: Code set up at run time to enter a nested function. | |
31 | * Library Calls:: Controlling how library routines are implicitly called. | |
32 | * Addressing Modes:: Defining addressing modes valid for memory operands. | |
33 | * Condition Code:: Defining how insns update the condition code. | |
34 | * Costs:: Defining relative costs of different operations. | |
35 | * Sections:: Dividing storage into text, data, and other sections. | |
36 | * PIC:: Macros for position independent code. | |
37 | * Assembler Format:: Defining how to write insns and pseudo-ops to output. | |
38 | * Debugging Info:: Defining the format of debugging output. | |
39 | * Cross-compilation:: Handling floating point for cross-compilers. | |
40 | * Misc:: Everything else. | |
41 | @end menu | |
42 | ||
43 | @node Driver | |
44 | @section Controlling the Compilation Driver, @file{gcc} | |
45 | @cindex driver | |
46 | @cindex controlling the compilation driver | |
47 | ||
48 | @c prevent bad page break with this line | |
49 | You can control the compilation driver. | |
50 | ||
51 | @table @code | |
52 | @findex SWITCH_TAKES_ARG | |
53 | @item SWITCH_TAKES_ARG (@var{char}) | |
54 | A C expression which determines whether the option @samp{-@var{char}} | |
55 | takes arguments. The value should be the number of arguments that | |
56 | option takes--zero, for many options. | |
57 | ||
58 | By default, this macro is defined as | |
59 | @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options | |
60 | properly. You need not define @code{SWITCH_TAKES_ARG} unless you | |
61 | wish to add additional options which take arguments. Any redefinition | |
62 | should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for | |
63 | additional options. | |
64 | ||
65 | @findex WORD_SWITCH_TAKES_ARG | |
66 | @item WORD_SWITCH_TAKES_ARG (@var{name}) | |
67 | A C expression which determines whether the option @samp{-@var{name}} | |
68 | takes arguments. The value should be the number of arguments that | |
69 | option takes--zero, for many options. This macro rather than | |
70 | @code{SWITCH_TAKES_ARG} is used for multi-character option names. | |
71 | ||
72 | By default, this macro is defined as | |
73 | @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options | |
74 | properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you | |
75 | wish to add additional options which take arguments. Any redefinition | |
76 | should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for | |
77 | additional options. | |
78 | ||
79 | @findex SWITCHES_NEED_SPACES | |
80 | @item SWITCHES_NEED_SPACES | |
81 | A string-valued C expression which enumerates the options for which | |
82 | the linker needs a space between the option and its argument. | |
83 | ||
84 | If this macro is not defined, the default value is @code{""}. | |
85 | ||
86 | @findex CPP_SPEC | |
87 | @item CPP_SPEC | |
88 | A C string constant that tells the GNU CC driver program options to | |
89 | pass to CPP. It can also specify how to translate options you | |
90 | give to GNU CC into options for GNU CC to pass to the CPP. | |
91 | ||
92 | Do not define this macro if it does not need to do anything. | |
93 | ||
94 | @findex NO_BUILTIN_SIZE_TYPE | |
95 | @item NO_BUILTIN_SIZE_TYPE | |
96 | If this macro is defined, the preprocessor will not define the builtin macro | |
97 | @code{__SIZE_TYPE__}. The macro @code{__SIZE_TYPE__} must then be defined | |
98 | by @code{CPP_SPEC} instead. | |
99 | ||
100 | This should be defined if @code{SIZE_TYPE} depends on target dependent flags | |
101 | which are not accessible to the preprocessor. Otherwise, it should not | |
102 | be defined. | |
103 | ||
104 | @findex NO_BUILTIN_PTRDIFF_TYPE | |
105 | @item NO_BUILTIN_PTRDIFF_TYPE | |
106 | If this macro is defined, the preprocessor will not define the builtin macro | |
107 | @code{__PTRDIFF_TYPE__}. The macro @code{__PTRDIFF_TYPE__} must then be | |
108 | defined by @code{CPP_SPEC} instead. | |
109 | ||
110 | This should be defined if @code{PTRDIFF_TYPE} depends on target dependent flags | |
111 | which are not accessible to the preprocessor. Otherwise, it should not | |
112 | be defined. | |
113 | ||
114 | @findex SIGNED_CHAR_SPEC | |
115 | @item SIGNED_CHAR_SPEC | |
116 | A C string constant that tells the GNU CC driver program options to | |
117 | pass to CPP. By default, this macro is defined to pass the option | |
118 | @samp{-D__CHAR_UNSIGNED__} to CPP if @code{char} will be treated as | |
119 | @code{unsigned char} by @code{cc1}. | |
120 | ||
121 | Do not define this macro unless you need to override the default | |
122 | definition. | |
123 | ||
124 | @findex CC1_SPEC | |
125 | @item CC1_SPEC | |
126 | A C string constant that tells the GNU CC driver program options to | |
127 | pass to @code{cc1}. It can also specify how to translate options you | |
128 | give to GNU CC into options for GNU CC to pass to the @code{cc1}. | |
129 | ||
130 | Do not define this macro if it does not need to do anything. | |
131 | ||
132 | @findex CC1PLUS_SPEC | |
133 | @item CC1PLUS_SPEC | |
134 | A C string constant that tells the GNU CC driver program options to | |
135 | pass to @code{cc1plus}. It can also specify how to translate options you | |
136 | give to GNU CC into options for GNU CC to pass to the @code{cc1plus}. | |
137 | ||
138 | Do not define this macro if it does not need to do anything. | |
139 | ||
140 | @findex ASM_SPEC | |
141 | @item ASM_SPEC | |
142 | A C string constant that tells the GNU CC driver program options to | |
143 | pass to the assembler. It can also specify how to translate options | |
144 | you give to GNU CC into options for GNU CC to pass to the assembler. | |
145 | See the file @file{sun3.h} for an example of this. | |
146 | ||
147 | Do not define this macro if it does not need to do anything. | |
148 | ||
149 | @findex ASM_FINAL_SPEC | |
150 | @item ASM_FINAL_SPEC | |
151 | A C string constant that tells the GNU CC driver program how to | |
152 | run any programs which cleanup after the normal assembler. | |
153 | Normally, this is not needed. See the file @file{mips.h} for | |
154 | an example of this. | |
155 | ||
156 | Do not define this macro if it does not need to do anything. | |
157 | ||
158 | @findex LINK_SPEC | |
159 | @item LINK_SPEC | |
160 | A C string constant that tells the GNU CC driver program options to | |
161 | pass to the linker. It can also specify how to translate options you | |
162 | give to GNU CC into options for GNU CC to pass to the linker. | |
163 | ||
164 | Do not define this macro if it does not need to do anything. | |
165 | ||
166 | @findex LIB_SPEC | |
167 | @item LIB_SPEC | |
168 | Another C string constant used much like @code{LINK_SPEC}. The difference | |
169 | between the two is that @code{LIB_SPEC} is used at the end of the | |
170 | command given to the linker. | |
171 | ||
172 | If this macro is not defined, a default is provided that | |
173 | loads the standard C library from the usual place. See @file{gcc.c}. | |
174 | ||
175 | @findex LIBGCC_SPEC | |
176 | @item LIBGCC_SPEC | |
177 | Another C string constant that tells the GNU CC driver program | |
178 | how and when to place a reference to @file{libgcc.a} into the | |
179 | linker command line. This constant is placed both before and after | |
180 | the value of @code{LIB_SPEC}. | |
181 | ||
182 | If this macro is not defined, the GNU CC driver provides a default that | |
183 | passes the string @samp{-lgcc} to the linker unless the @samp{-shared} | |
184 | option is specified. | |
185 | ||
186 | @findex STARTFILE_SPEC | |
187 | @item STARTFILE_SPEC | |
188 | Another C string constant used much like @code{LINK_SPEC}. The | |
189 | difference between the two is that @code{STARTFILE_SPEC} is used at | |
190 | the very beginning of the command given to the linker. | |
191 | ||
192 | If this macro is not defined, a default is provided that loads the | |
193 | standard C startup file from the usual place. See @file{gcc.c}. | |
194 | ||
195 | @findex ENDFILE_SPEC | |
196 | @item ENDFILE_SPEC | |
197 | Another C string constant used much like @code{LINK_SPEC}. The | |
198 | difference between the two is that @code{ENDFILE_SPEC} is used at | |
199 | the very end of the command given to the linker. | |
200 | ||
201 | Do not define this macro if it does not need to do anything. | |
202 | ||
203 | @findex EXTRA_SPECS | |
204 | @item EXTRA_SPECS | |
205 | Define this macro to provide additional specifications to put in the | |
206 | @file{specs} file that can be used in various specifications like | |
207 | @code{CC1_SPEC}. | |
208 | ||
209 | The definition should be an initializer for an array of structures, | |
210 | containing a string constant, that defines the specification name, and a | |
211 | string constant that provides the specification. | |
212 | ||
213 | Do not define this macro if it does not need to do anything. | |
214 | ||
215 | @code{EXTRA_SPECS} is useful when an architecture contains several | |
216 | related targets, which have various @code{..._SPECS} which are similar | |
217 | to each other, and the maintainer would like one central place to keep | |
218 | these definitions. | |
219 | ||
220 | For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to | |
221 | define either @code{_CALL_SYSV} when the System V calling sequence is | |
222 | used or @code{_CALL_AIX} when the older AIX-based calling sequence is | |
223 | used. | |
224 | ||
225 | The @file{config/rs6000/rs6000.h} target file defines: | |
226 | ||
227 | @example | |
228 | #define EXTRA_SPECS \ | |
229 | @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @}, | |
230 | ||
231 | #define CPP_SYS_DEFAULT "" | |
232 | @end example | |
233 | ||
234 | The @file{config/rs6000/sysv.h} target file defines: | |
235 | @smallexample | |
236 | #undef CPP_SPEC | |
237 | #define CPP_SPEC \ | |
238 | "%@{posix: -D_POSIX_SOURCE @} \ | |
239 | %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \ | |
240 | %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \ | |
241 | %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}" | |
242 | ||
243 | #undef CPP_SYSV_DEFAULT | |
244 | #define CPP_SYSV_DEFAULT "-D_CALL_SYSV" | |
245 | @end smallexample | |
246 | ||
247 | while the @file{config/rs6000/eabiaix.h} target file defines | |
248 | @code{CPP_SYSV_DEFAULT} as: | |
249 | ||
250 | @smallexample | |
251 | #undef CPP_SYSV_DEFAULT | |
252 | #define CPP_SYSV_DEFAULT "-D_CALL_AIX" | |
253 | @end smallexample | |
254 | ||
255 | @findex LINK_LIBGCC_SPECIAL | |
256 | @item LINK_LIBGCC_SPECIAL | |
257 | Define this macro if the driver program should find the library | |
258 | @file{libgcc.a} itself and should not pass @samp{-L} options to the | |
259 | linker. If you do not define this macro, the driver program will pass | |
260 | the argument @samp{-lgcc} to tell the linker to do the search and will | |
261 | pass @samp{-L} options to it. | |
262 | ||
263 | @findex LINK_LIBGCC_SPECIAL_1 | |
264 | @item LINK_LIBGCC_SPECIAL_1 | |
265 | Define this macro if the driver program should find the library | |
266 | @file{libgcc.a}. If you do not define this macro, the driver program will pass | |
267 | the argument @samp{-lgcc} to tell the linker to do the search. | |
268 | This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does | |
269 | not affect @samp{-L} options. | |
270 | ||
271 | @findex MULTILIB_DEFAULTS | |
272 | @item MULTILIB_DEFAULTS | |
273 | Define this macro as a C expression for the initializer of an array of | |
274 | string to tell the driver program which options are defaults for this | |
275 | target and thus do not need to be handled specially when using | |
276 | @code{MULTILIB_OPTIONS}. | |
277 | ||
278 | Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in | |
279 | the target makefile fragment or if none of the options listed in | |
280 | @code{MULTILIB_OPTIONS} are set by default. | |
281 | @xref{Target Fragment}. | |
282 | ||
283 | @findex RELATIVE_PREFIX_NOT_LINKDIR | |
284 | @item RELATIVE_PREFIX_NOT_LINKDIR | |
285 | Define this macro to tell @code{gcc} that it should only translate | |
286 | a @samp{-B} prefix into a @samp{-L} linker option if the prefix | |
287 | indicates an absolute file name. | |
288 | ||
289 | @findex STANDARD_EXEC_PREFIX | |
290 | @item STANDARD_EXEC_PREFIX | |
291 | Define this macro as a C string constant if you wish to override the | |
292 | standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to | |
293 | try when searching for the executable files of the compiler. | |
294 | ||
295 | @findex MD_EXEC_PREFIX | |
296 | @item MD_EXEC_PREFIX | |
297 | If defined, this macro is an additional prefix to try after | |
298 | @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched | |
299 | when the @samp{-b} option is used, or the compiler is built as a cross | |
300 | compiler. | |
301 | ||
302 | @findex STANDARD_STARTFILE_PREFIX | |
303 | @item STANDARD_STARTFILE_PREFIX | |
304 | Define this macro as a C string constant if you wish to override the | |
305 | standard choice of @file{/usr/local/lib/} as the default prefix to | |
306 | try when searching for startup files such as @file{crt0.o}. | |
307 | ||
308 | @findex MD_STARTFILE_PREFIX | |
309 | @item MD_STARTFILE_PREFIX | |
310 | If defined, this macro supplies an additional prefix to try after the | |
311 | standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the | |
312 | @samp{-b} option is used, or when the compiler is built as a cross | |
313 | compiler. | |
314 | ||
315 | @findex MD_STARTFILE_PREFIX_1 | |
316 | @item MD_STARTFILE_PREFIX_1 | |
317 | If defined, this macro supplies yet another prefix to try after the | |
318 | standard prefixes. It is not searched when the @samp{-b} option is | |
319 | used, or when the compiler is built as a cross compiler. | |
320 | ||
321 | @findex INIT_ENVIRONMENT | |
322 | @item INIT_ENVIRONMENT | |
323 | Define this macro as a C string constant if you with to set environment | |
324 | variables for programs called by the driver, such as the assembler and | |
325 | loader. The driver passes the value of this macro to @code{putenv} to | |
326 | initialize the necessary environment variables. | |
327 | ||
328 | @findex LOCAL_INCLUDE_DIR | |
329 | @item LOCAL_INCLUDE_DIR | |
330 | Define this macro as a C string constant if you wish to override the | |
331 | standard choice of @file{/usr/local/include} as the default prefix to | |
332 | try when searching for local header files. @code{LOCAL_INCLUDE_DIR} | |
333 | comes before @code{SYSTEM_INCLUDE_DIR} in the search order. | |
334 | ||
335 | Cross compilers do not use this macro and do not search either | |
336 | @file{/usr/local/include} or its replacement. | |
337 | ||
338 | @findex SYSTEM_INCLUDE_DIR | |
339 | @item SYSTEM_INCLUDE_DIR | |
340 | Define this macro as a C string constant if you wish to specify a | |
341 | system-specific directory to search for header files before the standard | |
342 | directory. @code{SYSTEM_INCLUDE_DIR} comes before | |
343 | @code{STANDARD_INCLUDE_DIR} in the search order. | |
344 | ||
345 | Cross compilers do not use this macro and do not search the directory | |
346 | specified. | |
347 | ||
348 | @findex STANDARD_INCLUDE_DIR | |
349 | @item STANDARD_INCLUDE_DIR | |
350 | Define this macro as a C string constant if you wish to override the | |
351 | standard choice of @file{/usr/include} as the default prefix to | |
352 | try when searching for header files. | |
353 | ||
354 | Cross compilers do not use this macro and do not search either | |
355 | @file{/usr/include} or its replacement. | |
356 | ||
357 | @findex INCLUDE_DEFAULTS | |
358 | @item INCLUDE_DEFAULTS | |
359 | Define this macro if you wish to override the entire default search path | |
360 | for include files. The default search path includes | |
361 | @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR}, | |
362 | @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and | |
363 | @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR} | |
364 | and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile}, | |
365 | and specify private search areas for GCC. The directory | |
366 | @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs. | |
367 | ||
368 | The definition should be an initializer for an array of structures. | |
369 | Each array element should have two elements: the directory name (a | |
370 | string constant) and a flag for C++-only directories. Mark the end of | |
371 | the array with a null element. For example, here is the definition used | |
372 | for VMS: | |
373 | ||
374 | @example | |
375 | #define INCLUDE_DEFAULTS \ | |
376 | @{ \ | |
377 | @{ "GNU_GXX_INCLUDE:", 1@}, \ | |
378 | @{ "GNU_CC_INCLUDE:", 0@}, \ | |
379 | @{ "SYS$SYSROOT:[SYSLIB.]", 0@}, \ | |
380 | @{ ".", 0@}, \ | |
381 | @{ 0, 0@} \ | |
382 | @} | |
383 | @end example | |
384 | @end table | |
385 | ||
386 | Here is the order of prefixes tried for exec files: | |
387 | ||
388 | @enumerate | |
389 | @item | |
390 | Any prefixes specified by the user with @samp{-B}. | |
391 | ||
392 | @item | |
393 | The environment variable @code{GCC_EXEC_PREFIX}, if any. | |
394 | ||
395 | @item | |
396 | The directories specified by the environment variable @code{COMPILER_PATH}. | |
397 | ||
398 | @item | |
399 | The macro @code{STANDARD_EXEC_PREFIX}. | |
400 | ||
401 | @item | |
402 | @file{/usr/lib/gcc/}. | |
403 | ||
404 | @item | |
405 | The macro @code{MD_EXEC_PREFIX}, if any. | |
406 | @end enumerate | |
407 | ||
408 | Here is the order of prefixes tried for startfiles: | |
409 | ||
410 | @enumerate | |
411 | @item | |
412 | Any prefixes specified by the user with @samp{-B}. | |
413 | ||
414 | @item | |
415 | The environment variable @code{GCC_EXEC_PREFIX}, if any. | |
416 | ||
417 | @item | |
418 | The directories specified by the environment variable @code{LIBRARY_PATH} | |
419 | (native only, cross compilers do not use this). | |
420 | ||
421 | @item | |
422 | The macro @code{STANDARD_EXEC_PREFIX}. | |
423 | ||
424 | @item | |
425 | @file{/usr/lib/gcc/}. | |
426 | ||
427 | @item | |
428 | The macro @code{MD_EXEC_PREFIX}, if any. | |
429 | ||
430 | @item | |
431 | The macro @code{MD_STARTFILE_PREFIX}, if any. | |
432 | ||
433 | @item | |
434 | The macro @code{STANDARD_STARTFILE_PREFIX}. | |
435 | ||
436 | @item | |
437 | @file{/lib/}. | |
438 | ||
439 | @item | |
440 | @file{/usr/lib/}. | |
441 | @end enumerate | |
442 | ||
443 | @node Run-time Target | |
444 | @section Run-time Target Specification | |
445 | @cindex run-time target specification | |
446 | @cindex predefined macros | |
447 | @cindex target specifications | |
448 | ||
449 | @c prevent bad page break with this line | |
450 | Here are run-time target specifications. | |
451 | ||
452 | @table @code | |
453 | @findex CPP_PREDEFINES | |
454 | @item CPP_PREDEFINES | |
455 | Define this to be a string constant containing @samp{-D} options to | |
456 | define the predefined macros that identify this machine and system. | |
457 | These macros will be predefined unless the @samp{-ansi} option is | |
458 | specified. | |
459 | ||
460 | In addition, a parallel set of macros are predefined, whose names are | |
461 | made by appending @samp{__} at the beginning and at the end. These | |
462 | @samp{__} macros are permitted by the ANSI standard, so they are | |
463 | predefined regardless of whether @samp{-ansi} is specified. | |
464 | ||
465 | For example, on the Sun, one can use the following value: | |
466 | ||
467 | @smallexample | |
468 | "-Dmc68000 -Dsun -Dunix" | |
469 | @end smallexample | |
470 | ||
471 | The result is to define the macros @code{__mc68000__}, @code{__sun__} | |
472 | and @code{__unix__} unconditionally, and the macros @code{mc68000}, | |
473 | @code{sun} and @code{unix} provided @samp{-ansi} is not specified. | |
474 | ||
475 | @findex extern int target_flags | |
476 | @item extern int target_flags; | |
477 | This declaration should be present. | |
478 | ||
479 | @cindex optional hardware or system features | |
480 | @cindex features, optional, in system conventions | |
481 | @item TARGET_@dots{} | |
482 | This series of macros is to allow compiler command arguments to | |
483 | enable or disable the use of optional features of the target machine. | |
484 | For example, one machine description serves both the 68000 and | |
485 | the 68020; a command argument tells the compiler whether it should | |
486 | use 68020-only instructions or not. This command argument works | |
487 | by means of a macro @code{TARGET_68020} that tests a bit in | |
488 | @code{target_flags}. | |
489 | ||
490 | Define a macro @code{TARGET_@var{featurename}} for each such option. | |
491 | Its definition should test a bit in @code{target_flags}; for example: | |
492 | ||
493 | @smallexample | |
494 | #define TARGET_68020 (target_flags & 1) | |
495 | @end smallexample | |
496 | ||
497 | One place where these macros are used is in the condition-expressions | |
498 | of instruction patterns. Note how @code{TARGET_68020} appears | |
499 | frequently in the 68000 machine description file, @file{m68k.md}. | |
500 | Another place they are used is in the definitions of the other | |
501 | macros in the @file{@var{machine}.h} file. | |
502 | ||
503 | @findex TARGET_SWITCHES | |
504 | @item TARGET_SWITCHES | |
505 | This macro defines names of command options to set and clear | |
506 | bits in @code{target_flags}. Its definition is an initializer | |
507 | with a subgrouping for each command option. | |
508 | ||
509 | Each subgrouping contains a string constant, that defines the option | |
510 | name, and a number, which contains the bits to set in | |
511 | @code{target_flags}. A negative number says to clear bits instead; | |
512 | the negative of the number is which bits to clear. The actual option | |
513 | name is made by appending @samp{-m} to the specified name. | |
514 | ||
515 | One of the subgroupings should have a null string. The number in | |
516 | this grouping is the default value for @code{target_flags}. Any | |
517 | target options act starting with that value. | |
518 | ||
519 | Here is an example which defines @samp{-m68000} and @samp{-m68020} | |
520 | with opposite meanings, and picks the latter as the default: | |
521 | ||
522 | @smallexample | |
523 | #define TARGET_SWITCHES \ | |
524 | @{ @{ "68020", 1@}, \ | |
525 | @{ "68000", -1@}, \ | |
526 | @{ "", 1@}@} | |
527 | @end smallexample | |
528 | ||
529 | @findex TARGET_OPTIONS | |
530 | @item TARGET_OPTIONS | |
531 | This macro is similar to @code{TARGET_SWITCHES} but defines names of command | |
532 | options that have values. Its definition is an initializer with a | |
533 | subgrouping for each command option. | |
534 | ||
535 | Each subgrouping contains a string constant, that defines the fixed part | |
536 | of the option name, and the address of a variable. The variable, type | |
537 | @code{char *}, is set to the variable part of the given option if the fixed | |
538 | part matches. The actual option name is made by appending @samp{-m} to the | |
539 | specified name. | |
540 | ||
541 | Here is an example which defines @samp{-mshort-data-@var{number}}. If the | |
542 | given option is @samp{-mshort-data-512}, the variable @code{m88k_short_data} | |
543 | will be set to the string @code{"512"}. | |
544 | ||
545 | @smallexample | |
546 | extern char *m88k_short_data; | |
547 | #define TARGET_OPTIONS \ | |
548 | @{ @{ "short-data-", &m88k_short_data @} @} | |
549 | @end smallexample | |
550 | ||
551 | @findex TARGET_VERSION | |
552 | @item TARGET_VERSION | |
553 | This macro is a C statement to print on @code{stderr} a string | |
554 | describing the particular machine description choice. Every machine | |
555 | description should define @code{TARGET_VERSION}. For example: | |
556 | ||
557 | @smallexample | |
558 | #ifdef MOTOROLA | |
559 | #define TARGET_VERSION \ | |
560 | fprintf (stderr, " (68k, Motorola syntax)"); | |
561 | #else | |
562 | #define TARGET_VERSION \ | |
563 | fprintf (stderr, " (68k, MIT syntax)"); | |
564 | #endif | |
565 | @end smallexample | |
566 | ||
567 | @findex OVERRIDE_OPTIONS | |
568 | @item OVERRIDE_OPTIONS | |
569 | Sometimes certain combinations of command options do not make sense on | |
570 | a particular target machine. You can define a macro | |
571 | @code{OVERRIDE_OPTIONS} to take account of this. This macro, if | |
572 | defined, is executed once just after all the command options have been | |
573 | parsed. | |
574 | ||
575 | Don't use this macro to turn on various extra optimizations for | |
576 | @samp{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for. | |
577 | ||
578 | @findex OPTIMIZATION_OPTIONS | |
579 | @item OPTIMIZATION_OPTIONS (@var{level}) | |
580 | Some machines may desire to change what optimizations are performed for | |
581 | various optimization levels. This macro, if defined, is executed once | |
582 | just after the optimization level is determined and before the remainder | |
583 | of the command options have been parsed. Values set in this macro are | |
584 | used as the default values for the other command line options. | |
585 | ||
586 | @var{level} is the optimization level specified; 2 if @samp{-O2} is | |
587 | specified, 1 if @samp{-O} is specified, and 0 if neither is specified. | |
588 | ||
589 | You should not use this macro to change options that are not | |
590 | machine-specific. These should uniformly selected by the same | |
591 | optimization level on all supported machines. Use this macro to enable | |
592 | machine-specific optimizations. | |
593 | ||
594 | @strong{Do not examine @code{write_symbols} in | |
595 | this macro!} The debugging options are not supposed to alter the | |
596 | generated code. | |
597 | ||
598 | @findex CAN_DEBUG_WITHOUT_FP | |
599 | @item CAN_DEBUG_WITHOUT_FP | |
600 | Define this macro if debugging can be performed even without a frame | |
601 | pointer. If this macro is defined, GNU CC will turn on the | |
602 | @samp{-fomit-frame-pointer} option whenever @samp{-O} is specified. | |
603 | @end table | |
604 | ||
605 | @node Storage Layout | |
606 | @section Storage Layout | |
607 | @cindex storage layout | |
608 | ||
609 | Note that the definitions of the macros in this table which are sizes or | |
610 | alignments measured in bits do not need to be constant. They can be C | |
611 | expressions that refer to static variables, such as the @code{target_flags}. | |
612 | @xref{Run-time Target}. | |
613 | ||
614 | @table @code | |
615 | @findex BITS_BIG_ENDIAN | |
616 | @item BITS_BIG_ENDIAN | |
617 | Define this macro to have the value 1 if the most significant bit in a | |
618 | byte has the lowest number; otherwise define it to have the value zero. | |
619 | This means that bit-field instructions count from the most significant | |
620 | bit. If the machine has no bit-field instructions, then this must still | |
621 | be defined, but it doesn't matter which value it is defined to. This | |
622 | macro need not be a constant. | |
623 | ||
624 | This macro does not affect the way structure fields are packed into | |
625 | bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}. | |
626 | ||
627 | @findex BYTES_BIG_ENDIAN | |
628 | @item BYTES_BIG_ENDIAN | |
629 | Define this macro to have the value 1 if the most significant byte in a | |
630 | word has the lowest number. This macro need not be a constant. | |
631 | ||
632 | @findex WORDS_BIG_ENDIAN | |
633 | @item WORDS_BIG_ENDIAN | |
634 | Define this macro to have the value 1 if, in a multiword object, the | |
635 | most significant word has the lowest number. This applies to both | |
636 | memory locations and registers; GNU CC fundamentally assumes that the | |
637 | order of words in memory is the same as the order in registers. This | |
638 | macro need not be a constant. | |
639 | ||
640 | @findex LIBGCC2_WORDS_BIG_ENDIAN | |
641 | @item LIBGCC2_WORDS_BIG_ENDIAN | |
642 | Define this macro if WORDS_BIG_ENDIAN is not constant. This must be a | |
643 | constant value with the same meaning as WORDS_BIG_ENDIAN, which will be | |
644 | used only when compiling libgcc2.c. Typically the value will be set | |
645 | based on preprocessor defines. | |
646 | ||
647 | @findex FLOAT_WORDS_BIG_ENDIAN | |
648 | @item FLOAT_WORDS_BIG_ENDIAN | |
649 | Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or | |
650 | @code{TFmode} floating point numbers are stored in memory with the word | |
651 | containing the sign bit at the lowest address; otherwise define it to | |
652 | have the value 0. This macro need not be a constant. | |
653 | ||
654 | You need not define this macro if the ordering is the same as for | |
655 | multi-word integers. | |
656 | ||
657 | @findex BITS_PER_UNIT | |
658 | @item BITS_PER_UNIT | |
659 | Define this macro to be the number of bits in an addressable storage | |
660 | unit (byte); normally 8. | |
661 | ||
662 | @findex BITS_PER_WORD | |
663 | @item BITS_PER_WORD | |
664 | Number of bits in a word; normally 32. | |
665 | ||
666 | @findex MAX_BITS_PER_WORD | |
667 | @item MAX_BITS_PER_WORD | |
668 | Maximum number of bits in a word. If this is undefined, the default is | |
669 | @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the | |
670 | largest value that @code{BITS_PER_WORD} can have at run-time. | |
671 | ||
672 | @findex UNITS_PER_WORD | |
673 | @item UNITS_PER_WORD | |
674 | Number of storage units in a word; normally 4. | |
675 | ||
676 | @findex MIN_UNITS_PER_WORD | |
677 | @item MIN_UNITS_PER_WORD | |
678 | Minimum number of units in a word. If this is undefined, the default is | |
679 | @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the | |
680 | smallest value that @code{UNITS_PER_WORD} can have at run-time. | |
681 | ||
682 | @findex POINTER_SIZE | |
683 | @item POINTER_SIZE | |
684 | Width of a pointer, in bits. You must specify a value no wider than the | |
685 | width of @code{Pmode}. If it is not equal to the width of @code{Pmode}, | |
686 | you must define @code{POINTERS_EXTEND_UNSIGNED}. | |
687 | ||
688 | @findex POINTERS_EXTEND_UNSIGNED | |
689 | @item POINTERS_EXTEND_UNSIGNED | |
690 | A C expression whose value is nonzero if pointers that need to be | |
691 | extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} | |
692 | are sign-extended and zero if they are zero-extended. | |
693 | ||
694 | You need not define this macro if the @code{POINTER_SIZE} is equal | |
695 | to the width of @code{Pmode}. | |
696 | ||
697 | @findex PROMOTE_MODE | |
698 | @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type}) | |
699 | A macro to update @var{m} and @var{unsignedp} when an object whose type | |
700 | is @var{type} and which has the specified mode and signedness is to be | |
701 | stored in a register. This macro is only called when @var{type} is a | |
702 | scalar type. | |
703 | ||
704 | On most RISC machines, which only have operations that operate on a full | |
705 | register, define this macro to set @var{m} to @code{word_mode} if | |
706 | @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most | |
707 | cases, only integer modes should be widened because wider-precision | |
708 | floating-point operations are usually more expensive than their narrower | |
709 | counterparts. | |
710 | ||
711 | For most machines, the macro definition does not change @var{unsignedp}. | |
712 | However, some machines, have instructions that preferentially handle | |
713 | either signed or unsigned quantities of certain modes. For example, on | |
714 | the DEC Alpha, 32-bit loads from memory and 32-bit add instructions | |
715 | sign-extend the result to 64 bits. On such machines, set | |
716 | @var{unsignedp} according to which kind of extension is more efficient. | |
717 | ||
718 | Do not define this macro if it would never modify @var{m}. | |
719 | ||
720 | @findex PROMOTE_FUNCTION_ARGS | |
721 | @item PROMOTE_FUNCTION_ARGS | |
722 | Define this macro if the promotion described by @code{PROMOTE_MODE} | |
723 | should also be done for outgoing function arguments. | |
724 | ||
725 | @findex PROMOTE_FUNCTION_RETURN | |
726 | @item PROMOTE_FUNCTION_RETURN | |
727 | Define this macro if the promotion described by @code{PROMOTE_MODE} | |
728 | should also be done for the return value of functions. | |
729 | ||
730 | If this macro is defined, @code{FUNCTION_VALUE} must perform the same | |
731 | promotions done by @code{PROMOTE_MODE}. | |
732 | ||
733 | @findex PROMOTE_FOR_CALL_ONLY | |
734 | @item PROMOTE_FOR_CALL_ONLY | |
735 | Define this macro if the promotion described by @code{PROMOTE_MODE} | |
736 | should @emph{only} be performed for outgoing function arguments or | |
737 | function return values, as specified by @code{PROMOTE_FUNCTION_ARGS} | |
738 | and @code{PROMOTE_FUNCTION_RETURN}, respectively. | |
739 | ||
740 | @findex PARM_BOUNDARY | |
741 | @item PARM_BOUNDARY | |
742 | Normal alignment required for function parameters on the stack, in | |
743 | bits. All stack parameters receive at least this much alignment | |
744 | regardless of data type. On most machines, this is the same as the | |
745 | size of an integer. | |
746 | ||
747 | @findex STACK_BOUNDARY | |
748 | @item STACK_BOUNDARY | |
749 | Define this macro if you wish to preserve a certain alignment for | |
750 | the stack pointer. The definition is a C expression | |
751 | for the desired alignment (measured in bits). | |
752 | ||
753 | @cindex @code{PUSH_ROUNDING}, interaction with @code{STACK_BOUNDARY} | |
754 | If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned | |
755 | to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies a | |
756 | less strict alignment than @code{STACK_BOUNDARY}, the stack may be | |
757 | momentarily unaligned while pushing arguments. | |
758 | ||
759 | @findex FUNCTION_BOUNDARY | |
760 | @item FUNCTION_BOUNDARY | |
761 | Alignment required for a function entry point, in bits. | |
762 | ||
763 | @findex BIGGEST_ALIGNMENT | |
764 | @item BIGGEST_ALIGNMENT | |
765 | Biggest alignment that any data type can require on this machine, in bits. | |
766 | ||
767 | @findex BIGGEST_FIELD_ALIGNMENT | |
768 | @item BIGGEST_FIELD_ALIGNMENT | |
769 | Biggest alignment that any structure field can require on this machine, | |
770 | in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for | |
771 | structure fields only. | |
772 | ||
773 | @findex ADJUST_FIELD_ALIGN | |
774 | @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed}) | |
775 | An expression for the alignment of a structure field @var{field} if the | |
776 | alignment computed in the usual way is @var{computed}. GNU CC uses | |
777 | this value instead of the value in @code{BIGGEST_ALIGNMENT} or | |
778 | @code{BIGGEST_FIELD_ALIGNMENT}, if defined, for structure fields only. | |
779 | ||
780 | @findex MAX_OFILE_ALIGNMENT | |
781 | @item MAX_OFILE_ALIGNMENT | |
782 | Biggest alignment supported by the object file format of this machine. | |
783 | Use this macro to limit the alignment which can be specified using the | |
784 | @code{__attribute__ ((aligned (@var{n})))} construct. If not defined, | |
785 | the default value is @code{BIGGEST_ALIGNMENT}. | |
786 | ||
787 | @findex DATA_ALIGNMENT | |
788 | @item DATA_ALIGNMENT (@var{type}, @var{basic-align}) | |
789 | If defined, a C expression to compute the alignment for a static | |
790 | variable. @var{type} is the data type, and @var{basic-align} is the | |
791 | alignment that the object would ordinarily have. The value of this | |
792 | macro is used instead of that alignment to align the object. | |
793 | ||
794 | If this macro is not defined, then @var{basic-align} is used. | |
795 | ||
796 | @findex strcpy | |
797 | One use of this macro is to increase alignment of medium-size data to | |
798 | make it all fit in fewer cache lines. Another is to cause character | |
799 | arrays to be word-aligned so that @code{strcpy} calls that copy | |
800 | constants to character arrays can be done inline. | |
801 | ||
802 | @findex CONSTANT_ALIGNMENT | |
803 | @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align}) | |
804 | If defined, a C expression to compute the alignment given to a constant | |
805 | that is being placed in memory. @var{constant} is the constant and | |
806 | @var{basic-align} is the alignment that the object would ordinarily | |
807 | have. The value of this macro is used instead of that alignment to | |
808 | align the object. | |
809 | ||
810 | If this macro is not defined, then @var{basic-align} is used. | |
811 | ||
812 | The typical use of this macro is to increase alignment for string | |
813 | constants to be word aligned so that @code{strcpy} calls that copy | |
814 | constants can be done inline. | |
815 | ||
816 | @findex EMPTY_FIELD_BOUNDARY | |
817 | @item EMPTY_FIELD_BOUNDARY | |
818 | Alignment in bits to be given to a structure bit field that follows an | |
819 | empty field such as @code{int : 0;}. | |
820 | ||
821 | Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment | |
822 | that results from an empty field. | |
823 | ||
824 | @findex STRUCTURE_SIZE_BOUNDARY | |
825 | @item STRUCTURE_SIZE_BOUNDARY | |
826 | Number of bits which any structure or union's size must be a multiple of. | |
827 | Each structure or union's size is rounded up to a multiple of this. | |
828 | ||
829 | If you do not define this macro, the default is the same as | |
830 | @code{BITS_PER_UNIT}. | |
831 | ||
832 | @findex STRICT_ALIGNMENT | |
833 | @item STRICT_ALIGNMENT | |
834 | Define this macro to be the value 1 if instructions will fail to work | |
835 | if given data not on the nominal alignment. If instructions will merely | |
836 | go slower in that case, define this macro as 0. | |
837 | ||
838 | @findex PCC_BITFIELD_TYPE_MATTERS | |
839 | @item PCC_BITFIELD_TYPE_MATTERS | |
840 | Define this if you wish to imitate the way many other C compilers handle | |
841 | alignment of bitfields and the structures that contain them. | |
842 | ||
843 | The behavior is that the type written for a bitfield (@code{int}, | |
844 | @code{short}, or other integer type) imposes an alignment for the | |
845 | entire structure, as if the structure really did contain an ordinary | |
846 | field of that type. In addition, the bitfield is placed within the | |
847 | structure so that it would fit within such a field, not crossing a | |
848 | boundary for it. | |
849 | ||
850 | Thus, on most machines, a bitfield whose type is written as @code{int} | |
851 | would not cross a four-byte boundary, and would force four-byte | |
852 | alignment for the whole structure. (The alignment used may not be four | |
853 | bytes; it is controlled by the other alignment parameters.) | |
854 | ||
855 | If the macro is defined, its definition should be a C expression; | |
856 | a nonzero value for the expression enables this behavior. | |
857 | ||
858 | Note that if this macro is not defined, or its value is zero, some | |
859 | bitfields may cross more than one alignment boundary. The compiler can | |
860 | support such references if there are @samp{insv}, @samp{extv}, and | |
861 | @samp{extzv} insns that can directly reference memory. | |
862 | ||
863 | The other known way of making bitfields work is to define | |
864 | @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}. | |
865 | Then every structure can be accessed with fullwords. | |
866 | ||
867 | Unless the machine has bitfield instructions or you define | |
868 | @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define | |
869 | @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value. | |
870 | ||
871 | If your aim is to make GNU CC use the same conventions for laying out | |
872 | bitfields as are used by another compiler, here is how to investigate | |
873 | what the other compiler does. Compile and run this program: | |
874 | ||
875 | @example | |
876 | struct foo1 | |
877 | @{ | |
878 | char x; | |
879 | char :0; | |
880 | char y; | |
881 | @}; | |
882 | ||
883 | struct foo2 | |
884 | @{ | |
885 | char x; | |
886 | int :0; | |
887 | char y; | |
888 | @}; | |
889 | ||
890 | main () | |
891 | @{ | |
892 | printf ("Size of foo1 is %d\n", | |
893 | sizeof (struct foo1)); | |
894 | printf ("Size of foo2 is %d\n", | |
895 | sizeof (struct foo2)); | |
896 | exit (0); | |
897 | @} | |
898 | @end example | |
899 | ||
900 | If this prints 2 and 5, then the compiler's behavior is what you would | |
901 | get from @code{PCC_BITFIELD_TYPE_MATTERS}. | |
902 | ||
903 | @findex BITFIELD_NBYTES_LIMITED | |
904 | @item BITFIELD_NBYTES_LIMITED | |
905 | Like PCC_BITFIELD_TYPE_MATTERS except that its effect is limited to | |
906 | aligning a bitfield within the structure. | |
907 | ||
908 | @findex ROUND_TYPE_SIZE | |
909 | @item ROUND_TYPE_SIZE (@var{struct}, @var{size}, @var{align}) | |
910 | Define this macro as an expression for the overall size of a structure | |
911 | (given by @var{struct} as a tree node) when the size computed from the | |
912 | fields is @var{size} and the alignment is @var{align}. | |
913 | ||
914 | The default is to round @var{size} up to a multiple of @var{align}. | |
915 | ||
916 | @findex ROUND_TYPE_ALIGN | |
917 | @item ROUND_TYPE_ALIGN (@var{struct}, @var{computed}, @var{specified}) | |
918 | Define this macro as an expression for the alignment of a structure | |
919 | (given by @var{struct} as a tree node) if the alignment computed in the | |
920 | usual way is @var{computed} and the alignment explicitly specified was | |
921 | @var{specified}. | |
922 | ||
923 | The default is to use @var{specified} if it is larger; otherwise, use | |
924 | the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT} | |
925 | ||
926 | @findex MAX_FIXED_MODE_SIZE | |
927 | @item MAX_FIXED_MODE_SIZE | |
928 | An integer expression for the size in bits of the largest integer | |
929 | machine mode that should actually be used. All integer machine modes of | |
930 | this size or smaller can be used for structures and unions with the | |
931 | appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE | |
932 | (DImode)} is assumed. | |
933 | ||
934 | @findex CHECK_FLOAT_VALUE | |
935 | @item CHECK_FLOAT_VALUE (@var{mode}, @var{value}, @var{overflow}) | |
936 | A C statement to validate the value @var{value} (of type | |
937 | @code{double}) for mode @var{mode}. This means that you check whether | |
938 | @var{value} fits within the possible range of values for mode | |
939 | @var{mode} on this target machine. The mode @var{mode} is always | |
940 | a mode of class @code{MODE_FLOAT}. @var{overflow} is nonzero if | |
941 | the value is already known to be out of range. | |
942 | ||
943 | If @var{value} is not valid or if @var{overflow} is nonzero, you should | |
944 | set @var{overflow} to 1 and then assign some valid value to @var{value}. | |
945 | Allowing an invalid value to go through the compiler can produce | |
946 | incorrect assembler code which may even cause Unix assemblers to crash. | |
947 | ||
948 | This macro need not be defined if there is no work for it to do. | |
949 | ||
950 | @findex TARGET_FLOAT_FORMAT | |
951 | @item TARGET_FLOAT_FORMAT | |
952 | A code distinguishing the floating point format of the target machine. | |
953 | There are three defined values: | |
954 | ||
955 | @table @code | |
956 | @findex IEEE_FLOAT_FORMAT | |
957 | @item IEEE_FLOAT_FORMAT | |
958 | This code indicates IEEE floating point. It is the default; there is no | |
959 | need to define this macro when the format is IEEE. | |
960 | ||
961 | @findex VAX_FLOAT_FORMAT | |
962 | @item VAX_FLOAT_FORMAT | |
963 | This code indicates the peculiar format used on the Vax. | |
964 | ||
965 | @findex UNKNOWN_FLOAT_FORMAT | |
966 | @item UNKNOWN_FLOAT_FORMAT | |
967 | This code indicates any other format. | |
968 | @end table | |
969 | ||
970 | The value of this macro is compared with @code{HOST_FLOAT_FORMAT} | |
971 | (@pxref{Config}) to determine whether the target machine has the same | |
972 | format as the host machine. If any other formats are actually in use on | |
973 | supported machines, new codes should be defined for them. | |
974 | ||
975 | The ordering of the component words of floating point values stored in | |
976 | memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN} for the target | |
977 | machine and @code{HOST_FLOAT_WORDS_BIG_ENDIAN} for the host. | |
978 | @end table | |
979 | ||
980 | @node Type Layout | |
981 | @section Layout of Source Language Data Types | |
982 | ||
983 | These macros define the sizes and other characteristics of the standard | |
984 | basic data types used in programs being compiled. Unlike the macros in | |
985 | the previous section, these apply to specific features of C and related | |
986 | languages, rather than to fundamental aspects of storage layout. | |
987 | ||
988 | @table @code | |
989 | @findex INT_TYPE_SIZE | |
990 | @item INT_TYPE_SIZE | |
991 | A C expression for the size in bits of the type @code{int} on the | |
992 | target machine. If you don't define this, the default is one word. | |
993 | ||
994 | @findex MAX_INT_TYPE_SIZE | |
995 | @item MAX_INT_TYPE_SIZE | |
996 | Maximum number for the size in bits of the type @code{int} on the target | |
997 | machine. If this is undefined, the default is @code{INT_TYPE_SIZE}. | |
998 | Otherwise, it is the constant value that is the largest value that | |
999 | @code{INT_TYPE_SIZE} can have at run-time. This is used in @code{cpp}. | |
1000 | ||
1001 | @findex SHORT_TYPE_SIZE | |
1002 | @item SHORT_TYPE_SIZE | |
1003 | A C expression for the size in bits of the type @code{short} on the | |
1004 | target machine. If you don't define this, the default is half a word. | |
1005 | (If this would be less than one storage unit, it is rounded up to one | |
1006 | unit.) | |
1007 | ||
1008 | @findex LONG_TYPE_SIZE | |
1009 | @item LONG_TYPE_SIZE | |
1010 | A C expression for the size in bits of the type @code{long} on the | |
1011 | target machine. If you don't define this, the default is one word. | |
1012 | ||
1013 | @findex MAX_LONG_TYPE_SIZE | |
1014 | @item MAX_LONG_TYPE_SIZE | |
1015 | Maximum number for the size in bits of the type @code{long} on the | |
1016 | target machine. If this is undefined, the default is | |
1017 | @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the | |
1018 | largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is | |
1019 | used in @code{cpp}. | |
1020 | ||
1021 | @findex LONG_LONG_TYPE_SIZE | |
1022 | @item LONG_LONG_TYPE_SIZE | |
1023 | A C expression for the size in bits of the type @code{long long} on the | |
1024 | target machine. If you don't define this, the default is two | |
1025 | words. If you want to support GNU Ada on your machine, the value of | |
1026 | macro must be at least 64. | |
1027 | ||
1028 | @findex CHAR_TYPE_SIZE | |
1029 | @item CHAR_TYPE_SIZE | |
1030 | A C expression for the size in bits of the type @code{char} on the | |
1031 | target machine. If you don't define this, the default is one quarter | |
1032 | of a word. (If this would be less than one storage unit, it is rounded up | |
1033 | to one unit.) | |
1034 | ||
1035 | @findex MAX_CHAR_TYPE_SIZE | |
1036 | @item MAX_CHAR_TYPE_SIZE | |
1037 | Maximum number for the size in bits of the type @code{char} on the | |
1038 | target machine. If this is undefined, the default is | |
1039 | @code{CHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the | |
1040 | largest value that @code{CHAR_TYPE_SIZE} can have at run-time. This is | |
1041 | used in @code{cpp}. | |
1042 | ||
1043 | @findex FLOAT_TYPE_SIZE | |
1044 | @item FLOAT_TYPE_SIZE | |
1045 | A C expression for the size in bits of the type @code{float} on the | |
1046 | target machine. If you don't define this, the default is one word. | |
1047 | ||
1048 | @findex DOUBLE_TYPE_SIZE | |
1049 | @item DOUBLE_TYPE_SIZE | |
1050 | A C expression for the size in bits of the type @code{double} on the | |
1051 | target machine. If you don't define this, the default is two | |
1052 | words. | |
1053 | ||
1054 | @findex LONG_DOUBLE_TYPE_SIZE | |
1055 | @item LONG_DOUBLE_TYPE_SIZE | |
1056 | A C expression for the size in bits of the type @code{long double} on | |
1057 | the target machine. If you don't define this, the default is two | |
1058 | words. | |
1059 | ||
1060 | @findex DEFAULT_SIGNED_CHAR | |
1061 | @item DEFAULT_SIGNED_CHAR | |
1062 | An expression whose value is 1 or 0, according to whether the type | |
1063 | @code{char} should be signed or unsigned by default. The user can | |
1064 | always override this default with the options @samp{-fsigned-char} | |
1065 | and @samp{-funsigned-char}. | |
1066 | ||
1067 | @findex DEFAULT_SHORT_ENUMS | |
1068 | @item DEFAULT_SHORT_ENUMS | |
1069 | A C expression to determine whether to give an @code{enum} type | |
1070 | only as many bytes as it takes to represent the range of possible values | |
1071 | of that type. A nonzero value means to do that; a zero value means all | |
1072 | @code{enum} types should be allocated like @code{int}. | |
1073 | ||
1074 | If you don't define the macro, the default is 0. | |
1075 | ||
1076 | @findex SIZE_TYPE | |
1077 | @item SIZE_TYPE | |
1078 | A C expression for a string describing the name of the data type to use | |
1079 | for size values. The typedef name @code{size_t} is defined using the | |
1080 | contents of the string. | |
1081 | ||
1082 | The string can contain more than one keyword. If so, separate them with | |
1083 | spaces, and write first any length keyword, then @code{unsigned} if | |
1084 | appropriate, and finally @code{int}. The string must exactly match one | |
1085 | of the data type names defined in the function | |
1086 | @code{init_decl_processing} in the file @file{c-decl.c}. You may not | |
1087 | omit @code{int} or change the order---that would cause the compiler to | |
1088 | crash on startup. | |
1089 | ||
1090 | If you don't define this macro, the default is @code{"long unsigned | |
1091 | int"}. | |
1092 | ||
1093 | @findex PTRDIFF_TYPE | |
1094 | @item PTRDIFF_TYPE | |
1095 | A C expression for a string describing the name of the data type to use | |
1096 | for the result of subtracting two pointers. The typedef name | |
1097 | @code{ptrdiff_t} is defined using the contents of the string. See | |
1098 | @code{SIZE_TYPE} above for more information. | |
1099 | ||
1100 | If you don't define this macro, the default is @code{"long int"}. | |
1101 | ||
1102 | @findex WCHAR_TYPE | |
1103 | @item WCHAR_TYPE | |
1104 | A C expression for a string describing the name of the data type to use | |
1105 | for wide characters. The typedef name @code{wchar_t} is defined using | |
1106 | the contents of the string. See @code{SIZE_TYPE} above for more | |
1107 | information. | |
1108 | ||
1109 | If you don't define this macro, the default is @code{"int"}. | |
1110 | ||
1111 | @findex WCHAR_TYPE_SIZE | |
1112 | @item WCHAR_TYPE_SIZE | |
1113 | A C expression for the size in bits of the data type for wide | |
1114 | characters. This is used in @code{cpp}, which cannot make use of | |
1115 | @code{WCHAR_TYPE}. | |
1116 | ||
1117 | @findex MAX_WCHAR_TYPE_SIZE | |
1118 | @item MAX_WCHAR_TYPE_SIZE | |
1119 | Maximum number for the size in bits of the data type for wide | |
1120 | characters. If this is undefined, the default is | |
1121 | @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the | |
1122 | largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is | |
1123 | used in @code{cpp}. | |
1124 | ||
1125 | @findex OBJC_INT_SELECTORS | |
1126 | @item OBJC_INT_SELECTORS | |
1127 | Define this macro if the type of Objective C selectors should be | |
1128 | @code{int}. | |
1129 | ||
1130 | If this macro is not defined, then selectors should have the type | |
1131 | @code{struct objc_selector *}. | |
1132 | ||
1133 | @findex OBJC_SELECTORS_WITHOUT_LABELS | |
1134 | @item OBJC_SELECTORS_WITHOUT_LABELS | |
1135 | Define this macro if the compiler can group all the selectors together | |
1136 | into a vector and use just one label at the beginning of the vector. | |
1137 | Otherwise, the compiler must give each selector its own assembler | |
1138 | label. | |
1139 | ||
1140 | On certain machines, it is important to have a separate label for each | |
1141 | selector because this enables the linker to eliminate duplicate selectors. | |
1142 | ||
1143 | @findex TARGET_BELL | |
1144 | @item TARGET_BELL | |
1145 | A C constant expression for the integer value for escape sequence | |
1146 | @samp{\a}. | |
1147 | ||
1148 | @findex TARGET_TAB | |
1149 | @findex TARGET_BS | |
1150 | @findex TARGET_NEWLINE | |
1151 | @item TARGET_BS | |
1152 | @itemx TARGET_TAB | |
1153 | @itemx TARGET_NEWLINE | |
1154 | C constant expressions for the integer values for escape sequences | |
1155 | @samp{\b}, @samp{\t} and @samp{\n}. | |
1156 | ||
1157 | @findex TARGET_VT | |
1158 | @findex TARGET_FF | |
1159 | @findex TARGET_CR | |
1160 | @item TARGET_VT | |
1161 | @itemx TARGET_FF | |
1162 | @itemx TARGET_CR | |
1163 | C constant expressions for the integer values for escape sequences | |
1164 | @samp{\v}, @samp{\f} and @samp{\r}. | |
1165 | @end table | |
1166 | ||
1167 | @node Registers | |
1168 | @section Register Usage | |
1169 | @cindex register usage | |
1170 | ||
1171 | This section explains how to describe what registers the target machine | |
1172 | has, and how (in general) they can be used. | |
1173 | ||
1174 | The description of which registers a specific instruction can use is | |
1175 | done with register classes; see @ref{Register Classes}. For information | |
1176 | on using registers to access a stack frame, see @ref{Frame Registers}. | |
1177 | For passing values in registers, see @ref{Register Arguments}. | |
1178 | For returning values in registers, see @ref{Scalar Return}. | |
1179 | ||
1180 | @menu | |
1181 | * Register Basics:: Number and kinds of registers. | |
1182 | * Allocation Order:: Order in which registers are allocated. | |
1183 | * Values in Registers:: What kinds of values each reg can hold. | |
1184 | * Leaf Functions:: Renumbering registers for leaf functions. | |
1185 | * Stack Registers:: Handling a register stack such as 80387. | |
1186 | * Obsolete Register Macros:: Macros formerly used for the 80387. | |
1187 | @end menu | |
1188 | ||
1189 | @node Register Basics | |
1190 | @subsection Basic Characteristics of Registers | |
1191 | ||
1192 | @c prevent bad page break with this line | |
1193 | Registers have various characteristics. | |
1194 | ||
1195 | @table @code | |
1196 | @findex FIRST_PSEUDO_REGISTER | |
1197 | @item FIRST_PSEUDO_REGISTER | |
1198 | Number of hardware registers known to the compiler. They receive | |
1199 | numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first | |
1200 | pseudo register's number really is assigned the number | |
1201 | @code{FIRST_PSEUDO_REGISTER}. | |
1202 | ||
1203 | @item FIXED_REGISTERS | |
1204 | @findex FIXED_REGISTERS | |
1205 | @cindex fixed register | |
1206 | An initializer that says which registers are used for fixed purposes | |
1207 | all throughout the compiled code and are therefore not available for | |
1208 | general allocation. These would include the stack pointer, the frame | |
1209 | pointer (except on machines where that can be used as a general | |
1210 | register when no frame pointer is needed), the program counter on | |
1211 | machines where that is considered one of the addressable registers, | |
1212 | and any other numbered register with a standard use. | |
1213 | ||
1214 | This information is expressed as a sequence of numbers, separated by | |
1215 | commas and surrounded by braces. The @var{n}th number is 1 if | |
1216 | register @var{n} is fixed, 0 otherwise. | |
1217 | ||
1218 | The table initialized from this macro, and the table initialized by | |
1219 | the following one, may be overridden at run time either automatically, | |
1220 | by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by | |
1221 | the user with the command options @samp{-ffixed-@var{reg}}, | |
1222 | @samp{-fcall-used-@var{reg}} and @samp{-fcall-saved-@var{reg}}. | |
1223 | ||
1224 | @findex CALL_USED_REGISTERS | |
1225 | @item CALL_USED_REGISTERS | |
1226 | @cindex call-used register | |
1227 | @cindex call-clobbered register | |
1228 | @cindex call-saved register | |
1229 | Like @code{FIXED_REGISTERS} but has 1 for each register that is | |
1230 | clobbered (in general) by function calls as well as for fixed | |
1231 | registers. This macro therefore identifies the registers that are not | |
1232 | available for general allocation of values that must live across | |
1233 | function calls. | |
1234 | ||
1235 | If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler | |
1236 | automatically saves it on function entry and restores it on function | |
1237 | exit, if the register is used within the function. | |
1238 | ||
1239 | @findex CONDITIONAL_REGISTER_USAGE | |
1240 | @findex fixed_regs | |
1241 | @findex call_used_regs | |
1242 | @item CONDITIONAL_REGISTER_USAGE | |
1243 | Zero or more C statements that may conditionally modify two variables | |
1244 | @code{fixed_regs} and @code{call_used_regs} (both of type @code{char | |
1245 | []}) after they have been initialized from the two preceding macros. | |
1246 | ||
1247 | This is necessary in case the fixed or call-clobbered registers depend | |
1248 | on target flags. | |
1249 | ||
1250 | You need not define this macro if it has no work to do. | |
1251 | ||
1252 | @cindex disabling certain registers | |
1253 | @cindex controlling register usage | |
1254 | If the usage of an entire class of registers depends on the target | |
1255 | flags, you may indicate this to GCC by using this macro to modify | |
1256 | @code{fixed_regs} and @code{call_used_regs} to 1 for each of the | |
1257 | registers in the classes which should not be used by GCC. Also define | |
1258 | the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it | |
1259 | is called with a letter for a class that shouldn't be used. | |
1260 | ||
1261 | (However, if this class is not included in @code{GENERAL_REGS} and all | |
1262 | of the insn patterns whose constraints permit this class are | |
1263 | controlled by target switches, then GCC will automatically avoid using | |
1264 | these registers when the target switches are opposed to them.) | |
1265 | ||
1266 | @findex NON_SAVING_SETJMP | |
1267 | @item NON_SAVING_SETJMP | |
1268 | If this macro is defined and has a nonzero value, it means that | |
1269 | @code{setjmp} and related functions fail to save the registers, or that | |
1270 | @code{longjmp} fails to restore them. To compensate, the compiler | |
1271 | avoids putting variables in registers in functions that use | |
1272 | @code{setjmp}. | |
1273 | ||
1274 | @findex INCOMING_REGNO | |
1275 | @item INCOMING_REGNO (@var{out}) | |
1276 | Define this macro if the target machine has register windows. This C | |
1277 | expression returns the register number as seen by the called function | |
1278 | corresponding to the register number @var{out} as seen by the calling | |
1279 | function. Return @var{out} if register number @var{out} is not an | |
1280 | outbound register. | |
1281 | ||
1282 | @findex OUTGOING_REGNO | |
1283 | @item OUTGOING_REGNO (@var{in}) | |
1284 | Define this macro if the target machine has register windows. This C | |
1285 | expression returns the register number as seen by the calling function | |
1286 | corresponding to the register number @var{in} as seen by the called | |
1287 | function. Return @var{in} if register number @var{in} is not an inbound | |
1288 | register. | |
1289 | ||
1290 | @ignore | |
1291 | @findex PC_REGNUM | |
1292 | @item PC_REGNUM | |
1293 | If the program counter has a register number, define this as that | |
1294 | register number. Otherwise, do not define it. | |
1295 | @end ignore | |
1296 | @end table | |
1297 | ||
1298 | @node Allocation Order | |
1299 | @subsection Order of Allocation of Registers | |
1300 | @cindex order of register allocation | |
1301 | @cindex register allocation order | |
1302 | ||
1303 | @c prevent bad page break with this line | |
1304 | Registers are allocated in order. | |
1305 | ||
1306 | @table @code | |
1307 | @findex REG_ALLOC_ORDER | |
1308 | @item REG_ALLOC_ORDER | |
1309 | If defined, an initializer for a vector of integers, containing the | |
1310 | numbers of hard registers in the order in which GNU CC should prefer | |
1311 | to use them (from most preferred to least). | |
1312 | ||
1313 | If this macro is not defined, registers are used lowest numbered first | |
1314 | (all else being equal). | |
1315 | ||
1316 | One use of this macro is on machines where the highest numbered | |
1317 | registers must always be saved and the save-multiple-registers | |
1318 | instruction supports only sequences of consecutive registers. On such | |
1319 | machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists | |
1320 | the highest numbered allocatable register first. | |
1321 | ||
1322 | @findex ORDER_REGS_FOR_LOCAL_ALLOC | |
1323 | @item ORDER_REGS_FOR_LOCAL_ALLOC | |
1324 | A C statement (sans semicolon) to choose the order in which to allocate | |
1325 | hard registers for pseudo-registers local to a basic block. | |
1326 | ||
1327 | Store the desired register order in the array @code{reg_alloc_order}. | |
1328 | Element 0 should be the register to allocate first; element 1, the next | |
1329 | register; and so on. | |
1330 | ||
1331 | The macro body should not assume anything about the contents of | |
1332 | @code{reg_alloc_order} before execution of the macro. | |
1333 | ||
1334 | On most machines, it is not necessary to define this macro. | |
1335 | @end table | |
1336 | ||
1337 | @node Values in Registers | |
1338 | @subsection How Values Fit in Registers | |
1339 | ||
1340 | This section discusses the macros that describe which kinds of values | |
1341 | (specifically, which machine modes) each register can hold, and how many | |
1342 | consecutive registers are needed for a given mode. | |
1343 | ||
1344 | @table @code | |
1345 | @findex HARD_REGNO_NREGS | |
1346 | @item HARD_REGNO_NREGS (@var{regno}, @var{mode}) | |
1347 | A C expression for the number of consecutive hard registers, starting | |
1348 | at register number @var{regno}, required to hold a value of mode | |
1349 | @var{mode}. | |
1350 | ||
1351 | On a machine where all registers are exactly one word, a suitable | |
1352 | definition of this macro is | |
1353 | ||
1354 | @smallexample | |
1355 | #define HARD_REGNO_NREGS(REGNO, MODE) \ | |
1356 | ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \ | |
1357 | / UNITS_PER_WORD)) | |
1358 | @end smallexample | |
1359 | ||
1360 | @findex HARD_REGNO_MODE_OK | |
1361 | @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode}) | |
1362 | A C expression that is nonzero if it is permissible to store a value | |
1363 | of mode @var{mode} in hard register number @var{regno} (or in several | |
1364 | registers starting with that one). For a machine where all registers | |
1365 | are equivalent, a suitable definition is | |
1366 | ||
1367 | @smallexample | |
1368 | #define HARD_REGNO_MODE_OK(REGNO, MODE) 1 | |
1369 | @end smallexample | |
1370 | ||
1371 | It is not necessary for this macro to check for the numbers of fixed | |
1372 | registers, because the allocation mechanism considers them to be always | |
1373 | occupied. | |
1374 | ||
1375 | @cindex register pairs | |
1376 | On some machines, double-precision values must be kept in even/odd | |
1377 | register pairs. The way to implement that is to define this macro | |
1378 | to reject odd register numbers for such modes. | |
1379 | ||
1380 | @ignore | |
1381 | @c I think this is not true now | |
1382 | GNU CC assumes that it can always move values between registers and | |
1383 | (suitably addressed) memory locations. If it is impossible to move a | |
1384 | value of a certain mode between memory and certain registers, then | |
1385 | @code{HARD_REGNO_MODE_OK} must not allow this mode in those registers. | |
1386 | @end ignore | |
1387 | ||
1388 | The minimum requirement for a mode to be OK in a register is that the | |
1389 | @samp{mov@var{mode}} instruction pattern support moves between the | |
1390 | register and any other hard register for which the mode is OK; and that | |
1391 | moving a value into the register and back out not alter it. | |
1392 | ||
1393 | Since the same instruction used to move @code{SImode} will work for all | |
1394 | narrower integer modes, it is not necessary on any machine for | |
1395 | @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided | |
1396 | you define patterns @samp{movhi}, etc., to take advantage of this. This | |
1397 | is useful because of the interaction between @code{HARD_REGNO_MODE_OK} | |
1398 | and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes | |
1399 | to be tieable. | |
1400 | ||
1401 | Many machines have special registers for floating point arithmetic. | |
1402 | Often people assume that floating point machine modes are allowed only | |
1403 | in floating point registers. This is not true. Any registers that | |
1404 | can hold integers can safely @emph{hold} a floating point machine | |
1405 | mode, whether or not floating arithmetic can be done on it in those | |
1406 | registers. Integer move instructions can be used to move the values. | |
1407 | ||
1408 | On some machines, though, the converse is true: fixed-point machine | |
1409 | modes may not go in floating registers. This is true if the floating | |
1410 | registers normalize any value stored in them, because storing a | |
1411 | non-floating value there would garble it. In this case, | |
1412 | @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in | |
1413 | floating registers. But if the floating registers do not automatically | |
1414 | normalize, if you can store any bit pattern in one and retrieve it | |
1415 | unchanged without a trap, then any machine mode may go in a floating | |
1416 | register, so you can define this macro to say so. | |
1417 | ||
1418 | The primary significance of special floating registers is rather that | |
1419 | they are the registers acceptable in floating point arithmetic | |
1420 | instructions. However, this is of no concern to | |
1421 | @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper | |
1422 | constraints for those instructions. | |
1423 | ||
1424 | On some machines, the floating registers are especially slow to access, | |
1425 | so that it is better to store a value in a stack frame than in such a | |
1426 | register if floating point arithmetic is not being done. As long as the | |
1427 | floating registers are not in class @code{GENERAL_REGS}, they will not | |
1428 | be used unless some pattern's constraint asks for one. | |
1429 | ||
1430 | @findex MODES_TIEABLE_P | |
1431 | @item MODES_TIEABLE_P (@var{mode1}, @var{mode2}) | |
1432 | A C expression that is nonzero if it is desirable to choose register | |
1433 | allocation so as to avoid move instructions between a value of mode | |
1434 | @var{mode1} and a value of mode @var{mode2}. | |
1435 | ||
1436 | If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and | |
1437 | @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are ever different | |
1438 | for any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, | |
1439 | @var{mode2})} must be zero. | |
1440 | @end table | |
1441 | ||
1442 | @node Leaf Functions | |
1443 | @subsection Handling Leaf Functions | |
1444 | ||
1445 | @cindex leaf functions | |
1446 | @cindex functions, leaf | |
1447 | On some machines, a leaf function (i.e., one which makes no calls) can run | |
1448 | more efficiently if it does not make its own register window. Often this | |
1449 | means it is required to receive its arguments in the registers where they | |
1450 | are passed by the caller, instead of the registers where they would | |
1451 | normally arrive. | |
1452 | ||
1453 | The special treatment for leaf functions generally applies only when | |
1454 | other conditions are met; for example, often they may use only those | |
1455 | registers for its own variables and temporaries. We use the term ``leaf | |
1456 | function'' to mean a function that is suitable for this special | |
1457 | handling, so that functions with no calls are not necessarily ``leaf | |
1458 | functions''. | |
1459 | ||
1460 | GNU CC assigns register numbers before it knows whether the function is | |
1461 | suitable for leaf function treatment. So it needs to renumber the | |
1462 | registers in order to output a leaf function. The following macros | |
1463 | accomplish this. | |
1464 | ||
1465 | @table @code | |
1466 | @findex LEAF_REGISTERS | |
1467 | @item LEAF_REGISTERS | |
1468 | A C initializer for a vector, indexed by hard register number, which | |
1469 | contains 1 for a register that is allowable in a candidate for leaf | |
1470 | function treatment. | |
1471 | ||
1472 | If leaf function treatment involves renumbering the registers, then the | |
1473 | registers marked here should be the ones before renumbering---those that | |
1474 | GNU CC would ordinarily allocate. The registers which will actually be | |
1475 | used in the assembler code, after renumbering, should not be marked with 1 | |
1476 | in this vector. | |
1477 | ||
1478 | Define this macro only if the target machine offers a way to optimize | |
1479 | the treatment of leaf functions. | |
1480 | ||
1481 | @findex LEAF_REG_REMAP | |
1482 | @item LEAF_REG_REMAP (@var{regno}) | |
1483 | A C expression whose value is the register number to which @var{regno} | |
1484 | should be renumbered, when a function is treated as a leaf function. | |
1485 | ||
1486 | If @var{regno} is a register number which should not appear in a leaf | |
1487 | function before renumbering, then the expression should yield -1, which | |
1488 | will cause the compiler to abort. | |
1489 | ||
1490 | Define this macro only if the target machine offers a way to optimize the | |
1491 | treatment of leaf functions, and registers need to be renumbered to do | |
1492 | this. | |
1493 | @end table | |
1494 | ||
1495 | @findex leaf_function | |
1496 | Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must | |
1497 | treat leaf functions specially. It can test the C variable | |
1498 | @code{leaf_function} which is nonzero for leaf functions. (The variable | |
1499 | @code{leaf_function} is defined only if @code{LEAF_REGISTERS} is | |
1500 | defined.) | |
1501 | @c changed this to fix overfull. ALSO: why the "it" at the beginning | |
1502 | @c of the next paragraph?! --mew 2feb93 | |
1503 | ||
1504 | @node Stack Registers | |
1505 | @subsection Registers That Form a Stack | |
1506 | ||
1507 | There are special features to handle computers where some of the | |
1508 | ``registers'' form a stack, as in the 80387 coprocessor for the 80386. | |
1509 | Stack registers are normally written by pushing onto the stack, and are | |
1510 | numbered relative to the top of the stack. | |
1511 | ||
1512 | Currently, GNU CC can only handle one group of stack-like registers, and | |
1513 | they must be consecutively numbered. | |
1514 | ||
1515 | @table @code | |
1516 | @findex STACK_REGS | |
1517 | @item STACK_REGS | |
1518 | Define this if the machine has any stack-like registers. | |
1519 | ||
1520 | @findex FIRST_STACK_REG | |
1521 | @item FIRST_STACK_REG | |
1522 | The number of the first stack-like register. This one is the top | |
1523 | of the stack. | |
1524 | ||
1525 | @findex LAST_STACK_REG | |
1526 | @item LAST_STACK_REG | |
1527 | The number of the last stack-like register. This one is the bottom of | |
1528 | the stack. | |
1529 | @end table | |
1530 | ||
1531 | @node Obsolete Register Macros | |
1532 | @subsection Obsolete Macros for Controlling Register Usage | |
1533 | ||
1534 | These features do not work very well. They exist because they used to | |
1535 | be required to generate correct code for the 80387 coprocessor of the | |
1536 | 80386. They are no longer used by that machine description and may be | |
1537 | removed in a later version of the compiler. Don't use them! | |
1538 | ||
1539 | @table @code | |
1540 | @findex OVERLAPPING_REGNO_P | |
1541 | @item OVERLAPPING_REGNO_P (@var{regno}) | |
1542 | If defined, this is a C expression whose value is nonzero if hard | |
1543 | register number @var{regno} is an overlapping register. This means a | |
1544 | hard register which overlaps a hard register with a different number. | |
1545 | (Such overlap is undesirable, but occasionally it allows a machine to | |
1546 | be supported which otherwise could not be.) This macro must return | |
1547 | nonzero for @emph{all} the registers which overlap each other. GNU CC | |
1548 | can use an overlapping register only in certain limited ways. It can | |
1549 | be used for allocation within a basic block, and may be spilled for | |
1550 | reloading; that is all. | |
1551 | ||
1552 | If this macro is not defined, it means that none of the hard registers | |
1553 | overlap each other. This is the usual situation. | |
1554 | ||
1555 | @findex INSN_CLOBBERS_REGNO_P | |
1556 | @item INSN_CLOBBERS_REGNO_P (@var{insn}, @var{regno}) | |
1557 | If defined, this is a C expression whose value should be nonzero if | |
1558 | the insn @var{insn} has the effect of mysteriously clobbering the | |
1559 | contents of hard register number @var{regno}. By ``mysterious'' we | |
1560 | mean that the insn's RTL expression doesn't describe such an effect. | |
1561 | ||
1562 | If this macro is not defined, it means that no insn clobbers registers | |
1563 | mysteriously. This is the usual situation; all else being equal, | |
1564 | it is best for the RTL expression to show all the activity. | |
1565 | ||
1566 | @cindex death notes | |
1567 | @findex PRESERVE_DEATH_INFO_REGNO_P | |
1568 | @item PRESERVE_DEATH_INFO_REGNO_P (@var{regno}) | |
1569 | If defined, this is a C expression whose value is nonzero if correct | |
1570 | @code{REG_DEAD} notes are needed for hard register number @var{regno} | |
1571 | after reload. | |
1572 | ||
1573 | You would arrange to preserve death info for a register when some of the | |
1574 | code in the machine description which is executed to write the assembler | |
1575 | code looks at the death notes. This is necessary only when the actual | |
1576 | hardware feature which GNU CC thinks of as a register is not actually a | |
1577 | register of the usual sort. (It might, for example, be a hardware | |
1578 | stack.) | |
1579 | ||
1580 | It is also useful for peepholes and linker relaxation. | |
1581 | ||
1582 | If this macro is not defined, it means that no death notes need to be | |
1583 | preserved, and some may even be incorrect. This is the usual situation. | |
1584 | @end table | |
1585 | ||
1586 | @node Register Classes | |
1587 | @section Register Classes | |
1588 | @cindex register class definitions | |
1589 | @cindex class definitions, register | |
1590 | ||
1591 | On many machines, the numbered registers are not all equivalent. | |
1592 | For example, certain registers may not be allowed for indexed addressing; | |
1593 | certain registers may not be allowed in some instructions. These machine | |
1594 | restrictions are described to the compiler using @dfn{register classes}. | |
1595 | ||
1596 | You define a number of register classes, giving each one a name and saying | |
1597 | which of the registers belong to it. Then you can specify register classes | |
1598 | that are allowed as operands to particular instruction patterns. | |
1599 | ||
1600 | @findex ALL_REGS | |
1601 | @findex NO_REGS | |
1602 | In general, each register will belong to several classes. In fact, one | |
1603 | class must be named @code{ALL_REGS} and contain all the registers. Another | |
1604 | class must be named @code{NO_REGS} and contain no registers. Often the | |
1605 | union of two classes will be another class; however, this is not required. | |
1606 | ||
1607 | @findex GENERAL_REGS | |
1608 | One of the classes must be named @code{GENERAL_REGS}. There is nothing | |
1609 | terribly special about the name, but the operand constraint letters | |
1610 | @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is | |
1611 | the same as @code{ALL_REGS}, just define it as a macro which expands | |
1612 | to @code{ALL_REGS}. | |
1613 | ||
1614 | Order the classes so that if class @var{x} is contained in class @var{y} | |
1615 | then @var{x} has a lower class number than @var{y}. | |
1616 | ||
1617 | The way classes other than @code{GENERAL_REGS} are specified in operand | |
1618 | constraints is through machine-dependent operand constraint letters. | |
1619 | You can define such letters to correspond to various classes, then use | |
1620 | them in operand constraints. | |
1621 | ||
1622 | You should define a class for the union of two classes whenever some | |
1623 | instruction allows both classes. For example, if an instruction allows | |
1624 | either a floating point (coprocessor) register or a general register for a | |
1625 | certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS} | |
1626 | which includes both of them. Otherwise you will get suboptimal code. | |
1627 | ||
1628 | You must also specify certain redundant information about the register | |
1629 | classes: for each class, which classes contain it and which ones are | |
1630 | contained in it; for each pair of classes, the largest class contained | |
1631 | in their union. | |
1632 | ||
1633 | When a value occupying several consecutive registers is expected in a | |
1634 | certain class, all the registers used must belong to that class. | |
1635 | Therefore, register classes cannot be used to enforce a requirement for | |
1636 | a register pair to start with an even-numbered register. The way to | |
1637 | specify this requirement is with @code{HARD_REGNO_MODE_OK}. | |
1638 | ||
1639 | Register classes used for input-operands of bitwise-and or shift | |
1640 | instructions have a special requirement: each such class must have, for | |
1641 | each fixed-point machine mode, a subclass whose registers can transfer that | |
1642 | mode to or from memory. For example, on some machines, the operations for | |
1643 | single-byte values (@code{QImode}) are limited to certain registers. When | |
1644 | this is so, each register class that is used in a bitwise-and or shift | |
1645 | instruction must have a subclass consisting of registers from which | |
1646 | single-byte values can be loaded or stored. This is so that | |
1647 | @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return. | |
1648 | ||
1649 | @table @code | |
1650 | @findex enum reg_class | |
1651 | @item enum reg_class | |
1652 | An enumeral type that must be defined with all the register class names | |
1653 | as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS} | |
1654 | must be the last register class, followed by one more enumeral value, | |
1655 | @code{LIM_REG_CLASSES}, which is not a register class but rather | |
1656 | tells how many classes there are. | |
1657 | ||
1658 | Each register class has a number, which is the value of casting | |
1659 | the class name to type @code{int}. The number serves as an index | |
1660 | in many of the tables described below. | |
1661 | ||
1662 | @findex N_REG_CLASSES | |
1663 | @item N_REG_CLASSES | |
1664 | The number of distinct register classes, defined as follows: | |
1665 | ||
1666 | @example | |
1667 | #define N_REG_CLASSES (int) LIM_REG_CLASSES | |
1668 | @end example | |
1669 | ||
1670 | @findex REG_CLASS_NAMES | |
1671 | @item REG_CLASS_NAMES | |
1672 | An initializer containing the names of the register classes as C string | |
1673 | constants. These names are used in writing some of the debugging dumps. | |
1674 | ||
1675 | @findex REG_CLASS_CONTENTS | |
1676 | @item REG_CLASS_CONTENTS | |
1677 | An initializer containing the contents of the register classes, as integers | |
1678 | which are bit masks. The @var{n}th integer specifies the contents of class | |
1679 | @var{n}. The way the integer @var{mask} is interpreted is that | |
1680 | register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1. | |
1681 | ||
1682 | When the machine has more than 32 registers, an integer does not suffice. | |
1683 | Then the integers are replaced by sub-initializers, braced groupings containing | |
1684 | several integers. Each sub-initializer must be suitable as an initializer | |
1685 | for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}. | |
1686 | ||
1687 | @findex REGNO_REG_CLASS | |
1688 | @item REGNO_REG_CLASS (@var{regno}) | |
1689 | A C expression whose value is a register class containing hard register | |
1690 | @var{regno}. In general there is more than one such class; choose a class | |
1691 | which is @dfn{minimal}, meaning that no smaller class also contains the | |
1692 | register. | |
1693 | ||
1694 | @findex BASE_REG_CLASS | |
1695 | @item BASE_REG_CLASS | |
1696 | A macro whose definition is the name of the class to which a valid | |
1697 | base register must belong. A base register is one used in an address | |
1698 | which is the register value plus a displacement. | |
1699 | ||
1700 | @findex INDEX_REG_CLASS | |
1701 | @item INDEX_REG_CLASS | |
1702 | A macro whose definition is the name of the class to which a valid | |
1703 | index register must belong. An index register is one used in an | |
1704 | address where its value is either multiplied by a scale factor or | |
1705 | added to another register (as well as added to a displacement). | |
1706 | ||
1707 | @findex REG_CLASS_FROM_LETTER | |
1708 | @item REG_CLASS_FROM_LETTER (@var{char}) | |
1709 | A C expression which defines the machine-dependent operand constraint | |
1710 | letters for register classes. If @var{char} is such a letter, the | |
1711 | value should be the register class corresponding to it. Otherwise, | |
1712 | the value should be @code{NO_REGS}. The register letter @samp{r}, | |
1713 | corresponding to class @code{GENERAL_REGS}, will not be passed | |
1714 | to this macro; you do not need to handle it. | |
1715 | ||
1716 | @findex REGNO_OK_FOR_BASE_P | |
1717 | @item REGNO_OK_FOR_BASE_P (@var{num}) | |
1718 | A C expression which is nonzero if register number @var{num} is | |
1719 | suitable for use as a base register in operand addresses. It may be | |
1720 | either a suitable hard register or a pseudo register that has been | |
1721 | allocated such a hard register. | |
1722 | ||
1723 | @findex REGNO_OK_FOR_INDEX_P | |
1724 | @item REGNO_OK_FOR_INDEX_P (@var{num}) | |
1725 | A C expression which is nonzero if register number @var{num} is | |
1726 | suitable for use as an index register in operand addresses. It may be | |
1727 | either a suitable hard register or a pseudo register that has been | |
1728 | allocated such a hard register. | |
1729 | ||
1730 | The difference between an index register and a base register is that | |
1731 | the index register may be scaled. If an address involves the sum of | |
1732 | two registers, neither one of them scaled, then either one may be | |
1733 | labeled the ``base'' and the other the ``index''; but whichever | |
1734 | labeling is used must fit the machine's constraints of which registers | |
1735 | may serve in each capacity. The compiler will try both labelings, | |
1736 | looking for one that is valid, and will reload one or both registers | |
1737 | only if neither labeling works. | |
1738 | ||
1739 | @findex PREFERRED_RELOAD_CLASS | |
1740 | @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class}) | |
1741 | A C expression that places additional restrictions on the register class | |
1742 | to use when it is necessary to copy value @var{x} into a register in class | |
1743 | @var{class}. The value is a register class; perhaps @var{class}, or perhaps | |
1744 | another, smaller class. On many machines, the following definition is | |
1745 | safe: | |
1746 | ||
1747 | @example | |
1748 | #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS | |
1749 | @end example | |
1750 | ||
1751 | Sometimes returning a more restrictive class makes better code. For | |
1752 | example, on the 68000, when @var{x} is an integer constant that is in range | |
1753 | for a @samp{moveq} instruction, the value of this macro is always | |
1754 | @code{DATA_REGS} as long as @var{class} includes the data registers. | |
1755 | Requiring a data register guarantees that a @samp{moveq} will be used. | |
1756 | ||
1757 | If @var{x} is a @code{const_double}, by returning @code{NO_REGS} | |
1758 | you can force @var{x} into a memory constant. This is useful on | |
1759 | certain machines where immediate floating values cannot be loaded into | |
1760 | certain kinds of registers. | |
1761 | ||
1762 | @findex PREFERRED_OUTPUT_RELOAD_CLASS | |
1763 | @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class}) | |
1764 | Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of | |
1765 | input reloads. If you don't define this macro, the default is to use | |
1766 | @var{class}, unchanged. | |
1767 | ||
1768 | @findex LIMIT_RELOAD_CLASS | |
1769 | @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class}) | |
1770 | A C expression that places additional restrictions on the register class | |
1771 | to use when it is necessary to be able to hold a value of mode | |
1772 | @var{mode} in a reload register for which class @var{class} would | |
1773 | ordinarily be used. | |
1774 | ||
1775 | Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when | |
1776 | there are certain modes that simply can't go in certain reload classes. | |
1777 | ||
1778 | The value is a register class; perhaps @var{class}, or perhaps another, | |
1779 | smaller class. | |
1780 | ||
1781 | Don't define this macro unless the target machine has limitations which | |
1782 | require the macro to do something nontrivial. | |
1783 | ||
1784 | @findex SECONDARY_RELOAD_CLASS | |
1785 | @findex SECONDARY_INPUT_RELOAD_CLASS | |
1786 | @findex SECONDARY_OUTPUT_RELOAD_CLASS | |
1787 | @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) | |
1788 | @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) | |
1789 | @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) | |
1790 | Many machines have some registers that cannot be copied directly to or | |
1791 | from memory or even from other types of registers. An example is the | |
1792 | @samp{MQ} register, which on most machines, can only be copied to or | |
1793 | from general registers, but not memory. Some machines allow copying all | |
1794 | registers to and from memory, but require a scratch register for stores | |
1795 | to some memory locations (e.g., those with symbolic address on the RT, | |
1796 | and those with certain symbolic address on the Sparc when compiling | |
1797 | PIC). In some cases, both an intermediate and a scratch register are | |
1798 | required. | |
1799 | ||
1800 | You should define these macros to indicate to the reload phase that it may | |
1801 | need to allocate at least one register for a reload in addition to the | |
1802 | register to contain the data. Specifically, if copying @var{x} to a | |
1803 | register @var{class} in @var{mode} requires an intermediate register, | |
1804 | you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the | |
1805 | largest register class all of whose registers can be used as | |
1806 | intermediate registers or scratch registers. | |
1807 | ||
1808 | If copying a register @var{class} in @var{mode} to @var{x} requires an | |
1809 | intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS} | |
1810 | should be defined to return the largest register class required. If the | |
1811 | requirements for input and output reloads are the same, the macro | |
1812 | @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both | |
1813 | macros identically. | |
1814 | ||
1815 | The values returned by these macros are often @code{GENERAL_REGS}. | |
1816 | Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x} | |
1817 | can be directly copied to or from a register of @var{class} in | |
1818 | @var{mode} without requiring a scratch register. Do not define this | |
1819 | macro if it would always return @code{NO_REGS}. | |
1820 | ||
1821 | If a scratch register is required (either with or without an | |
1822 | intermediate register), you should define patterns for | |
1823 | @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required | |
1824 | (@pxref{Standard Names}. These patterns, which will normally be | |
1825 | implemented with a @code{define_expand}, should be similar to the | |
1826 | @samp{mov@var{m}} patterns, except that operand 2 is the scratch | |
1827 | register. | |
1828 | ||
1829 | Define constraints for the reload register and scratch register that | |
1830 | contain a single register class. If the original reload register (whose | |
1831 | class is @var{class}) can meet the constraint given in the pattern, the | |
1832 | value returned by these macros is used for the class of the scratch | |
1833 | register. Otherwise, two additional reload registers are required. | |
1834 | Their classes are obtained from the constraints in the insn pattern. | |
1835 | ||
1836 | @var{x} might be a pseudo-register or a @code{subreg} of a | |
1837 | pseudo-register, which could either be in a hard register or in memory. | |
1838 | Use @code{true_regnum} to find out; it will return -1 if the pseudo is | |
1839 | in memory and the hard register number if it is in a register. | |
1840 | ||
1841 | These macros should not be used in the case where a particular class of | |
1842 | registers can only be copied to memory and not to another class of | |
1843 | registers. In that case, secondary reload registers are not needed and | |
1844 | would not be helpful. Instead, a stack location must be used to perform | |
1845 | the copy and the @code{mov@var{m}} pattern should use memory as a | |
1846 | intermediate storage. This case often occurs between floating-point and | |
1847 | general registers. | |
1848 | ||
1849 | @findex SECONDARY_MEMORY_NEEDED | |
1850 | @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m}) | |
1851 | Certain machines have the property that some registers cannot be copied | |
1852 | to some other registers without using memory. Define this macro on | |
1853 | those machines to be a C expression that is non-zero if objects of mode | |
1854 | @var{m} in registers of @var{class1} can only be copied to registers of | |
1855 | class @var{class2} by storing a register of @var{class1} into memory | |
1856 | and loading that memory location into a register of @var{class2}. | |
1857 | ||
1858 | Do not define this macro if its value would always be zero. | |
1859 | ||
1860 | @findex SECONDARY_MEMORY_NEEDED_RTX | |
1861 | @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode}) | |
1862 | Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler | |
1863 | allocates a stack slot for a memory location needed for register copies. | |
1864 | If this macro is defined, the compiler instead uses the memory location | |
1865 | defined by this macro. | |
1866 | ||
1867 | Do not define this macro if you do not define | |
1868 | @code{SECONDARY_MEMORY_NEEDED}. | |
1869 | ||
1870 | @findex SECONDARY_MEMORY_NEEDED_MODE | |
1871 | @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode}) | |
1872 | When the compiler needs a secondary memory location to copy between two | |
1873 | registers of mode @var{mode}, it normally allocates sufficient memory to | |
1874 | hold a quantity of @code{BITS_PER_WORD} bits and performs the store and | |
1875 | load operations in a mode that many bits wide and whose class is the | |
1876 | same as that of @var{mode}. | |
1877 | ||
1878 | This is right thing to do on most machines because it ensures that all | |
1879 | bits of the register are copied and prevents accesses to the registers | |
1880 | in a narrower mode, which some machines prohibit for floating-point | |
1881 | registers. | |
1882 | ||
1883 | However, this default behavior is not correct on some machines, such as | |
1884 | the DEC Alpha, that store short integers in floating-point registers | |
1885 | differently than in integer registers. On those machines, the default | |
1886 | widening will not work correctly and you must define this macro to | |
1887 | suppress that widening in some cases. See the file @file{alpha.h} for | |
1888 | details. | |
1889 | ||
1890 | Do not define this macro if you do not define | |
1891 | @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that | |
1892 | is @code{BITS_PER_WORD} bits wide is correct for your machine. | |
1893 | ||
1894 | @findex SMALL_REGISTER_CLASSES | |
1895 | @item SMALL_REGISTER_CLASSES | |
1896 | Normally the compiler avoids choosing registers that have been | |
1897 | explicitly mentioned in the rtl as spill registers (these registers are | |
1898 | normally those used to pass parameters and return values). However, | |
1899 | some machines have so few registers of certain classes that there | |
1900 | would not be enough registers to use as spill registers if this were | |
1901 | done. | |
1902 | ||
1903 | Define @code{SMALL_REGISTER_CLASSES} on these machines. When it is | |
1904 | defined, the compiler allows registers explicitly used in the rtl to be | |
1905 | used as spill registers but avoids extending the lifetime of these | |
1906 | registers. | |
1907 | ||
1908 | It is always safe to define this macro, but if you unnecessarily define | |
1909 | it, you will reduce the amount of optimizations that can be performed in | |
1910 | some cases. If you do not define this macro when it is required, the | |
1911 | compiler will run out of spill registers and print a fatal error | |
1912 | message. For most machines, you should not define this macro. | |
1913 | ||
1914 | @findex CLASS_LIKELY_SPILLED_P | |
1915 | @item CLASS_LIKELY_SPILLED_P (@var{class}) | |
1916 | A C expression whose value is nonzero if pseudos that have been assigned | |
1917 | to registers of class @var{class} would likely be spilled because | |
1918 | registers of @var{class} are needed for spill registers. | |
1919 | ||
1920 | The default value of this macro returns 1 if @var{class} has exactly one | |
1921 | register and zero otherwise. On most machines, this default should be | |
1922 | used. Only define this macro to some other expression if pseudo | |
1923 | allocated by @file{local-alloc.c} end up in memory because their hard | |
1924 | registers were needed for spill registers. If this macro returns nonzero | |
1925 | for those classes, those pseudos will only be allocated by | |
1926 | @file{global.c}, which knows how to reallocate the pseudo to another | |
1927 | register. If there would not be another register available for | |
1928 | reallocation, you should not change the definition of this macro since | |
1929 | the only effect of such a definition would be to slow down register | |
1930 | allocation. | |
1931 | ||
1932 | @findex CLASS_MAX_NREGS | |
1933 | @item CLASS_MAX_NREGS (@var{class}, @var{mode}) | |
1934 | A C expression for the maximum number of consecutive registers | |
1935 | of class @var{class} needed to hold a value of mode @var{mode}. | |
1936 | ||
1937 | This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact, | |
1938 | the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})} | |
1939 | should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno}, | |
1940 | @var{mode})} for all @var{regno} values in the class @var{class}. | |
1941 | ||
1942 | This macro helps control the handling of multiple-word values | |
1943 | in the reload pass. | |
1944 | ||
1945 | @item CLASS_CANNOT_CHANGE_SIZE | |
1946 | If defined, a C expression for a class that contains registers which the | |
1947 | compiler must always access in a mode that is the same size as the mode | |
1948 | in which it loaded the register. | |
1949 | ||
1950 | For the example, loading 32-bit integer or floating-point objects into | |
1951 | floating-point registers on the Alpha extends them to 64-bits. | |
1952 | Therefore loading a 64-bit object and then storing it as a 32-bit object | |
1953 | does not store the low-order 32-bits, as would be the case for a normal | |
1954 | register. Therefore, @file{alpha.h} defines this macro as | |
1955 | @code{FLOAT_REGS}. | |
1956 | @end table | |
1957 | ||
1958 | Three other special macros describe which operands fit which constraint | |
1959 | letters. | |
1960 | ||
1961 | @table @code | |
1962 | @findex CONST_OK_FOR_LETTER_P | |
1963 | @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c}) | |
1964 | A C expression that defines the machine-dependent operand constraint letters | |
1965 | that specify particular ranges of integer values. If @var{c} is one | |
1966 | of those letters, the expression should check that @var{value}, an integer, | |
1967 | is in the appropriate range and return 1 if so, 0 otherwise. If @var{c} is | |
1968 | not one of those letters, the value should be 0 regardless of @var{value}. | |
1969 | ||
1970 | @findex CONST_DOUBLE_OK_FOR_LETTER_P | |
1971 | @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c}) | |
1972 | A C expression that defines the machine-dependent operand constraint | |
1973 | letters that specify particular ranges of @code{const_double} values. | |
1974 | ||
1975 | If @var{c} is one of those letters, the expression should check that | |
1976 | @var{value}, an RTX of code @code{const_double}, is in the appropriate | |
1977 | range and return 1 if so, 0 otherwise. If @var{c} is not one of those | |
1978 | letters, the value should be 0 regardless of @var{value}. | |
1979 | ||
1980 | @code{const_double} is used for all floating-point constants and for | |
1981 | @code{DImode} fixed-point constants. A given letter can accept either | |
1982 | or both kinds of values. It can use @code{GET_MODE} to distinguish | |
1983 | between these kinds. | |
1984 | ||
1985 | @findex EXTRA_CONSTRAINT | |
1986 | @item EXTRA_CONSTRAINT (@var{value}, @var{c}) | |
1987 | A C expression that defines the optional machine-dependent constraint | |
1988 | letters that can be used to segregate specific types of operands, | |
1989 | usually memory references, for the target machine. Normally this macro | |
1990 | will not be defined. If it is required for a particular target machine, | |
1991 | it should return 1 if @var{value} corresponds to the operand type | |
1992 | represented by the constraint letter @var{c}. If @var{c} is not defined | |
1993 | as an extra constraint, the value returned should be 0 regardless of | |
1994 | @var{value}. | |
1995 | ||
1996 | For example, on the ROMP, load instructions cannot have their output in r0 if | |
1997 | the memory reference contains a symbolic address. Constraint letter | |
1998 | @samp{Q} is defined as representing a memory address that does | |
1999 | @emph{not} contain a symbolic address. An alternative is specified with | |
2000 | a @samp{Q} constraint on the input and @samp{r} on the output. The next | |
2001 | alternative specifies @samp{m} on the input and a register class that | |
2002 | does not include r0 on the output. | |
2003 | @end table | |
2004 | ||
2005 | @node Stack and Calling | |
2006 | @section Stack Layout and Calling Conventions | |
2007 | @cindex calling conventions | |
2008 | ||
2009 | @c prevent bad page break with this line | |
2010 | This describes the stack layout and calling conventions. | |
2011 | ||
2012 | @menu | |
2013 | * Frame Layout:: | |
2014 | * Frame Registers:: | |
2015 | * Elimination:: | |
2016 | * Stack Arguments:: | |
2017 | * Register Arguments:: | |
2018 | * Scalar Return:: | |
2019 | * Aggregate Return:: | |
2020 | * Caller Saves:: | |
2021 | * Function Entry:: | |
2022 | * Profiling:: | |
2023 | @end menu | |
2024 | ||
2025 | @node Frame Layout | |
2026 | @subsection Basic Stack Layout | |
2027 | @cindex stack frame layout | |
2028 | @cindex frame layout | |
2029 | ||
2030 | @c prevent bad page break with this line | |
2031 | Here is the basic stack layout. | |
2032 | ||
2033 | @table @code | |
2034 | @findex STACK_GROWS_DOWNWARD | |
2035 | @item STACK_GROWS_DOWNWARD | |
2036 | Define this macro if pushing a word onto the stack moves the stack | |
2037 | pointer to a smaller address. | |
2038 | ||
2039 | When we say, ``define this macro if @dots{},'' it means that the | |
2040 | compiler checks this macro only with @code{#ifdef} so the precise | |
2041 | definition used does not matter. | |
2042 | ||
2043 | @findex FRAME_GROWS_DOWNWARD | |
2044 | @item FRAME_GROWS_DOWNWARD | |
2045 | Define this macro if the addresses of local variable slots are at negative | |
2046 | offsets from the frame pointer. | |
2047 | ||
2048 | @findex ARGS_GROW_DOWNWARD | |
2049 | @item ARGS_GROW_DOWNWARD | |
2050 | Define this macro if successive arguments to a function occupy decreasing | |
2051 | addresses on the stack. | |
2052 | ||
2053 | @findex STARTING_FRAME_OFFSET | |
2054 | @item STARTING_FRAME_OFFSET | |
2055 | Offset from the frame pointer to the first local variable slot to be allocated. | |
2056 | ||
2057 | If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by | |
2058 | subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}. | |
2059 | Otherwise, it is found by adding the length of the first slot to the | |
2060 | value @code{STARTING_FRAME_OFFSET}. | |
2061 | @c i'm not sure if the above is still correct.. had to change it to get | |
2062 | @c rid of an overfull. --mew 2feb93 | |
2063 | ||
2064 | @findex STACK_POINTER_OFFSET | |
2065 | @item STACK_POINTER_OFFSET | |
2066 | Offset from the stack pointer register to the first location at which | |
2067 | outgoing arguments are placed. If not specified, the default value of | |
2068 | zero is used. This is the proper value for most machines. | |
2069 | ||
2070 | If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above | |
2071 | the first location at which outgoing arguments are placed. | |
2072 | ||
2073 | @findex FIRST_PARM_OFFSET | |
2074 | @item FIRST_PARM_OFFSET (@var{fundecl}) | |
2075 | Offset from the argument pointer register to the first argument's | |
2076 | address. On some machines it may depend on the data type of the | |
2077 | function. | |
2078 | ||
2079 | If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above | |
2080 | the first argument's address. | |
2081 | ||
2082 | @findex STACK_DYNAMIC_OFFSET | |
2083 | @item STACK_DYNAMIC_OFFSET (@var{fundecl}) | |
2084 | Offset from the stack pointer register to an item dynamically allocated | |
2085 | on the stack, e.g., by @code{alloca}. | |
2086 | ||
2087 | The default value for this macro is @code{STACK_POINTER_OFFSET} plus the | |
2088 | length of the outgoing arguments. The default is correct for most | |
2089 | machines. See @file{function.c} for details. | |
2090 | ||
2091 | @findex DYNAMIC_CHAIN_ADDRESS | |
2092 | @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr}) | |
2093 | A C expression whose value is RTL representing the address in a stack | |
2094 | frame where the pointer to the caller's frame is stored. Assume that | |
2095 | @var{frameaddr} is an RTL expression for the address of the stack frame | |
2096 | itself. | |
2097 | ||
2098 | If you don't define this macro, the default is to return the value | |
2099 | of @var{frameaddr}---that is, the stack frame address is also the | |
2100 | address of the stack word that points to the previous frame. | |
2101 | ||
2102 | @findex SETUP_FRAME_ADDRESSES | |
2103 | @item SETUP_FRAME_ADDRESSES () | |
2104 | If defined, a C expression that produces the machine-specific code to | |
2105 | setup the stack so that arbitrary frames can be accessed. For example, | |
2106 | on the Sparc, we must flush all of the register windows to the stack | |
2107 | before we can access arbitrary stack frames. | |
2108 | This macro will seldom need to be defined. | |
2109 | ||
2110 | @findex RETURN_ADDR_RTX | |
2111 | @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr}) | |
2112 | A C expression whose value is RTL representing the value of the return | |
2113 | address for the frame @var{count} steps up from the current frame. | |
2114 | @var{frameaddr} is the frame pointer of the @var{count} frame, or | |
2115 | the frame pointer of the @var{count} @minus{} 1 frame if | |
2116 | @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined. | |
2117 | ||
2118 | @findex RETURN_ADDR_IN_PREVIOUS_FRAME | |
2119 | @item RETURN_ADDR_IN_PREVIOUS_FRAME | |
2120 | Define this if the return address of a particular stack frame is accessed | |
2121 | from the frame pointer of the previous stack frame. | |
2122 | @end table | |
2123 | ||
2124 | @need 2000 | |
2125 | @node Frame Registers | |
2126 | @subsection Registers That Address the Stack Frame | |
2127 | ||
2128 | @c prevent bad page break with this line | |
2129 | This discusses registers that address the stack frame. | |
2130 | ||
2131 | @table @code | |
2132 | @findex STACK_POINTER_REGNUM | |
2133 | @item STACK_POINTER_REGNUM | |
2134 | The register number of the stack pointer register, which must also be a | |
2135 | fixed register according to @code{FIXED_REGISTERS}. On most machines, | |
2136 | the hardware determines which register this is. | |
2137 | ||
2138 | @findex FRAME_POINTER_REGNUM | |
2139 | @item FRAME_POINTER_REGNUM | |
2140 | The register number of the frame pointer register, which is used to | |
2141 | access automatic variables in the stack frame. On some machines, the | |
2142 | hardware determines which register this is. On other machines, you can | |
2143 | choose any register you wish for this purpose. | |
2144 | ||
2145 | @findex HARD_FRAME_POINTER_REGNUM | |
2146 | @item HARD_FRAME_POINTER_REGNUM | |
2147 | On some machines the offset between the frame pointer and starting | |
2148 | offset of the automatic variables is not known until after register | |
2149 | allocation has been done (for example, because the saved registers are | |
2150 | between these two locations). On those machines, define | |
2151 | @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to | |
2152 | be used internally until the offset is known, and define | |
2153 | @code{HARD_FRAME_POINTER_REGNUM} to be actual the hard register number | |
2154 | used for the frame pointer. | |
2155 | ||
2156 | You should define this macro only in the very rare circumstances when it | |
2157 | is not possible to calculate the offset between the frame pointer and | |
2158 | the automatic variables until after register allocation has been | |
2159 | completed. When this macro is defined, you must also indicate in your | |
2160 | definition of @code{ELIMINABLE_REGS} how to eliminate | |
2161 | @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM} | |
2162 | or @code{STACK_POINTER_REGNUM}. | |
2163 | ||
2164 | Do not define this macro if it would be the same as | |
2165 | @code{FRAME_POINTER_REGNUM}. | |
2166 | ||
2167 | @findex ARG_POINTER_REGNUM | |
2168 | @item ARG_POINTER_REGNUM | |
2169 | The register number of the arg pointer register, which is used to access | |
2170 | the function's argument list. On some machines, this is the same as the | |
2171 | frame pointer register. On some machines, the hardware determines which | |
2172 | register this is. On other machines, you can choose any register you | |
2173 | wish for this purpose. If this is not the same register as the frame | |
2174 | pointer register, then you must mark it as a fixed register according to | |
2175 | @code{FIXED_REGISTERS}, or arrange to be able to eliminate it | |
2176 | (@pxref{Elimination}). | |
2177 | ||
2178 | @findex RETURN_ADDRESS_POINTER_REGNUM | |
2179 | @item RETURN_ADDRESS_POINTER_REGNUM | |
2180 | The register number of the return address pointer register, which is used to | |
2181 | access the current function's return address from the stack. On some | |
2182 | machines, the return address is not at a fixed offset from the frame | |
2183 | pointer or stack pointer or argument pointer. This register can be defined | |
2184 | to point to the return address on the stack, and then be converted by | |
2185 | @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer. | |
2186 | ||
2187 | Do not define this macro unless there is no other way to get the return | |
2188 | address from the stack. | |
2189 | ||
2190 | @findex STATIC_CHAIN_REGNUM | |
2191 | @findex STATIC_CHAIN_INCOMING_REGNUM | |
2192 | @item STATIC_CHAIN_REGNUM | |
2193 | @itemx STATIC_CHAIN_INCOMING_REGNUM | |
2194 | Register numbers used for passing a function's static chain pointer. If | |
2195 | register windows are used, the register number as seen by the called | |
2196 | function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register | |
2197 | number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If | |
2198 | these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need | |
2199 | not be defined.@refill | |
2200 | ||
2201 | The static chain register need not be a fixed register. | |
2202 | ||
2203 | If the static chain is passed in memory, these macros should not be | |
2204 | defined; instead, the next two macros should be defined. | |
2205 | ||
2206 | @findex STATIC_CHAIN | |
2207 | @findex STATIC_CHAIN_INCOMING | |
2208 | @item STATIC_CHAIN | |
2209 | @itemx STATIC_CHAIN_INCOMING | |
2210 | If the static chain is passed in memory, these macros provide rtx giving | |
2211 | @code{mem} expressions that denote where they are stored. | |
2212 | @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations | |
2213 | as seen by the calling and called functions, respectively. Often the former | |
2214 | will be at an offset from the stack pointer and the latter at an offset from | |
2215 | the frame pointer.@refill | |
2216 | ||
2217 | @findex stack_pointer_rtx | |
2218 | @findex frame_pointer_rtx | |
2219 | @findex arg_pointer_rtx | |
2220 | The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and | |
2221 | @code{arg_pointer_rtx} will have been initialized prior to the use of these | |
2222 | macros and should be used to refer to those items. | |
2223 | ||
2224 | If the static chain is passed in a register, the two previous macros should | |
2225 | be defined instead. | |
2226 | @end table | |
2227 | ||
2228 | @node Elimination | |
2229 | @subsection Eliminating Frame Pointer and Arg Pointer | |
2230 | ||
2231 | @c prevent bad page break with this line | |
2232 | This is about eliminating the frame pointer and arg pointer. | |
2233 | ||
2234 | @table @code | |
2235 | @findex FRAME_POINTER_REQUIRED | |
2236 | @item FRAME_POINTER_REQUIRED | |
2237 | A C expression which is nonzero if a function must have and use a frame | |
2238 | pointer. This expression is evaluated in the reload pass. If its value is | |
2239 | nonzero the function will have a frame pointer. | |
2240 | ||
2241 | The expression can in principle examine the current function and decide | |
2242 | according to the facts, but on most machines the constant 0 or the | |
2243 | constant 1 suffices. Use 0 when the machine allows code to be generated | |
2244 | with no frame pointer, and doing so saves some time or space. Use 1 | |
2245 | when there is no possible advantage to avoiding a frame pointer. | |
2246 | ||
2247 | In certain cases, the compiler does not know how to produce valid code | |
2248 | without a frame pointer. The compiler recognizes those cases and | |
2249 | automatically gives the function a frame pointer regardless of what | |
2250 | @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about | |
2251 | them.@refill | |
2252 | ||
2253 | In a function that does not require a frame pointer, the frame pointer | |
2254 | register can be allocated for ordinary usage, unless you mark it as a | |
2255 | fixed register. See @code{FIXED_REGISTERS} for more information. | |
2256 | ||
2257 | @findex INITIAL_FRAME_POINTER_OFFSET | |
2258 | @findex get_frame_size | |
2259 | @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var}) | |
2260 | A C statement to store in the variable @var{depth-var} the difference | |
2261 | between the frame pointer and the stack pointer values immediately after | |
2262 | the function prologue. The value would be computed from information | |
2263 | such as the result of @code{get_frame_size ()} and the tables of | |
2264 | registers @code{regs_ever_live} and @code{call_used_regs}. | |
2265 | ||
2266 | If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and | |
2267 | need not be defined. Otherwise, it must be defined even if | |
2268 | @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that | |
2269 | case, you may set @var{depth-var} to anything. | |
2270 | ||
2271 | @findex ELIMINABLE_REGS | |
2272 | @item ELIMINABLE_REGS | |
2273 | If defined, this macro specifies a table of register pairs used to | |
2274 | eliminate unneeded registers that point into the stack frame. If it is not | |
2275 | defined, the only elimination attempted by the compiler is to replace | |
2276 | references to the frame pointer with references to the stack pointer. | |
2277 | ||
2278 | The definition of this macro is a list of structure initializations, each | |
2279 | of which specifies an original and replacement register. | |
2280 | ||
2281 | On some machines, the position of the argument pointer is not known until | |
2282 | the compilation is completed. In such a case, a separate hard register | |
2283 | must be used for the argument pointer. This register can be eliminated by | |
2284 | replacing it with either the frame pointer or the argument pointer, | |
2285 | depending on whether or not the frame pointer has been eliminated. | |
2286 | ||
2287 | In this case, you might specify: | |
2288 | @example | |
2289 | #define ELIMINABLE_REGS \ | |
2290 | @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \ | |
2291 | @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \ | |
2292 | @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@} | |
2293 | @end example | |
2294 | ||
2295 | Note that the elimination of the argument pointer with the stack pointer is | |
2296 | specified first since that is the preferred elimination. | |
2297 | ||
2298 | @findex CAN_ELIMINATE | |
2299 | @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg}) | |
2300 | A C expression that returns non-zero if the compiler is allowed to try | |
2301 | to replace register number @var{from-reg} with register number | |
2302 | @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS} | |
2303 | is defined, and will usually be the constant 1, since most of the cases | |
2304 | preventing register elimination are things that the compiler already | |
2305 | knows about. | |
2306 | ||
2307 | @findex INITIAL_ELIMINATION_OFFSET | |
2308 | @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var}) | |
2309 | This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It | |
2310 | specifies the initial difference between the specified pair of | |
2311 | registers. This macro must be defined if @code{ELIMINABLE_REGS} is | |
2312 | defined. | |
2313 | ||
2314 | @findex LONGJMP_RESTORE_FROM_STACK | |
2315 | @item LONGJMP_RESTORE_FROM_STACK | |
2316 | Define this macro if the @code{longjmp} function restores registers from | |
2317 | the stack frames, rather than from those saved specifically by | |
2318 | @code{setjmp}. Certain quantities must not be kept in registers across | |
2319 | a call to @code{setjmp} on such machines. | |
2320 | @end table | |
2321 | ||
2322 | @node Stack Arguments | |
2323 | @subsection Passing Function Arguments on the Stack | |
2324 | @cindex arguments on stack | |
2325 | @cindex stack arguments | |
2326 | ||
2327 | The macros in this section control how arguments are passed | |
2328 | on the stack. See the following section for other macros that | |
2329 | control passing certain arguments in registers. | |
2330 | ||
2331 | @table @code | |
2332 | @findex PROMOTE_PROTOTYPES | |
2333 | @item PROMOTE_PROTOTYPES | |
2334 | Define this macro if an argument declared in a prototype as an | |
2335 | integral type smaller than @code{int} should actually be passed as an | |
2336 | @code{int}. In addition to avoiding errors in certain cases of | |
2337 | mismatch, it also makes for better code on certain machines. | |
2338 | ||
2339 | @findex PUSH_ROUNDING | |
2340 | @item PUSH_ROUNDING (@var{npushed}) | |
2341 | A C expression that is the number of bytes actually pushed onto the | |
2342 | stack when an instruction attempts to push @var{npushed} bytes. | |
2343 | ||
2344 | If the target machine does not have a push instruction, do not define | |
2345 | this macro. That directs GNU CC to use an alternate strategy: to | |
2346 | allocate the entire argument block and then store the arguments into | |
2347 | it. | |
2348 | ||
2349 | On some machines, the definition | |
2350 | ||
2351 | @example | |
2352 | #define PUSH_ROUNDING(BYTES) (BYTES) | |
2353 | @end example | |
2354 | ||
2355 | @noindent | |
2356 | will suffice. But on other machines, instructions that appear | |
2357 | to push one byte actually push two bytes in an attempt to maintain | |
2358 | alignment. Then the definition should be | |
2359 | ||
2360 | @example | |
2361 | #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1) | |
2362 | @end example | |
2363 | ||
2364 | @findex ACCUMULATE_OUTGOING_ARGS | |
2365 | @findex current_function_outgoing_args_size | |
2366 | @item ACCUMULATE_OUTGOING_ARGS | |
2367 | If defined, the maximum amount of space required for outgoing arguments | |
2368 | will be computed and placed into the variable | |
2369 | @code{current_function_outgoing_args_size}. No space will be pushed | |
2370 | onto the stack for each call; instead, the function prologue should | |
2371 | increase the stack frame size by this amount. | |
2372 | ||
2373 | Defining both @code{PUSH_ROUNDING} and @code{ACCUMULATE_OUTGOING_ARGS} | |
2374 | is not proper. | |
2375 | ||
2376 | @findex REG_PARM_STACK_SPACE | |
2377 | @item REG_PARM_STACK_SPACE (@var{fndecl}) | |
2378 | Define this macro if functions should assume that stack space has been | |
2379 | allocated for arguments even when their values are passed in | |
2380 | registers. | |
2381 | ||
2382 | The value of this macro is the size, in bytes, of the area reserved for | |
2383 | arguments passed in registers for the function represented by @var{fndecl}. | |
2384 | ||
2385 | This space can be allocated by the caller, or be a part of the | |
2386 | machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says | |
2387 | which. | |
2388 | @c above is overfull. not sure what to do. --mew 5feb93 did | |
2389 | @c something, not sure if it looks good. --mew 10feb93 | |
2390 | ||
2391 | @findex MAYBE_REG_PARM_STACK_SPACE | |
2392 | @findex FINAL_REG_PARM_STACK_SPACE | |
2393 | @item MAYBE_REG_PARM_STACK_SPACE | |
2394 | @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size}) | |
2395 | Define these macros in addition to the one above if functions might | |
2396 | allocate stack space for arguments even when their values are passed | |
2397 | in registers. These should be used when the stack space allocated | |
2398 | for arguments in registers is not a simple constant independent of the | |
2399 | function declaration. | |
2400 | ||
2401 | The value of the first macro is the size, in bytes, of the area that | |
2402 | we should initially assume would be reserved for arguments passed in registers. | |
2403 | ||
2404 | The value of the second macro is the actual size, in bytes, of the area | |
2405 | that will be reserved for arguments passed in registers. This takes two | |
2406 | arguments: an integer representing the number of bytes of fixed sized | |
2407 | arguments on the stack, and a tree representing the number of bytes of | |
2408 | variable sized arguments on the stack. | |
2409 | ||
2410 | When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be | |
2411 | called for libcall functions, the current function, or for a function | |
2412 | being called when it is known that such stack space must be allocated. | |
2413 | In each case this value can be easily computed. | |
2414 | ||
2415 | When deciding whether a called function needs such stack space, and how | |
2416 | much space to reserve, GNU CC uses these two macros instead of | |
2417 | @code{REG_PARM_STACK_SPACE}. | |
2418 | ||
2419 | @findex OUTGOING_REG_PARM_STACK_SPACE | |
2420 | @item OUTGOING_REG_PARM_STACK_SPACE | |
2421 | Define this if it is the responsibility of the caller to allocate the area | |
2422 | reserved for arguments passed in registers. | |
2423 | ||
2424 | If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls | |
2425 | whether the space for these arguments counts in the value of | |
2426 | @code{current_function_outgoing_args_size}. | |
2427 | ||
2428 | @findex STACK_PARMS_IN_REG_PARM_AREA | |
2429 | @item STACK_PARMS_IN_REG_PARM_AREA | |
2430 | Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the | |
2431 | stack parameters don't skip the area specified by it. | |
2432 | @c i changed this, makes more sens and it should have taken care of the | |
2433 | @c overfull.. not as specific, tho. --mew 5feb93 | |
2434 | ||
2435 | Normally, when a parameter is not passed in registers, it is placed on the | |
2436 | stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro | |
2437 | suppresses this behavior and causes the parameter to be passed on the | |
2438 | stack in its natural location. | |
2439 | ||
2440 | @findex RETURN_POPS_ARGS | |
2441 | @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size}) | |
2442 | A C expression that should indicate the number of bytes of its own | |
2443 | arguments that a function pops on returning, or 0 if the | |
2444 | function pops no arguments and the caller must therefore pop them all | |
2445 | after the function returns. | |
2446 | ||
2447 | @var{fundecl} is a C variable whose value is a tree node that describes | |
2448 | the function in question. Normally it is a node of type | |
2449 | @code{FUNCTION_DECL} that describes the declaration of the function. | |
2450 | From this you can obtain the DECL_MACHINE_ATTRIBUTES of the function. | |
2451 | ||
2452 | @var{funtype} is a C variable whose value is a tree node that | |
2453 | describes the function in question. Normally it is a node of type | |
2454 | @code{FUNCTION_TYPE} that describes the data type of the function. | |
2455 | From this it is possible to obtain the data types of the value and | |
2456 | arguments (if known). | |
2457 | ||
2458 | When a call to a library function is being considered, @var{funtype} | |
2459 | will contain an identifier node for the library function. Thus, if | |
2460 | you need to distinguish among various library functions, you can do so | |
2461 | by their names. Note that ``library function'' in this context means | |
2462 | a function used to perform arithmetic, whose name is known specially | |
2463 | in the compiler and was not mentioned in the C code being compiled. | |
2464 | ||
2465 | @var{stack-size} is the number of bytes of arguments passed on the | |
2466 | stack. If a variable number of bytes is passed, it is zero, and | |
2467 | argument popping will always be the responsibility of the calling function. | |
2468 | ||
2469 | On the Vax, all functions always pop their arguments, so the definition | |
2470 | of this macro is @var{stack-size}. On the 68000, using the standard | |
2471 | calling convention, no functions pop their arguments, so the value of | |
2472 | the macro is always 0 in this case. But an alternative calling | |
2473 | convention is available in which functions that take a fixed number of | |
2474 | arguments pop them but other functions (such as @code{printf}) pop | |
2475 | nothing (the caller pops all). When this convention is in use, | |
2476 | @var{funtype} is examined to determine whether a function takes a fixed | |
2477 | number of arguments. | |
2478 | @end table | |
2479 | ||
2480 | @node Register Arguments | |
2481 | @subsection Passing Arguments in Registers | |
2482 | @cindex arguments in registers | |
2483 | @cindex registers arguments | |
2484 | ||
2485 | This section describes the macros which let you control how various | |
2486 | types of arguments are passed in registers or how they are arranged in | |
2487 | the stack. | |
2488 | ||
2489 | @table @code | |
2490 | @findex FUNCTION_ARG | |
2491 | @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
2492 | A C expression that controls whether a function argument is passed | |
2493 | in a register, and which register. | |
2494 | ||
2495 | The arguments are @var{cum}, which summarizes all the previous | |
2496 | arguments; @var{mode}, the machine mode of the argument; @var{type}, | |
2497 | the data type of the argument as a tree node or 0 if that is not known | |
2498 | (which happens for C support library functions); and @var{named}, | |
2499 | which is 1 for an ordinary argument and 0 for nameless arguments that | |
2500 | correspond to @samp{@dots{}} in the called function's prototype. | |
2501 | ||
2502 | The value of the expression is usually either a @code{reg} RTX for the | |
2503 | hard register in which to pass the argument, or zero to pass the | |
2504 | argument on the stack. | |
2505 | ||
2506 | For machines like the Vax and 68000, where normally all arguments are | |
2507 | pushed, zero suffices as a definition. | |
2508 | ||
2509 | The value of the expression can also be a @code{parallel} RTX. This is | |
2510 | used when an argument is passed in multiple locations. The mode of the | |
2511 | of the @code{parallel} should be the mode of the entire argument. The | |
2512 | @code{parallel} holds any number of @code{expr_list} pairs; each one | |
2513 | describes where part of the argument is passed. In each @code{expr_list}, | |
2514 | the first operand can be either a @code{reg} RTX for the hard register | |
2515 | in which to pass this part of the argument, or zero to pass the argument | |
2516 | on the stack. If this operand is a @code{reg}, then the mode indicates | |
2517 | how large this part of the argument is. The second operand of the | |
2518 | @code{expr_list} is a @code{const_int} which gives the offset in bytes | |
2519 | into the entire argument where this part starts. | |
2520 | ||
2521 | @cindex @file{stdarg.h} and register arguments | |
2522 | The usual way to make the ANSI library @file{stdarg.h} work on a machine | |
2523 | where some arguments are usually passed in registers, is to cause | |
2524 | nameless arguments to be passed on the stack instead. This is done | |
2525 | by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0. | |
2526 | ||
2527 | @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG} | |
2528 | @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG} | |
2529 | You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})} | |
2530 | in the definition of this macro to determine if this argument is of a | |
2531 | type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE} | |
2532 | is not defined and @code{FUNCTION_ARG} returns non-zero for such an | |
2533 | argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is | |
2534 | defined, the argument will be computed in the stack and then loaded into | |
2535 | a register. | |
2536 | ||
2537 | @findex FUNCTION_INCOMING_ARG | |
2538 | @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
2539 | Define this macro if the target machine has ``register windows'', so | |
2540 | that the register in which a function sees an arguments is not | |
2541 | necessarily the same as the one in which the caller passed the | |
2542 | argument. | |
2543 | ||
2544 | For such machines, @code{FUNCTION_ARG} computes the register in which | |
2545 | the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should | |
2546 | be defined in a similar fashion to tell the function being called | |
2547 | where the arguments will arrive. | |
2548 | ||
2549 | If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG} | |
2550 | serves both purposes.@refill | |
2551 | ||
2552 | @findex FUNCTION_ARG_PARTIAL_NREGS | |
2553 | @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
2554 | A C expression for the number of words, at the beginning of an | |
2555 | argument, must be put in registers. The value must be zero for | |
2556 | arguments that are passed entirely in registers or that are entirely | |
2557 | pushed on the stack. | |
2558 | ||
2559 | On some machines, certain arguments must be passed partially in | |
2560 | registers and partially in memory. On these machines, typically the | |
2561 | first @var{n} words of arguments are passed in registers, and the rest | |
2562 | on the stack. If a multi-word argument (a @code{double} or a | |
2563 | structure) crosses that boundary, its first few words must be passed | |
2564 | in registers and the rest must be pushed. This macro tells the | |
2565 | compiler when this occurs, and how many of the words should go in | |
2566 | registers. | |
2567 | ||
2568 | @code{FUNCTION_ARG} for these arguments should return the first | |
2569 | register to be used by the caller for this argument; likewise | |
2570 | @code{FUNCTION_INCOMING_ARG}, for the called function. | |
2571 | ||
2572 | @findex FUNCTION_ARG_PASS_BY_REFERENCE | |
2573 | @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
2574 | A C expression that indicates when an argument must be passed by reference. | |
2575 | If nonzero for an argument, a copy of that argument is made in memory and a | |
2576 | pointer to the argument is passed instead of the argument itself. | |
2577 | The pointer is passed in whatever way is appropriate for passing a pointer | |
2578 | to that type. | |
2579 | ||
2580 | On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable | |
2581 | definition of this macro might be | |
2582 | @smallexample | |
2583 | #define FUNCTION_ARG_PASS_BY_REFERENCE\ | |
2584 | (CUM, MODE, TYPE, NAMED) \ | |
2585 | MUST_PASS_IN_STACK (MODE, TYPE) | |
2586 | @end smallexample | |
2587 | @c this is *still* too long. --mew 5feb93 | |
2588 | ||
2589 | @findex FUNCTION_ARG_CALLEE_COPIES | |
2590 | @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
2591 | If defined, a C expression that indicates when it is the called function's | |
2592 | responsibility to make a copy of arguments passed by invisible reference. | |
2593 | Normally, the caller makes a copy and passes the address of the copy to the | |
2594 | routine being called. When FUNCTION_ARG_CALLEE_COPIES is defined and is | |
2595 | nonzero, the caller does not make a copy. Instead, it passes a pointer to the | |
2596 | ``live'' value. The called function must not modify this value. If it can be | |
2597 | determined that the value won't be modified, it need not make a copy; | |
2598 | otherwise a copy must be made. | |
2599 | ||
2600 | @findex CUMULATIVE_ARGS | |
2601 | @item CUMULATIVE_ARGS | |
2602 | A C type for declaring a variable that is used as the first argument of | |
2603 | @code{FUNCTION_ARG} and other related values. For some target machines, | |
2604 | the type @code{int} suffices and can hold the number of bytes of | |
2605 | argument so far. | |
2606 | ||
2607 | There is no need to record in @code{CUMULATIVE_ARGS} anything about the | |
2608 | arguments that have been passed on the stack. The compiler has other | |
2609 | variables to keep track of that. For target machines on which all | |
2610 | arguments are passed on the stack, there is no need to store anything in | |
2611 | @code{CUMULATIVE_ARGS}; however, the data structure must exist and | |
2612 | should not be empty, so use @code{int}. | |
2613 | ||
2614 | @findex INIT_CUMULATIVE_ARGS | |
2615 | @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect}) | |
2616 | A C statement (sans semicolon) for initializing the variable @var{cum} | |
2617 | for the state at the beginning of the argument list. The variable has | |
2618 | type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node | |
2619 | for the data type of the function which will receive the args, or 0 | |
2620 | if the args are to a compiler support library function. The value of | |
2621 | @var{indirect} is nonzero when processing an indirect call, for example | |
2622 | a call through a function pointer. The value of @var{indirect} is zero | |
2623 | for a call to an explicitly named function, a library function call, or when | |
2624 | @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function | |
2625 | being compiled. | |
2626 | ||
2627 | When processing a call to a compiler support library function, | |
2628 | @var{libname} identifies which one. It is a @code{symbol_ref} rtx which | |
2629 | contains the name of the function, as a string. @var{libname} is 0 when | |
2630 | an ordinary C function call is being processed. Thus, each time this | |
2631 | macro is called, either @var{libname} or @var{fntype} is nonzero, but | |
2632 | never both of them at once. | |
2633 | ||
2634 | @findex INIT_CUMULATIVE_INCOMING_ARGS | |
2635 | @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname}) | |
2636 | Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of | |
2637 | finding the arguments for the function being compiled. If this macro is | |
2638 | undefined, @code{INIT_CUMULATIVE_ARGS} is used instead. | |
2639 | ||
2640 | The value passed for @var{libname} is always 0, since library routines | |
2641 | with special calling conventions are never compiled with GNU CC. The | |
2642 | argument @var{libname} exists for symmetry with | |
2643 | @code{INIT_CUMULATIVE_ARGS}. | |
2644 | @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe. | |
2645 | @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93 | |
2646 | ||
2647 | @findex FUNCTION_ARG_ADVANCE | |
2648 | @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
2649 | A C statement (sans semicolon) to update the summarizer variable | |
2650 | @var{cum} to advance past an argument in the argument list. The | |
2651 | values @var{mode}, @var{type} and @var{named} describe that argument. | |
2652 | Once this is done, the variable @var{cum} is suitable for analyzing | |
2653 | the @emph{following} argument with @code{FUNCTION_ARG}, etc.@refill | |
2654 | ||
2655 | This macro need not do anything if the argument in question was passed | |
2656 | on the stack. The compiler knows how to track the amount of stack space | |
2657 | used for arguments without any special help. | |
2658 | ||
2659 | @findex FUNCTION_ARG_PADDING | |
2660 | @item FUNCTION_ARG_PADDING (@var{mode}, @var{type}) | |
2661 | If defined, a C expression which determines whether, and in which direction, | |
2662 | to pad out an argument with extra space. The value should be of type | |
2663 | @code{enum direction}: either @code{upward} to pad above the argument, | |
2664 | @code{downward} to pad below, or @code{none} to inhibit padding. | |
2665 | ||
2666 | The @emph{amount} of padding is always just enough to reach the next | |
2667 | multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control | |
2668 | it. | |
2669 | ||
2670 | This macro has a default definition which is right for most systems. | |
2671 | For little-endian machines, the default is to pad upward. For | |
2672 | big-endian machines, the default is to pad downward for an argument of | |
2673 | constant size shorter than an @code{int}, and upward otherwise. | |
2674 | ||
2675 | @findex FUNCTION_ARG_BOUNDARY | |
2676 | @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type}) | |
2677 | If defined, a C expression that gives the alignment boundary, in bits, | |
2678 | of an argument with the specified mode and type. If it is not defined, | |
2679 | @code{PARM_BOUNDARY} is used for all arguments. | |
2680 | ||
2681 | @findex FUNCTION_ARG_REGNO_P | |
2682 | @item FUNCTION_ARG_REGNO_P (@var{regno}) | |
2683 | A C expression that is nonzero if @var{regno} is the number of a hard | |
2684 | register in which function arguments are sometimes passed. This does | |
2685 | @emph{not} include implicit arguments such as the static chain and | |
2686 | the structure-value address. On many machines, no registers can be | |
2687 | used for this purpose since all function arguments are pushed on the | |
2688 | stack. | |
2689 | @end table | |
2690 | ||
2691 | @node Scalar Return | |
2692 | @subsection How Scalar Function Values Are Returned | |
2693 | @cindex return values in registers | |
2694 | @cindex values, returned by functions | |
2695 | @cindex scalars, returned as values | |
2696 | ||
2697 | This section discusses the macros that control returning scalars as | |
2698 | values---values that can fit in registers. | |
2699 | ||
2700 | @table @code | |
2701 | @findex TRADITIONAL_RETURN_FLOAT | |
2702 | @item TRADITIONAL_RETURN_FLOAT | |
2703 | Define this macro if @samp{-traditional} should not cause functions | |
2704 | declared to return @code{float} to convert the value to @code{double}. | |
2705 | ||
2706 | @findex FUNCTION_VALUE | |
2707 | @item FUNCTION_VALUE (@var{valtype}, @var{func}) | |
2708 | A C expression to create an RTX representing the place where a | |
2709 | function returns a value of data type @var{valtype}. @var{valtype} is | |
2710 | a tree node representing a data type. Write @code{TYPE_MODE | |
2711 | (@var{valtype})} to get the machine mode used to represent that type. | |
2712 | On many machines, only the mode is relevant. (Actually, on most | |
2713 | machines, scalar values are returned in the same place regardless of | |
2714 | mode).@refill | |
2715 | ||
2716 | The value of the expression is usually a @code{reg} RTX for the hard | |
2717 | register where the return value is stored. The value can also be a | |
2718 | @code{parallel} RTX, if the return value is in multiple places. See | |
2719 | @code{FUNCTION_ARG} for an explanation of the @code{parallel} form. | |
2720 | ||
2721 | If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same | |
2722 | promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a | |
2723 | scalar type. | |
2724 | ||
2725 | If the precise function being called is known, @var{func} is a tree | |
2726 | node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null | |
2727 | pointer. This makes it possible to use a different value-returning | |
2728 | convention for specific functions when all their calls are | |
2729 | known.@refill | |
2730 | ||
2731 | @code{FUNCTION_VALUE} is not used for return vales with aggregate data | |
2732 | types, because these are returned in another way. See | |
2733 | @code{STRUCT_VALUE_REGNUM} and related macros, below. | |
2734 | ||
2735 | @findex FUNCTION_OUTGOING_VALUE | |
2736 | @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func}) | |
2737 | Define this macro if the target machine has ``register windows'' | |
2738 | so that the register in which a function returns its value is not | |
2739 | the same as the one in which the caller sees the value. | |
2740 | ||
2741 | For such machines, @code{FUNCTION_VALUE} computes the register in which | |
2742 | the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be | |
2743 | defined in a similar fashion to tell the function where to put the | |
2744 | value.@refill | |
2745 | ||
2746 | If @code{FUNCTION_OUTGOING_VALUE} is not defined, | |
2747 | @code{FUNCTION_VALUE} serves both purposes.@refill | |
2748 | ||
2749 | @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with | |
2750 | aggregate data types, because these are returned in another way. See | |
2751 | @code{STRUCT_VALUE_REGNUM} and related macros, below. | |
2752 | ||
2753 | @findex LIBCALL_VALUE | |
2754 | @item LIBCALL_VALUE (@var{mode}) | |
2755 | A C expression to create an RTX representing the place where a library | |
2756 | function returns a value of mode @var{mode}. If the precise function | |
2757 | being called is known, @var{func} is a tree node | |
2758 | (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null | |
2759 | pointer. This makes it possible to use a different value-returning | |
2760 | convention for specific functions when all their calls are | |
2761 | known.@refill | |
2762 | ||
2763 | Note that ``library function'' in this context means a compiler | |
2764 | support routine, used to perform arithmetic, whose name is known | |
2765 | specially by the compiler and was not mentioned in the C code being | |
2766 | compiled. | |
2767 | ||
2768 | The definition of @code{LIBRARY_VALUE} need not be concerned aggregate | |
2769 | data types, because none of the library functions returns such types. | |
2770 | ||
2771 | @findex FUNCTION_VALUE_REGNO_P | |
2772 | @item FUNCTION_VALUE_REGNO_P (@var{regno}) | |
2773 | A C expression that is nonzero if @var{regno} is the number of a hard | |
2774 | register in which the values of called function may come back. | |
2775 | ||
2776 | A register whose use for returning values is limited to serving as the | |
2777 | second of a pair (for a value of type @code{double}, say) need not be | |
2778 | recognized by this macro. So for most machines, this definition | |
2779 | suffices: | |
2780 | ||
2781 | @example | |
2782 | #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0) | |
2783 | @end example | |
2784 | ||
2785 | If the machine has register windows, so that the caller and the called | |
2786 | function use different registers for the return value, this macro | |
2787 | should recognize only the caller's register numbers. | |
2788 | ||
2789 | @findex APPLY_RESULT_SIZE | |
2790 | @item APPLY_RESULT_SIZE | |
2791 | Define this macro if @samp{untyped_call} and @samp{untyped_return} | |
2792 | need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for | |
2793 | saving and restoring an arbitrary return value. | |
2794 | @end table | |
2795 | ||
2796 | @node Aggregate Return | |
2797 | @subsection How Large Values Are Returned | |
2798 | @cindex aggregates as return values | |
2799 | @cindex large return values | |
2800 | @cindex returning aggregate values | |
2801 | @cindex structure value address | |
2802 | ||
2803 | When a function value's mode is @code{BLKmode} (and in some other | |
2804 | cases), the value is not returned according to @code{FUNCTION_VALUE} | |
2805 | (@pxref{Scalar Return}). Instead, the caller passes the address of a | |
2806 | block of memory in which the value should be stored. This address | |
2807 | is called the @dfn{structure value address}. | |
2808 | ||
2809 | This section describes how to control returning structure values in | |
2810 | memory. | |
2811 | ||
2812 | @table @code | |
2813 | @findex RETURN_IN_MEMORY | |
2814 | @item RETURN_IN_MEMORY (@var{type}) | |
2815 | A C expression which can inhibit the returning of certain function | |
2816 | values in registers, based on the type of value. A nonzero value says | |
2817 | to return the function value in memory, just as large structures are | |
2818 | always returned. Here @var{type} will be a C expression of type | |
2819 | @code{tree}, representing the data type of the value. | |
2820 | ||
2821 | Note that values of mode @code{BLKmode} must be explicitly handled | |
2822 | by this macro. Also, the option @samp{-fpcc-struct-return} | |
2823 | takes effect regardless of this macro. On most systems, it is | |
2824 | possible to leave the macro undefined; this causes a default | |
2825 | definition to be used, whose value is the constant 1 for @code{BLKmode} | |
2826 | values, and 0 otherwise. | |
2827 | ||
2828 | Do not use this macro to indicate that structures and unions should always | |
2829 | be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN} | |
2830 | to indicate this. | |
2831 | ||
2832 | @findex DEFAULT_PCC_STRUCT_RETURN | |
2833 | @item DEFAULT_PCC_STRUCT_RETURN | |
2834 | Define this macro to be 1 if all structure and union return values must be | |
2835 | in memory. Since this results in slower code, this should be defined | |
2836 | only if needed for compatibility with other compilers or with an ABI. | |
2837 | If you define this macro to be 0, then the conventions used for structure | |
2838 | and union return values are decided by the @code{RETURN_IN_MEMORY} macro. | |
2839 | ||
2840 | If not defined, this defaults to the value 1. | |
2841 | ||
2842 | @findex STRUCT_VALUE_REGNUM | |
2843 | @item STRUCT_VALUE_REGNUM | |
2844 | If the structure value address is passed in a register, then | |
2845 | @code{STRUCT_VALUE_REGNUM} should be the number of that register. | |
2846 | ||
2847 | @findex STRUCT_VALUE | |
2848 | @item STRUCT_VALUE | |
2849 | If the structure value address is not passed in a register, define | |
2850 | @code{STRUCT_VALUE} as an expression returning an RTX for the place | |
2851 | where the address is passed. If it returns 0, the address is passed as | |
2852 | an ``invisible'' first argument. | |
2853 | ||
2854 | @findex STRUCT_VALUE_INCOMING_REGNUM | |
2855 | @item STRUCT_VALUE_INCOMING_REGNUM | |
2856 | On some architectures the place where the structure value address | |
2857 | is found by the called function is not the same place that the | |
2858 | caller put it. This can be due to register windows, or it could | |
2859 | be because the function prologue moves it to a different place. | |
2860 | ||
2861 | If the incoming location of the structure value address is in a | |
2862 | register, define this macro as the register number. | |
2863 | ||
2864 | @findex STRUCT_VALUE_INCOMING | |
2865 | @item STRUCT_VALUE_INCOMING | |
2866 | If the incoming location is not a register, then you should define | |
2867 | @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the | |
2868 | called function should find the value. If it should find the value on | |
2869 | the stack, define this to create a @code{mem} which refers to the frame | |
2870 | pointer. A definition of 0 means that the address is passed as an | |
2871 | ``invisible'' first argument. | |
2872 | ||
2873 | @findex PCC_STATIC_STRUCT_RETURN | |
2874 | @item PCC_STATIC_STRUCT_RETURN | |
2875 | Define this macro if the usual system convention on the target machine | |
2876 | for returning structures and unions is for the called function to return | |
2877 | the address of a static variable containing the value. | |
2878 | ||
2879 | Do not define this if the usual system convention is for the caller to | |
2880 | pass an address to the subroutine. | |
2881 | ||
2882 | This macro has effect in @samp{-fpcc-struct-return} mode, but it does | |
2883 | nothing when you use @samp{-freg-struct-return} mode. | |
2884 | @end table | |
2885 | ||
2886 | @node Caller Saves | |
2887 | @subsection Caller-Saves Register Allocation | |
2888 | ||
2889 | If you enable it, GNU CC can save registers around function calls. This | |
2890 | makes it possible to use call-clobbered registers to hold variables that | |
2891 | must live across calls. | |
2892 | ||
2893 | @table @code | |
2894 | @findex DEFAULT_CALLER_SAVES | |
2895 | @item DEFAULT_CALLER_SAVES | |
2896 | Define this macro if function calls on the target machine do not preserve | |
2897 | any registers; in other words, if @code{CALL_USED_REGISTERS} has 1 | |
2898 | for all registers. This macro enables @samp{-fcaller-saves} by default. | |
2899 | Eventually that option will be enabled by default on all machines and both | |
2900 | the option and this macro will be eliminated. | |
2901 | ||
2902 | @findex CALLER_SAVE_PROFITABLE | |
2903 | @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls}) | |
2904 | A C expression to determine whether it is worthwhile to consider placing | |
2905 | a pseudo-register in a call-clobbered hard register and saving and | |
2906 | restoring it around each function call. The expression should be 1 when | |
2907 | this is worth doing, and 0 otherwise. | |
2908 | ||
2909 | If you don't define this macro, a default is used which is good on most | |
2910 | machines: @code{4 * @var{calls} < @var{refs}}. | |
2911 | @end table | |
2912 | ||
2913 | @node Function Entry | |
2914 | @subsection Function Entry and Exit | |
2915 | @cindex function entry and exit | |
2916 | @cindex prologue | |
2917 | @cindex epilogue | |
2918 | ||
2919 | This section describes the macros that output function entry | |
2920 | (@dfn{prologue}) and exit (@dfn{epilogue}) code. | |
2921 | ||
2922 | @table @code | |
2923 | @findex FUNCTION_PROLOGUE | |
2924 | @item FUNCTION_PROLOGUE (@var{file}, @var{size}) | |
2925 | A C compound statement that outputs the assembler code for entry to a | |
2926 | function. The prologue is responsible for setting up the stack frame, | |
2927 | initializing the frame pointer register, saving registers that must be | |
2928 | saved, and allocating @var{size} additional bytes of storage for the | |
2929 | local variables. @var{size} is an integer. @var{file} is a stdio | |
2930 | stream to which the assembler code should be output. | |
2931 | ||
2932 | The label for the beginning of the function need not be output by this | |
2933 | macro. That has already been done when the macro is run. | |
2934 | ||
2935 | @findex regs_ever_live | |
2936 | To determine which registers to save, the macro can refer to the array | |
2937 | @code{regs_ever_live}: element @var{r} is nonzero if hard register | |
2938 | @var{r} is used anywhere within the function. This implies the function | |
2939 | prologue should save register @var{r}, provided it is not one of the | |
2940 | call-used registers. (@code{FUNCTION_EPILOGUE} must likewise use | |
2941 | @code{regs_ever_live}.) | |
2942 | ||
2943 | On machines that have ``register windows'', the function entry code does | |
2944 | not save on the stack the registers that are in the windows, even if | |
2945 | they are supposed to be preserved by function calls; instead it takes | |
2946 | appropriate steps to ``push'' the register stack, if any non-call-used | |
2947 | registers are used in the function. | |
2948 | ||
2949 | @findex frame_pointer_needed | |
2950 | On machines where functions may or may not have frame-pointers, the | |
2951 | function entry code must vary accordingly; it must set up the frame | |
2952 | pointer if one is wanted, and not otherwise. To determine whether a | |
2953 | frame pointer is in wanted, the macro can refer to the variable | |
2954 | @code{frame_pointer_needed}. The variable's value will be 1 at run | |
2955 | time in a function that needs a frame pointer. @xref{Elimination}. | |
2956 | ||
2957 | The function entry code is responsible for allocating any stack space | |
2958 | required for the function. This stack space consists of the regions | |
2959 | listed below. In most cases, these regions are allocated in the | |
2960 | order listed, with the last listed region closest to the top of the | |
2961 | stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and | |
2962 | the highest address if it is not defined). You can use a different order | |
2963 | for a machine if doing so is more convenient or required for | |
2964 | compatibility reasons. Except in cases where required by standard | |
2965 | or by a debugger, there is no reason why the stack layout used by GCC | |
2966 | need agree with that used by other compilers for a machine. | |
2967 | ||
2968 | @itemize @bullet | |
2969 | @item | |
2970 | @findex current_function_pretend_args_size | |
2971 | A region of @code{current_function_pretend_args_size} bytes of | |
2972 | uninitialized space just underneath the first argument arriving on the | |
2973 | stack. (This may not be at the very start of the allocated stack region | |
2974 | if the calling sequence has pushed anything else since pushing the stack | |
2975 | arguments. But usually, on such machines, nothing else has been pushed | |
2976 | yet, because the function prologue itself does all the pushing.) This | |
2977 | region is used on machines where an argument may be passed partly in | |
2978 | registers and partly in memory, and, in some cases to support the | |
2979 | features in @file{varargs.h} and @file{stdargs.h}. | |
2980 | ||
2981 | @item | |
2982 | An area of memory used to save certain registers used by the function. | |
2983 | The size of this area, which may also include space for such things as | |
2984 | the return address and pointers to previous stack frames, is | |
2985 | machine-specific and usually depends on which registers have been used | |
2986 | in the function. Machines with register windows often do not require | |
2987 | a save area. | |
2988 | ||
2989 | @item | |
2990 | A region of at least @var{size} bytes, possibly rounded up to an allocation | |
2991 | boundary, to contain the local variables of the function. On some machines, | |
2992 | this region and the save area may occur in the opposite order, with the | |
2993 | save area closer to the top of the stack. | |
2994 | ||
2995 | @item | |
2996 | @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames | |
2997 | Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of | |
2998 | @code{current_function_outgoing_args_size} bytes to be used for outgoing | |
2999 | argument lists of the function. @xref{Stack Arguments}. | |
3000 | @end itemize | |
3001 | ||
3002 | Normally, it is necessary for the macros @code{FUNCTION_PROLOGUE} and | |
3003 | @code{FUNCTION_EPILOGUE} to treat leaf functions specially. The C | |
3004 | variable @code{leaf_function} is nonzero for such a function. | |
3005 | ||
3006 | @findex EXIT_IGNORE_STACK | |
3007 | @item EXIT_IGNORE_STACK | |
3008 | Define this macro as a C expression that is nonzero if the return | |
3009 | instruction or the function epilogue ignores the value of the stack | |
3010 | pointer; in other words, if it is safe to delete an instruction to | |
3011 | adjust the stack pointer before a return from the function. | |
3012 | ||
3013 | Note that this macro's value is relevant only for functions for which | |
3014 | frame pointers are maintained. It is never safe to delete a final | |
3015 | stack adjustment in a function that has no frame pointer, and the | |
3016 | compiler knows this regardless of @code{EXIT_IGNORE_STACK}. | |
3017 | ||
3018 | @findex EPILOGUE_USES | |
3019 | @item EPILOGUE_USES (@var{regno}) | |
3020 | Define this macro as a C expression that is nonzero for registers are | |
3021 | used by the epilogue or the @samp{return} pattern. The stack and frame | |
3022 | pointer registers are already be assumed to be used as needed. | |
3023 | ||
3024 | @findex FUNCTION_EPILOGUE | |
3025 | @item FUNCTION_EPILOGUE (@var{file}, @var{size}) | |
3026 | A C compound statement that outputs the assembler code for exit from a | |
3027 | function. The epilogue is responsible for restoring the saved | |
3028 | registers and stack pointer to their values when the function was | |
3029 | called, and returning control to the caller. This macro takes the | |
3030 | same arguments as the macro @code{FUNCTION_PROLOGUE}, and the | |
3031 | registers to restore are determined from @code{regs_ever_live} and | |
3032 | @code{CALL_USED_REGISTERS} in the same way. | |
3033 | ||
3034 | On some machines, there is a single instruction that does all the work | |
3035 | of returning from the function. On these machines, give that | |
3036 | instruction the name @samp{return} and do not define the macro | |
3037 | @code{FUNCTION_EPILOGUE} at all. | |
3038 | ||
3039 | Do not define a pattern named @samp{return} if you want the | |
3040 | @code{FUNCTION_EPILOGUE} to be used. If you want the target switches | |
3041 | to control whether return instructions or epilogues are used, define a | |
3042 | @samp{return} pattern with a validity condition that tests the target | |
3043 | switches appropriately. If the @samp{return} pattern's validity | |
3044 | condition is false, epilogues will be used. | |
3045 | ||
3046 | On machines where functions may or may not have frame-pointers, the | |
3047 | function exit code must vary accordingly. Sometimes the code for these | |
3048 | two cases is completely different. To determine whether a frame pointer | |
3049 | is wanted, the macro can refer to the variable | |
3050 | @code{frame_pointer_needed}. The variable's value will be 1 when compiling | |
3051 | a function that needs a frame pointer. | |
3052 | ||
3053 | Normally, @code{FUNCTION_PROLOGUE} and @code{FUNCTION_EPILOGUE} must | |
3054 | treat leaf functions specially. The C variable @code{leaf_function} is | |
3055 | nonzero for such a function. @xref{Leaf Functions}. | |
3056 | ||
3057 | On some machines, some functions pop their arguments on exit while | |
3058 | others leave that for the caller to do. For example, the 68020 when | |
3059 | given @samp{-mrtd} pops arguments in functions that take a fixed | |
3060 | number of arguments. | |
3061 | ||
3062 | @findex current_function_pops_args | |
3063 | Your definition of the macro @code{RETURN_POPS_ARGS} decides which | |
3064 | functions pop their own arguments. @code{FUNCTION_EPILOGUE} needs to | |
3065 | know what was decided. The variable that is called | |
3066 | @code{current_function_pops_args} is the number of bytes of its | |
3067 | arguments that a function should pop. @xref{Scalar Return}. | |
3068 | @c what is the "its arguments" in the above sentence referring to, pray | |
3069 | @c tell? --mew 5feb93 | |
3070 | ||
3071 | @findex DELAY_SLOTS_FOR_EPILOGUE | |
3072 | @item DELAY_SLOTS_FOR_EPILOGUE | |
3073 | Define this macro if the function epilogue contains delay slots to which | |
3074 | instructions from the rest of the function can be ``moved''. The | |
3075 | definition should be a C expression whose value is an integer | |
3076 | representing the number of delay slots there. | |
3077 | ||
3078 | @findex ELIGIBLE_FOR_EPILOGUE_DELAY | |
3079 | @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n}) | |
3080 | A C expression that returns 1 if @var{insn} can be placed in delay | |
3081 | slot number @var{n} of the epilogue. | |
3082 | ||
3083 | The argument @var{n} is an integer which identifies the delay slot now | |
3084 | being considered (since different slots may have different rules of | |
3085 | eligibility). It is never negative and is always less than the number | |
3086 | of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns). | |
3087 | If you reject a particular insn for a given delay slot, in principle, it | |
3088 | may be reconsidered for a subsequent delay slot. Also, other insns may | |
3089 | (at least in principle) be considered for the so far unfilled delay | |
3090 | slot. | |
3091 | ||
3092 | @findex current_function_epilogue_delay_list | |
3093 | @findex final_scan_insn | |
3094 | The insns accepted to fill the epilogue delay slots are put in an RTL | |
3095 | list made with @code{insn_list} objects, stored in the variable | |
3096 | @code{current_function_epilogue_delay_list}. The insn for the first | |
3097 | delay slot comes first in the list. Your definition of the macro | |
3098 | @code{FUNCTION_EPILOGUE} should fill the delay slots by outputting the | |
3099 | insns in this list, usually by calling @code{final_scan_insn}. | |
3100 | ||
3101 | You need not define this macro if you did not define | |
3102 | @code{DELAY_SLOTS_FOR_EPILOGUE}. | |
3103 | ||
3104 | @findex ASM_OUTPUT_MI_THUNK | |
3105 | @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function}) | |
3106 | A C compound statement that outputs the assembler code for a thunk | |
3107 | function, used to implement C++ virtual function calls with multiple | |
3108 | inheritance. The thunk acts as a wrapper around a virtual function, | |
3109 | adjusting the implicit object parameter before handing control off to | |
3110 | the real function. | |
3111 | ||
3112 | First, emit code to add the integer @var{delta} to the location that | |
3113 | contains the incoming first argument. Assume that this argument | |
3114 | contains a pointer, and is the one used to pass the @code{this} pointer | |
3115 | in C++. This is the incoming argument @emph{before} the function prologue, | |
3116 | e.g. @samp{%o0} on a sparc. The addition must preserve the values of | |
3117 | all other incoming arguments. | |
3118 | ||
3119 | After the addition, emit code to jump to @var{function}, which is a | |
3120 | @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does | |
3121 | not touch the return address. Hence returning from @var{FUNCTION} will | |
3122 | return to whoever called the current @samp{thunk}. | |
3123 | ||
3124 | The effect must be as if @var{function} had been called directly with | |
3125 | the adjusted first argument. This macro is responsible for emitting all | |
3126 | of the code for a thunk function; @code{FUNCTION_PROLOGUE} and | |
3127 | @code{FUNCTION_EPILOGUE} are not invoked. | |
3128 | ||
3129 | The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function} | |
3130 | have already been extracted from it.) It might possibly be useful on | |
3131 | some targets, but probably not. | |
3132 | ||
3133 | For many targets, the target-independent code in the C++ frontend will | |
3134 | be sufficient and you can leave this macro undefined. You need to | |
3135 | define this macro if the code generated by default would clobber any of | |
3136 | the incoming arguments; this is only likely if parameters can be passed | |
3137 | in registers. You should also define this macro if the default code is | |
3138 | sub-optimal. | |
3139 | @end table | |
3140 | ||
3141 | @node Profiling | |
3142 | @subsection Generating Code for Profiling | |
3143 | @cindex profiling, code generation | |
3144 | ||
3145 | These macros will help you generate code for profiling. | |
3146 | ||
3147 | @table @code | |
3148 | @findex FUNCTION_PROFILER | |
3149 | @item FUNCTION_PROFILER (@var{file}, @var{labelno}) | |
3150 | A C statement or compound statement to output to @var{file} some | |
3151 | assembler code to call the profiling subroutine @code{mcount}. | |
3152 | Before calling, the assembler code must load the address of a | |
3153 | counter variable into a register where @code{mcount} expects to | |
3154 | find the address. The name of this variable is @samp{LP} followed | |
3155 | by the number @var{labelno}, so you would generate the name using | |
3156 | @samp{LP%d} in a @code{fprintf}. | |
3157 | ||
3158 | @findex mcount | |
3159 | The details of how the address should be passed to @code{mcount} are | |
3160 | determined by your operating system environment, not by GNU CC. To | |
3161 | figure them out, compile a small program for profiling using the | |
3162 | system's installed C compiler and look at the assembler code that | |
3163 | results. | |
3164 | ||
3165 | @findex PROFILE_BEFORE_PROLOGUE | |
3166 | @item PROFILE_BEFORE_PROLOGUE | |
3167 | Define this macro if the code for function profiling should come before | |
3168 | the function prologue. Normally, the profiling code comes after. | |
3169 | ||
3170 | @findex FUNCTION_BLOCK_PROFILER | |
3171 | @vindex profile_block_flag | |
3172 | @item FUNCTION_BLOCK_PROFILER (@var{file}, @var{labelno}) | |
3173 | A C statement or compound statement to output to @var{file} some | |
3174 | assembler code to initialize basic-block profiling for the current | |
3175 | object module. The global compile flag @code{profile_block_flag} | |
3176 | distingishes two profile modes. | |
3177 | ||
3178 | @table @code | |
3179 | @findex __bb_init_func | |
3180 | @item profile_block_flag != 2 | |
3181 | Output code to call the subroutine @code{__bb_init_func} once per | |
3182 | object module, passing it as its sole argument the address of a block | |
3183 | allocated in the object module. | |
3184 | ||
3185 | The name of the block is a local symbol made with this statement: | |
3186 | ||
3187 | @smallexample | |
3188 | ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0); | |
3189 | @end smallexample | |
3190 | ||
3191 | Of course, since you are writing the definition of | |
3192 | @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you | |
3193 | can take a short cut in the definition of this macro and use the name | |
3194 | that you know will result. | |
3195 | ||
3196 | The first word of this block is a flag which will be nonzero if the | |
3197 | object module has already been initialized. So test this word first, | |
3198 | and do not call @code{__bb_init_func} if the flag is | |
3199 | nonzero. BLOCK_OR_LABEL contains a unique number which may be used to | |
3200 | generate a label as a branch destination when @code{__bb_init_func} | |
3201 | will not be called. | |
3202 | ||
3203 | Described in assembler language, the code to be output looks like: | |
3204 | ||
3205 | @example | |
3206 | cmp (LPBX0),0 | |
3207 | bne local_label | |
3208 | parameter1 <- LPBX0 | |
3209 | call __bb_init_func | |
3210 | local_label: | |
3211 | @end example | |
3212 | ||
3213 | @findex __bb_init_trace_func | |
3214 | @item profile_block_flag == 2 | |
3215 | Output code to call the subroutine @code{__bb_init_trace_func} | |
3216 | and pass two parameters to it. The first parameter is the same as | |
3217 | for @code{__bb_init_func}. The second parameter is the number of the | |
3218 | first basic block of the function as given by BLOCK_OR_LABEL. Note | |
3219 | that @code{__bb_init_trace_func} has to be called, even if the object | |
3220 | module has been initialized already. | |
3221 | ||
3222 | Described in assembler language, the code to be output looks like: | |
3223 | @example | |
3224 | parameter1 <- LPBX0 | |
3225 | parameter2 <- BLOCK_OR_LABEL | |
3226 | call __bb_init_trace_func | |
3227 | @end example | |
3228 | @end table | |
3229 | ||
3230 | @findex BLOCK_PROFILER | |
3231 | @vindex profile_block_flag | |
3232 | @item BLOCK_PROFILER (@var{file}, @var{blockno}) | |
3233 | A C statement or compound statement to output to @var{file} some | |
3234 | assembler code to increment the count associated with the basic | |
3235 | block number @var{blockno}. The global compile flag | |
3236 | @code{profile_block_flag} distingishes two profile modes. | |
3237 | ||
3238 | @table @code | |
3239 | @item profile_block_flag != 2 | |
3240 | Output code to increment the counter directly. Basic blocks are | |
3241 | numbered separately from zero within each compilation. The count | |
3242 | associated with block number @var{blockno} is at index | |
3243 | @var{blockno} in a vector of words; the name of this array is a local | |
3244 | symbol made with this statement: | |
3245 | ||
3246 | @smallexample | |
3247 | ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 2); | |
3248 | @end smallexample | |
3249 | ||
3250 | @c This paragraph is the same as one a few paragraphs up. | |
3251 | @c That is not an error. | |
3252 | Of course, since you are writing the definition of | |
3253 | @code{ASM_GENERATE_INTERNAL_LABEL} as well as that of this macro, you | |
3254 | can take a short cut in the definition of this macro and use the name | |
3255 | that you know will result. | |
3256 | ||
3257 | Described in assembler language, the code to be output looks like: | |
3258 | ||
3259 | @smallexample | |
3260 | inc (LPBX2+4*BLOCKNO) | |
3261 | @end smallexample | |
3262 | ||
3263 | @vindex __bb | |
3264 | @findex __bb_trace_func | |
3265 | @item profile_block_flag == 2 | |
3266 | Output code to initialize the global structure @code{__bb} and | |
3267 | call the function @code{__bb_trace_func}, which will increment the | |
3268 | counter. | |
3269 | ||
3270 | @code{__bb} consists of two words. In the first word, the current | |
3271 | basic block number, as given by BLOCKNO, has to be stored. In | |
3272 | the second word, the address of a block allocated in the object | |
3273 | module has to be stored. The address is given by the label created | |
3274 | with this statement: | |
3275 | ||
3276 | @smallexample | |
3277 | ASM_GENERATE_INTERNAL_LABEL (@var{buffer}, "LPBX", 0); | |
3278 | @end smallexample | |
3279 | ||
3280 | Described in assembler language, the code to be output looks like: | |
3281 | @example | |
3282 | move BLOCKNO -> (__bb) | |
3283 | move LPBX0 -> (__bb+4) | |
3284 | call __bb_trace_func | |
3285 | @end example | |
3286 | @end table | |
3287 | ||
3288 | @findex FUNCTION_BLOCK_PROFILER_EXIT | |
3289 | @findex __bb_trace_ret | |
3290 | @vindex profile_block_flag | |
3291 | @item FUNCTION_BLOCK_PROFILER_EXIT (@var{file}) | |
3292 | A C statement or compound statement to output to @var{file} | |
3293 | assembler code to call function @code{__bb_trace_ret}. The | |
3294 | assembler code should only be output | |
3295 | if the global compile flag @code{profile_block_flag} == 2. This | |
3296 | macro has to be used at every place where code for returning from | |
3297 | a function is generated (e.g. @code{FUNCTION_EPILOGUE}). Although | |
3298 | you have to write the definition of @code{FUNCTION_EPILOGUE} | |
3299 | as well, you have to define this macro to tell the compiler, that | |
3300 | the proper call to @code{__bb_trace_ret} is produced. | |
3301 | ||
3302 | @findex MACHINE_STATE_SAVE | |
3303 | @findex __bb_init_trace_func | |
3304 | @findex __bb_trace_func | |
3305 | @findex __bb_trace_ret | |
3306 | @item MACHINE_STATE_SAVE (@var{id}) | |
3307 | A C statement or compound statement to save all registers, which may | |
3308 | be clobbered by a function call, including condition codes. The | |
3309 | @code{asm} statement will be mostly likely needed to handle this | |
3310 | task. Local labels in the assembler code can be concatenated with the | |
3311 | string @var{id}, to obtain a unique lable name. | |
3312 | ||
3313 | Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or | |
3314 | @code{FUNCTION_EPILOGUE} must be saved in the macros | |
3315 | @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and | |
3316 | @code{BLOCK_PROFILER} prior calling @code{__bb_init_trace_func}, | |
3317 | @code{__bb_trace_ret} and @code{__bb_trace_func} respectively. | |
3318 | ||
3319 | @findex MACHINE_STATE_RESTORE | |
3320 | @findex __bb_init_trace_func | |
3321 | @findex __bb_trace_func | |
3322 | @findex __bb_trace_ret | |
3323 | @item MACHINE_STATE_RESTORE (@var{id}) | |
3324 | A C statement or compound statement to restore all registers, including | |
3325 | condition codes, saved by @code{MACHINE_STATE_SAVE}. | |
3326 | ||
3327 | Registers or condition codes clobbered by @code{FUNCTION_PROLOGUE} or | |
3328 | @code{FUNCTION_EPILOGUE} must be restored in the macros | |
3329 | @code{FUNCTION_BLOCK_PROFILER}, @code{FUNCTION_BLOCK_PROFILER_EXIT} and | |
3330 | @code{BLOCK_PROFILER} after calling @code{__bb_init_trace_func}, | |
3331 | @code{__bb_trace_ret} and @code{__bb_trace_func} respectively. | |
3332 | ||
3333 | @findex BLOCK_PROFILER_CODE | |
3334 | @item BLOCK_PROFILER_CODE | |
3335 | A C function or functions which are needed in the library to | |
3336 | support block profiling. | |
3337 | @end table | |
3338 | ||
3339 | @node Varargs | |
3340 | @section Implementing the Varargs Macros | |
3341 | @cindex varargs implementation | |
3342 | ||
3343 | GNU CC comes with an implementation of @file{varargs.h} and | |
3344 | @file{stdarg.h} that work without change on machines that pass arguments | |
3345 | on the stack. Other machines require their own implementations of | |
3346 | varargs, and the two machine independent header files must have | |
3347 | conditionals to include it. | |
3348 | ||
3349 | ANSI @file{stdarg.h} differs from traditional @file{varargs.h} mainly in | |
3350 | the calling convention for @code{va_start}. The traditional | |
3351 | implementation takes just one argument, which is the variable in which | |
3352 | to store the argument pointer. The ANSI implementation of | |
3353 | @code{va_start} takes an additional second argument. The user is | |
3354 | supposed to write the last named argument of the function here. | |
3355 | ||
3356 | However, @code{va_start} should not use this argument. The way to find | |
3357 | the end of the named arguments is with the built-in functions described | |
3358 | below. | |
3359 | ||
3360 | @table @code | |
3361 | @findex __builtin_saveregs | |
3362 | @item __builtin_saveregs () | |
3363 | Use this built-in function to save the argument registers in memory so | |
3364 | that the varargs mechanism can access them. Both ANSI and traditional | |
3365 | versions of @code{va_start} must use @code{__builtin_saveregs}, unless | |
3366 | you use @code{SETUP_INCOMING_VARARGS} (see below) instead. | |
3367 | ||
3368 | On some machines, @code{__builtin_saveregs} is open-coded under the | |
3369 | control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines, | |
3370 | it calls a routine written in assembler language, found in | |
3371 | @file{libgcc2.c}. | |
3372 | ||
3373 | Code generated for the call to @code{__builtin_saveregs} appears at the | |
3374 | beginning of the function, as opposed to where the call to | |
3375 | @code{__builtin_saveregs} is written, regardless of what the code is. | |
3376 | This is because the registers must be saved before the function starts | |
3377 | to use them for its own purposes. | |
3378 | @c i rewrote the first sentence above to fix an overfull hbox. --mew | |
3379 | @c 10feb93 | |
3380 | ||
3381 | @findex __builtin_args_info | |
3382 | @item __builtin_args_info (@var{category}) | |
3383 | Use this built-in function to find the first anonymous arguments in | |
3384 | registers. | |
3385 | ||
3386 | In general, a machine may have several categories of registers used for | |
3387 | arguments, each for a particular category of data types. (For example, | |
3388 | on some machines, floating-point registers are used for floating-point | |
3389 | arguments while other arguments are passed in the general registers.) | |
3390 | To make non-varargs functions use the proper calling convention, you | |
3391 | have defined the @code{CUMULATIVE_ARGS} data type to record how many | |
3392 | registers in each category have been used so far | |
3393 | ||
3394 | @code{__builtin_args_info} accesses the same data structure of type | |
3395 | @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished | |
3396 | with it, with @var{category} specifying which word to access. Thus, the | |
3397 | value indicates the first unused register in a given category. | |
3398 | ||
3399 | Normally, you would use @code{__builtin_args_info} in the implementation | |
3400 | of @code{va_start}, accessing each category just once and storing the | |
3401 | value in the @code{va_list} object. This is because @code{va_list} will | |
3402 | have to update the values, and there is no way to alter the | |
3403 | values accessed by @code{__builtin_args_info}. | |
3404 | ||
3405 | @findex __builtin_next_arg | |
3406 | @item __builtin_next_arg (@var{lastarg}) | |
3407 | This is the equivalent of @code{__builtin_args_info}, for stack | |
3408 | arguments. It returns the address of the first anonymous stack | |
3409 | argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it | |
3410 | returns the address of the location above the first anonymous stack | |
3411 | argument. Use it in @code{va_start} to initialize the pointer for | |
3412 | fetching arguments from the stack. Also use it in @code{va_start} to | |
3413 | verify that the second parameter @var{lastarg} is the last named argument | |
3414 | of the current function. | |
3415 | ||
3416 | @findex __builtin_classify_type | |
3417 | @item __builtin_classify_type (@var{object}) | |
3418 | Since each machine has its own conventions for which data types are | |
3419 | passed in which kind of register, your implementation of @code{va_arg} | |
3420 | has to embody these conventions. The easiest way to categorize the | |
3421 | specified data type is to use @code{__builtin_classify_type} together | |
3422 | with @code{sizeof} and @code{__alignof__}. | |
3423 | ||
3424 | @code{__builtin_classify_type} ignores the value of @var{object}, | |
3425 | considering only its data type. It returns an integer describing what | |
3426 | kind of type that is---integer, floating, pointer, structure, and so on. | |
3427 | ||
3428 | The file @file{typeclass.h} defines an enumeration that you can use to | |
3429 | interpret the values of @code{__builtin_classify_type}. | |
3430 | @end table | |
3431 | ||
3432 | These machine description macros help implement varargs: | |
3433 | ||
3434 | @table @code | |
3435 | @findex EXPAND_BUILTIN_SAVEREGS | |
3436 | @item EXPAND_BUILTIN_SAVEREGS (@var{args}) | |
3437 | If defined, is a C expression that produces the machine-specific code | |
3438 | for a call to @code{__builtin_saveregs}. This code will be moved to the | |
3439 | very beginning of the function, before any parameter access are made. | |
3440 | The return value of this function should be an RTX that contains the | |
3441 | value to use as the return of @code{__builtin_saveregs}. | |
3442 | ||
3443 | The argument @var{args} is a @code{tree_list} containing the arguments | |
3444 | that were passed to @code{__builtin_saveregs}. | |
3445 | ||
3446 | If this macro is not defined, the compiler will output an ordinary | |
3447 | call to the library function @samp{__builtin_saveregs}. | |
3448 | ||
3449 | @c !!! a bug in texinfo; how to make the entry on the @item line allow | |
3450 | @c more than one line of text... help... --mew 10feb93 | |
3451 | @findex SETUP_INCOMING_VARARGS | |
3452 | @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, | |
3453 | @var{pretend_args_size}, @var{second_time}) | |
3454 | This macro offers an alternative to using @code{__builtin_saveregs} and | |
3455 | defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the | |
3456 | anonymous register arguments into the stack so that all the arguments | |
3457 | appear to have been passed consecutively on the stack. Once this is | |
3458 | done, you can use the standard implementation of varargs that works for | |
3459 | machines that pass all their arguments on the stack. | |
3460 | ||
3461 | The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data | |
3462 | structure, containing the values that obtain after processing of the | |
3463 | named arguments. The arguments @var{mode} and @var{type} describe the | |
3464 | last named argument---its machine mode and its data type as a tree node. | |
3465 | ||
3466 | The macro implementation should do two things: first, push onto the | |
3467 | stack all the argument registers @emph{not} used for the named | |
3468 | arguments, and second, store the size of the data thus pushed into the | |
3469 | @code{int}-valued variable whose name is supplied as the argument | |
3470 | @var{pretend_args_size}. The value that you store here will serve as | |
3471 | additional offset for setting up the stack frame. | |
3472 | ||
3473 | Because you must generate code to push the anonymous arguments at | |
3474 | compile time without knowing their data types, | |
3475 | @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just | |
3476 | a single category of argument register and use it uniformly for all data | |
3477 | types. | |
3478 | ||
3479 | If the argument @var{second_time} is nonzero, it means that the | |
3480 | arguments of the function are being analyzed for the second time. This | |
3481 | happens for an inline function, which is not actually compiled until the | |
3482 | end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should | |
3483 | not generate any instructions in this case. | |
3484 | ||
3485 | @findex STRICT_ARGUMENT_NAMING | |
3486 | @item STRICT_ARGUMENT_NAMING | |
3487 | Define this macro if the location where a function argument is passed | |
3488 | depends on whether or not it is a named argument. | |
3489 | ||
3490 | This macro controls how the @var{named} argument to @code{FUNCTION_ARG} | |
3491 | is set for varargs and stdarg functions. With this macro defined, | |
3492 | the @var{named} argument is always true for named arguments, and false for | |
3493 | unnamed arguments. If this is not defined, but @code{SETUP_INCOMING_VARARGS} | |
3494 | is defined, then all arguments are treated as named. Otherwise, all named | |
3495 | arguments except the last are treated as named. | |
3496 | @end table | |
3497 | ||
3498 | @node Trampolines | |
3499 | @section Trampolines for Nested Functions | |
3500 | @cindex trampolines for nested functions | |
3501 | @cindex nested functions, trampolines for | |
3502 | ||
3503 | A @dfn{trampoline} is a small piece of code that is created at run time | |
3504 | when the address of a nested function is taken. It normally resides on | |
3505 | the stack, in the stack frame of the containing function. These macros | |
3506 | tell GNU CC how to generate code to allocate and initialize a | |
3507 | trampoline. | |
3508 | ||
3509 | The instructions in the trampoline must do two things: load a constant | |
3510 | address into the static chain register, and jump to the real address of | |
3511 | the nested function. On CISC machines such as the m68k, this requires | |
3512 | two instructions, a move immediate and a jump. Then the two addresses | |
3513 | exist in the trampoline as word-long immediate operands. On RISC | |
3514 | machines, it is often necessary to load each address into a register in | |
3515 | two parts. Then pieces of each address form separate immediate | |
3516 | operands. | |
3517 | ||
3518 | The code generated to initialize the trampoline must store the variable | |
3519 | parts---the static chain value and the function address---into the | |
3520 | immediate operands of the instructions. On a CISC machine, this is | |
3521 | simply a matter of copying each address to a memory reference at the | |
3522 | proper offset from the start of the trampoline. On a RISC machine, it | |
3523 | may be necessary to take out pieces of the address and store them | |
3524 | separately. | |
3525 | ||
3526 | @table @code | |
3527 | @findex TRAMPOLINE_TEMPLATE | |
3528 | @item TRAMPOLINE_TEMPLATE (@var{file}) | |
3529 | A C statement to output, on the stream @var{file}, assembler code for a | |
3530 | block of data that contains the constant parts of a trampoline. This | |
3531 | code should not include a label---the label is taken care of | |
3532 | automatically. | |
3533 | ||
3534 | If you do not define this macro, it means no template is needed | |
3535 | for the target. Do not define this macro on systems where the block move | |
3536 | code to copy the trampoline into place would be larger than the code | |
3537 | to generate it on the spot. | |
3538 | ||
3539 | @findex TRAMPOLINE_SECTION | |
3540 | @item TRAMPOLINE_SECTION | |
3541 | The name of a subroutine to switch to the section in which the | |
3542 | trampoline template is to be placed (@pxref{Sections}). The default is | |
3543 | a value of @samp{readonly_data_section}, which places the trampoline in | |
3544 | the section containing read-only data. | |
3545 | ||
3546 | @findex TRAMPOLINE_SIZE | |
3547 | @item TRAMPOLINE_SIZE | |
3548 | A C expression for the size in bytes of the trampoline, as an integer. | |
3549 | ||
3550 | @findex TRAMPOLINE_ALIGNMENT | |
3551 | @item TRAMPOLINE_ALIGNMENT | |
3552 | Alignment required for trampolines, in bits. | |
3553 | ||
3554 | If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT} | |
3555 | is used for aligning trampolines. | |
3556 | ||
3557 | @findex INITIALIZE_TRAMPOLINE | |
3558 | @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain}) | |
3559 | A C statement to initialize the variable parts of a trampoline. | |
3560 | @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is | |
3561 | an RTX for the address of the nested function; @var{static_chain} is an | |
3562 | RTX for the static chain value that should be passed to the function | |
3563 | when it is called. | |
3564 | ||
3565 | @findex ALLOCATE_TRAMPOLINE | |
3566 | @item ALLOCATE_TRAMPOLINE (@var{fp}) | |
3567 | A C expression to allocate run-time space for a trampoline. The | |
3568 | expression value should be an RTX representing a memory reference to the | |
3569 | space for the trampoline. | |
3570 | ||
3571 | @cindex @code{FUNCTION_EPILOGUE} and trampolines | |
3572 | @cindex @code{FUNCTION_PROLOGUE} and trampolines | |
3573 | If this macro is not defined, by default the trampoline is allocated as | |
3574 | a stack slot. This default is right for most machines. The exceptions | |
3575 | are machines where it is impossible to execute instructions in the stack | |
3576 | area. On such machines, you may have to implement a separate stack, | |
3577 | using this macro in conjunction with @code{FUNCTION_PROLOGUE} and | |
3578 | @code{FUNCTION_EPILOGUE}. | |
3579 | ||
3580 | @var{fp} points to a data structure, a @code{struct function}, which | |
3581 | describes the compilation status of the immediate containing function of | |
3582 | the function which the trampoline is for. Normally (when | |
3583 | @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the | |
3584 | trampoline is in the stack frame of this containing function. Other | |
3585 | allocation strategies probably must do something analogous with this | |
3586 | information. | |
3587 | @end table | |
3588 | ||
3589 | Implementing trampolines is difficult on many machines because they have | |
3590 | separate instruction and data caches. Writing into a stack location | |
3591 | fails to clear the memory in the instruction cache, so when the program | |
3592 | jumps to that location, it executes the old contents. | |
3593 | ||
3594 | Here are two possible solutions. One is to clear the relevant parts of | |
3595 | the instruction cache whenever a trampoline is set up. The other is to | |
3596 | make all trampolines identical, by having them jump to a standard | |
3597 | subroutine. The former technique makes trampoline execution faster; the | |
3598 | latter makes initialization faster. | |
3599 | ||
3600 | To clear the instruction cache when a trampoline is initialized, define | |
3601 | the following macros which describe the shape of the cache. | |
3602 | ||
3603 | @table @code | |
3604 | @findex INSN_CACHE_SIZE | |
3605 | @item INSN_CACHE_SIZE | |
3606 | The total size in bytes of the cache. | |
3607 | ||
3608 | @findex INSN_CACHE_LINE_WIDTH | |
3609 | @item INSN_CACHE_LINE_WIDTH | |
3610 | The length in bytes of each cache line. The cache is divided into cache | |
3611 | lines which are disjoint slots, each holding a contiguous chunk of data | |
3612 | fetched from memory. Each time data is brought into the cache, an | |
3613 | entire line is read at once. The data loaded into a cache line is | |
3614 | always aligned on a boundary equal to the line size. | |
3615 | ||
3616 | @findex INSN_CACHE_DEPTH | |
3617 | @item INSN_CACHE_DEPTH | |
3618 | The number of alternative cache lines that can hold any particular memory | |
3619 | location. | |
3620 | @end table | |
3621 | ||
3622 | Alternatively, if the machine has system calls or instructions to clear | |
3623 | the instruction cache directly, you can define the following macro. | |
3624 | ||
3625 | @table @code | |
3626 | @findex CLEAR_INSN_CACHE | |
3627 | @item CLEAR_INSN_CACHE (@var{BEG}, @var{END}) | |
3628 | If defined, expands to a C expression clearing the @emph{instruction | |
3629 | cache} in the specified interval. If it is not defined, and the macro | |
3630 | INSN_CACHE_SIZE is defined, some generic code is generated to clear the | |
3631 | cache. The definition of this macro would typically be a series of | |
3632 | @code{asm} statements. Both @var{BEG} and @var{END} are both pointer | |
3633 | expressions. | |
3634 | @end table | |
3635 | ||
3636 | To use a standard subroutine, define the following macro. In addition, | |
3637 | you must make sure that the instructions in a trampoline fill an entire | |
3638 | cache line with identical instructions, or else ensure that the | |
3639 | beginning of the trampoline code is always aligned at the same point in | |
3640 | its cache line. Look in @file{m68k.h} as a guide. | |
3641 | ||
3642 | @table @code | |
3643 | @findex TRANSFER_FROM_TRAMPOLINE | |
3644 | @item TRANSFER_FROM_TRAMPOLINE | |
3645 | Define this macro if trampolines need a special subroutine to do their | |
3646 | work. The macro should expand to a series of @code{asm} statements | |
3647 | which will be compiled with GNU CC. They go in a library function named | |
3648 | @code{__transfer_from_trampoline}. | |
3649 | ||
3650 | If you need to avoid executing the ordinary prologue code of a compiled | |
3651 | C function when you jump to the subroutine, you can do so by placing a | |
3652 | special label of your own in the assembler code. Use one @code{asm} | |
3653 | statement to generate an assembler label, and another to make the label | |
3654 | global. Then trampolines can use that label to jump directly to your | |
3655 | special assembler code. | |
3656 | @end table | |
3657 | ||
3658 | @node Library Calls | |
3659 | @section Implicit Calls to Library Routines | |
3660 | @cindex library subroutine names | |
3661 | @cindex @file{libgcc.a} | |
3662 | ||
3663 | @c prevent bad page break with this line | |
3664 | Here is an explanation of implicit calls to library routines. | |
3665 | ||
3666 | @table @code | |
3667 | @findex MULSI3_LIBCALL | |
3668 | @item MULSI3_LIBCALL | |
3669 | A C string constant giving the name of the function to call for | |
3670 | multiplication of one signed full-word by another. If you do not | |
3671 | define this macro, the default name is used, which is @code{__mulsi3}, | |
3672 | a function defined in @file{libgcc.a}. | |
3673 | ||
3674 | @findex DIVSI3_LIBCALL | |
3675 | @item DIVSI3_LIBCALL | |
3676 | A C string constant giving the name of the function to call for | |
3677 | division of one signed full-word by another. If you do not define | |
3678 | this macro, the default name is used, which is @code{__divsi3}, a | |
3679 | function defined in @file{libgcc.a}. | |
3680 | ||
3681 | @findex UDIVSI3_LIBCALL | |
3682 | @item UDIVSI3_LIBCALL | |
3683 | A C string constant giving the name of the function to call for | |
3684 | division of one unsigned full-word by another. If you do not define | |
3685 | this macro, the default name is used, which is @code{__udivsi3}, a | |
3686 | function defined in @file{libgcc.a}. | |
3687 | ||
3688 | @findex MODSI3_LIBCALL | |
3689 | @item MODSI3_LIBCALL | |
3690 | A C string constant giving the name of the function to call for the | |
3691 | remainder in division of one signed full-word by another. If you do | |
3692 | not define this macro, the default name is used, which is | |
3693 | @code{__modsi3}, a function defined in @file{libgcc.a}. | |
3694 | ||
3695 | @findex UMODSI3_LIBCALL | |
3696 | @item UMODSI3_LIBCALL | |
3697 | A C string constant giving the name of the function to call for the | |
3698 | remainder in division of one unsigned full-word by another. If you do | |
3699 | not define this macro, the default name is used, which is | |
3700 | @code{__umodsi3}, a function defined in @file{libgcc.a}. | |
3701 | ||
3702 | @findex MULDI3_LIBCALL | |
3703 | @item MULDI3_LIBCALL | |
3704 | A C string constant giving the name of the function to call for | |
3705 | multiplication of one signed double-word by another. If you do not | |
3706 | define this macro, the default name is used, which is @code{__muldi3}, | |
3707 | a function defined in @file{libgcc.a}. | |
3708 | ||
3709 | @findex DIVDI3_LIBCALL | |
3710 | @item DIVDI3_LIBCALL | |
3711 | A C string constant giving the name of the function to call for | |
3712 | division of one signed double-word by another. If you do not define | |
3713 | this macro, the default name is used, which is @code{__divdi3}, a | |
3714 | function defined in @file{libgcc.a}. | |
3715 | ||
3716 | @findex UDIVDI3_LIBCALL | |
3717 | @item UDIVDI3_LIBCALL | |
3718 | A C string constant giving the name of the function to call for | |
3719 | division of one unsigned full-word by another. If you do not define | |
3720 | this macro, the default name is used, which is @code{__udivdi3}, a | |
3721 | function defined in @file{libgcc.a}. | |
3722 | ||
3723 | @findex MODDI3_LIBCALL | |
3724 | @item MODDI3_LIBCALL | |
3725 | A C string constant giving the name of the function to call for the | |
3726 | remainder in division of one signed double-word by another. If you do | |
3727 | not define this macro, the default name is used, which is | |
3728 | @code{__moddi3}, a function defined in @file{libgcc.a}. | |
3729 | ||
3730 | @findex UMODDI3_LIBCALL | |
3731 | @item UMODDI3_LIBCALL | |
3732 | A C string constant giving the name of the function to call for the | |
3733 | remainder in division of one unsigned full-word by another. If you do | |
3734 | not define this macro, the default name is used, which is | |
3735 | @code{__umoddi3}, a function defined in @file{libgcc.a}. | |
3736 | ||
3737 | @findex INIT_TARGET_OPTABS | |
3738 | @item INIT_TARGET_OPTABS | |
3739 | Define this macro as a C statement that declares additional library | |
3740 | routines renames existing ones. @code{init_optabs} calls this macro after | |
3741 | initializing all the normal library routines. | |
3742 | ||
3743 | @findex TARGET_EDOM | |
3744 | @cindex @code{EDOM}, implicit usage | |
3745 | @item TARGET_EDOM | |
3746 | The value of @code{EDOM} on the target machine, as a C integer constant | |
3747 | expression. If you don't define this macro, GNU CC does not attempt to | |
3748 | deposit the value of @code{EDOM} into @code{errno} directly. Look in | |
3749 | @file{/usr/include/errno.h} to find the value of @code{EDOM} on your | |
3750 | system. | |
3751 | ||
3752 | If you do not define @code{TARGET_EDOM}, then compiled code reports | |
3753 | domain errors by calling the library function and letting it report the | |
3754 | error. If mathematical functions on your system use @code{matherr} when | |
3755 | there is an error, then you should leave @code{TARGET_EDOM} undefined so | |
3756 | that @code{matherr} is used normally. | |
3757 | ||
3758 | @findex GEN_ERRNO_RTX | |
3759 | @cindex @code{errno}, implicit usage | |
3760 | @item GEN_ERRNO_RTX | |
3761 | Define this macro as a C expression to create an rtl expression that | |
3762 | refers to the global ``variable'' @code{errno}. (On certain systems, | |
3763 | @code{errno} may not actually be a variable.) If you don't define this | |
3764 | macro, a reasonable default is used. | |
3765 | ||
3766 | @findex TARGET_MEM_FUNCTIONS | |
3767 | @cindex @code{bcopy}, implicit usage | |
3768 | @cindex @code{memcpy}, implicit usage | |
3769 | @cindex @code{bzero}, implicit usage | |
3770 | @cindex @code{memset}, implicit usage | |
3771 | @item TARGET_MEM_FUNCTIONS | |
3772 | Define this macro if GNU CC should generate calls to the System V | |
3773 | (and ANSI C) library functions @code{memcpy} and @code{memset} | |
3774 | rather than the BSD functions @code{bcopy} and @code{bzero}. | |
3775 | ||
3776 | @findex LIBGCC_NEEDS_DOUBLE | |
3777 | @item LIBGCC_NEEDS_DOUBLE | |
3778 | Define this macro if only @code{float} arguments cannot be passed to | |
3779 | library routines (so they must be converted to @code{double}). This | |
3780 | macro affects both how library calls are generated and how the library | |
3781 | routines in @file{libgcc1.c} accept their arguments. It is useful on | |
3782 | machines where floating and fixed point arguments are passed | |
3783 | differently, such as the i860. | |
3784 | ||
3785 | @findex FLOAT_ARG_TYPE | |
3786 | @item FLOAT_ARG_TYPE | |
3787 | Define this macro to override the type used by the library routines to | |
3788 | pick up arguments of type @code{float}. (By default, they use a union | |
3789 | of @code{float} and @code{int}.) | |
3790 | ||
3791 | The obvious choice would be @code{float}---but that won't work with | |
3792 | traditional C compilers that expect all arguments declared as @code{float} | |
3793 | to arrive as @code{double}. To avoid this conversion, the library routines | |
3794 | ask for the value as some other type and then treat it as a @code{float}. | |
3795 | ||
3796 | On some systems, no other type will work for this. For these systems, | |
3797 | you must use @code{LIBGCC_NEEDS_DOUBLE} instead, to force conversion of | |
3798 | the values @code{double} before they are passed. | |
3799 | ||
3800 | @findex FLOATIFY | |
3801 | @item FLOATIFY (@var{passed-value}) | |
3802 | Define this macro to override the way library routines redesignate a | |
3803 | @code{float} argument as a @code{float} instead of the type it was | |
3804 | passed as. The default is an expression which takes the @code{float} | |
3805 | field of the union. | |
3806 | ||
3807 | @findex FLOAT_VALUE_TYPE | |
3808 | @item FLOAT_VALUE_TYPE | |
3809 | Define this macro to override the type used by the library routines to | |
3810 | return values that ought to have type @code{float}. (By default, they | |
3811 | use @code{int}.) | |
3812 | ||
3813 | The obvious choice would be @code{float}---but that won't work with | |
3814 | traditional C compilers gratuitously convert values declared as | |
3815 | @code{float} into @code{double}. | |
3816 | ||
3817 | @findex INTIFY | |
3818 | @item INTIFY (@var{float-value}) | |
3819 | Define this macro to override the way the value of a | |
3820 | @code{float}-returning library routine should be packaged in order to | |
3821 | return it. These functions are actually declared to return type | |
3822 | @code{FLOAT_VALUE_TYPE} (normally @code{int}). | |
3823 | ||
3824 | These values can't be returned as type @code{float} because traditional | |
3825 | C compilers would gratuitously convert the value to a @code{double}. | |
3826 | ||
3827 | A local variable named @code{intify} is always available when the macro | |
3828 | @code{INTIFY} is used. It is a union of a @code{float} field named | |
3829 | @code{f} and a field named @code{i} whose type is | |
3830 | @code{FLOAT_VALUE_TYPE} or @code{int}. | |
3831 | ||
3832 | If you don't define this macro, the default definition works by copying | |
3833 | the value through that union. | |
3834 | ||
3835 | @findex nongcc_SI_type | |
3836 | @item nongcc_SI_type | |
3837 | Define this macro as the name of the data type corresponding to | |
3838 | @code{SImode} in the system's own C compiler. | |
3839 | ||
3840 | You need not define this macro if that type is @code{long int}, as it usually | |
3841 | is. | |
3842 | ||
3843 | @findex nongcc_word_type | |
3844 | @item nongcc_word_type | |
3845 | Define this macro as the name of the data type corresponding to the | |
3846 | word_mode in the system's own C compiler. | |
3847 | ||
3848 | You need not define this macro if that type is @code{long int}, as it usually | |
3849 | is. | |
3850 | ||
3851 | @findex perform_@dots{} | |
3852 | @item perform_@dots{} | |
3853 | Define these macros to supply explicit C statements to carry out various | |
3854 | arithmetic operations on types @code{float} and @code{double} in the | |
3855 | library routines in @file{libgcc1.c}. See that file for a full list | |
3856 | of these macros and their arguments. | |
3857 | ||
3858 | On most machines, you don't need to define any of these macros, because | |
3859 | the C compiler that comes with the system takes care of doing them. | |
3860 | ||
3861 | @findex NEXT_OBJC_RUNTIME | |
3862 | @item NEXT_OBJC_RUNTIME | |
3863 | Define this macro to generate code for Objective C message sending using | |
3864 | the calling convention of the NeXT system. This calling convention | |
3865 | involves passing the object, the selector and the method arguments all | |
3866 | at once to the method-lookup library function. | |
3867 | ||
3868 | The default calling convention passes just the object and the selector | |
3869 | to the lookup function, which returns a pointer to the method. | |
3870 | @end table | |
3871 | ||
3872 | @node Addressing Modes | |
3873 | @section Addressing Modes | |
3874 | @cindex addressing modes | |
3875 | ||
3876 | @c prevent bad page break with this line | |
3877 | This is about addressing modes. | |
3878 | ||
3879 | @table @code | |
3880 | @findex HAVE_POST_INCREMENT | |
3881 | @item HAVE_POST_INCREMENT | |
3882 | Define this macro if the machine supports post-increment addressing. | |
3883 | ||
3884 | @findex HAVE_PRE_INCREMENT | |
3885 | @findex HAVE_POST_DECREMENT | |
3886 | @findex HAVE_PRE_DECREMENT | |
3887 | @item HAVE_PRE_INCREMENT | |
3888 | @itemx HAVE_POST_DECREMENT | |
3889 | @itemx HAVE_PRE_DECREMENT | |
3890 | Similar for other kinds of addressing. | |
3891 | ||
3892 | @findex CONSTANT_ADDRESS_P | |
3893 | @item CONSTANT_ADDRESS_P (@var{x}) | |
3894 | A C expression that is 1 if the RTX @var{x} is a constant which | |
3895 | is a valid address. On most machines, this can be defined as | |
3896 | @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive | |
3897 | in which constant addresses are supported. | |
3898 | ||
3899 | @findex CONSTANT_P | |
3900 | @code{CONSTANT_P} accepts integer-values expressions whose values are | |
3901 | not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and | |
3902 | @code{high} expressions and @code{const} arithmetic expressions, in | |
3903 | addition to @code{const_int} and @code{const_double} expressions. | |
3904 | ||
3905 | @findex MAX_REGS_PER_ADDRESS | |
3906 | @item MAX_REGS_PER_ADDRESS | |
3907 | A number, the maximum number of registers that can appear in a valid | |
3908 | memory address. Note that it is up to you to specify a value equal to | |
3909 | the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever | |
3910 | accept. | |
3911 | ||
3912 | @findex GO_IF_LEGITIMATE_ADDRESS | |
3913 | @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label}) | |
3914 | A C compound statement with a conditional @code{goto @var{label};} | |
3915 | executed if @var{x} (an RTX) is a legitimate memory address on the | |
3916 | target machine for a memory operand of mode @var{mode}. | |
3917 | ||
3918 | It usually pays to define several simpler macros to serve as | |
3919 | subroutines for this one. Otherwise it may be too complicated to | |
3920 | understand. | |
3921 | ||
3922 | This macro must exist in two variants: a strict variant and a | |
3923 | non-strict one. The strict variant is used in the reload pass. It | |
3924 | must be defined so that any pseudo-register that has not been | |
3925 | allocated a hard register is considered a memory reference. In | |
3926 | contexts where some kind of register is required, a pseudo-register | |
3927 | with no hard register must be rejected. | |
3928 | ||
3929 | The non-strict variant is used in other passes. It must be defined to | |
3930 | accept all pseudo-registers in every context where some kind of | |
3931 | register is required. | |
3932 | ||
3933 | @findex REG_OK_STRICT | |
3934 | Compiler source files that want to use the strict variant of this | |
3935 | macro define the macro @code{REG_OK_STRICT}. You should use an | |
3936 | @code{#ifdef REG_OK_STRICT} conditional to define the strict variant | |
3937 | in that case and the non-strict variant otherwise. | |
3938 | ||
3939 | Subroutines to check for acceptable registers for various purposes (one | |
3940 | for base registers, one for index registers, and so on) are typically | |
3941 | among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}. | |
3942 | Then only these subroutine macros need have two variants; the higher | |
3943 | levels of macros may be the same whether strict or not.@refill | |
3944 | ||
3945 | Normally, constant addresses which are the sum of a @code{symbol_ref} | |
3946 | and an integer are stored inside a @code{const} RTX to mark them as | |
3947 | constant. Therefore, there is no need to recognize such sums | |
3948 | specifically as legitimate addresses. Normally you would simply | |
3949 | recognize any @code{const} as legitimate. | |
3950 | ||
3951 | Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant | |
3952 | sums that are not marked with @code{const}. It assumes that a naked | |
3953 | @code{plus} indicates indexing. If so, then you @emph{must} reject such | |
3954 | naked constant sums as illegitimate addresses, so that none of them will | |
3955 | be given to @code{PRINT_OPERAND_ADDRESS}. | |
3956 | ||
3957 | @cindex @code{ENCODE_SECTION_INFO} and address validation | |
3958 | On some machines, whether a symbolic address is legitimate depends on | |
3959 | the section that the address refers to. On these machines, define the | |
3960 | macro @code{ENCODE_SECTION_INFO} to store the information into the | |
3961 | @code{symbol_ref}, and then check for it here. When you see a | |
3962 | @code{const}, you will have to look inside it to find the | |
3963 | @code{symbol_ref} in order to determine the section. @xref{Assembler | |
3964 | Format}. | |
3965 | ||
3966 | @findex saveable_obstack | |
3967 | The best way to modify the name string is by adding text to the | |
3968 | beginning, with suitable punctuation to prevent any ambiguity. Allocate | |
3969 | the new name in @code{saveable_obstack}. You will have to modify | |
3970 | @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and | |
3971 | output the name accordingly, and define @code{STRIP_NAME_ENCODING} to | |
3972 | access the original name string. | |
3973 | ||
3974 | You can check the information stored here into the @code{symbol_ref} in | |
3975 | the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and | |
3976 | @code{PRINT_OPERAND_ADDRESS}. | |
3977 | ||
3978 | @findex REG_OK_FOR_BASE_P | |
3979 | @item REG_OK_FOR_BASE_P (@var{x}) | |
3980 | A C expression that is nonzero if @var{x} (assumed to be a @code{reg} | |
3981 | RTX) is valid for use as a base register. For hard registers, it | |
3982 | should always accept those which the hardware permits and reject the | |
3983 | others. Whether the macro accepts or rejects pseudo registers must be | |
3984 | controlled by @code{REG_OK_STRICT} as described above. This usually | |
3985 | requires two variant definitions, of which @code{REG_OK_STRICT} | |
3986 | controls the one actually used. | |
3987 | ||
3988 | @findex REG_OK_FOR_INDEX_P | |
3989 | @item REG_OK_FOR_INDEX_P (@var{x}) | |
3990 | A C expression that is nonzero if @var{x} (assumed to be a @code{reg} | |
3991 | RTX) is valid for use as an index register. | |
3992 | ||
3993 | The difference between an index register and a base register is that | |
3994 | the index register may be scaled. If an address involves the sum of | |
3995 | two registers, neither one of them scaled, then either one may be | |
3996 | labeled the ``base'' and the other the ``index''; but whichever | |
3997 | labeling is used must fit the machine's constraints of which registers | |
3998 | may serve in each capacity. The compiler will try both labelings, | |
3999 | looking for one that is valid, and will reload one or both registers | |
4000 | only if neither labeling works. | |
4001 | ||
4002 | @findex LEGITIMIZE_ADDRESS | |
4003 | @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win}) | |
4004 | A C compound statement that attempts to replace @var{x} with a valid | |
4005 | memory address for an operand of mode @var{mode}. @var{win} will be a | |
4006 | C statement label elsewhere in the code; the macro definition may use | |
4007 | ||
4008 | @example | |
4009 | GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win}); | |
4010 | @end example | |
4011 | ||
4012 | @noindent | |
4013 | to avoid further processing if the address has become legitimate. | |
4014 | ||
4015 | @findex break_out_memory_refs | |
4016 | @var{x} will always be the result of a call to @code{break_out_memory_refs}, | |
4017 | and @var{oldx} will be the operand that was given to that function to produce | |
4018 | @var{x}. | |
4019 | ||
4020 | The code generated by this macro should not alter the substructure of | |
4021 | @var{x}. If it transforms @var{x} into a more legitimate form, it | |
4022 | should assign @var{x} (which will always be a C variable) a new value. | |
4023 | ||
4024 | It is not necessary for this macro to come up with a legitimate | |
4025 | address. The compiler has standard ways of doing so in all cases. In | |
4026 | fact, it is safe for this macro to do nothing. But often a | |
4027 | machine-dependent strategy can generate better code. | |
4028 | ||
4029 | @findex GO_IF_MODE_DEPENDENT_ADDRESS | |
4030 | @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label}) | |
4031 | A C statement or compound statement with a conditional @code{goto | |
4032 | @var{label};} executed if memory address @var{x} (an RTX) can have | |
4033 | different meanings depending on the machine mode of the memory | |
4034 | reference it is used for or if the address is valid for some modes | |
4035 | but not others. | |
4036 | ||
4037 | Autoincrement and autodecrement addresses typically have mode-dependent | |
4038 | effects because the amount of the increment or decrement is the size | |
4039 | of the operand being addressed. Some machines have other mode-dependent | |
4040 | addresses. Many RISC machines have no mode-dependent addresses. | |
4041 | ||
4042 | You may assume that @var{addr} is a valid address for the machine. | |
4043 | ||
4044 | @findex LEGITIMATE_CONSTANT_P | |
4045 | @item LEGITIMATE_CONSTANT_P (@var{x}) | |
4046 | A C expression that is nonzero if @var{x} is a legitimate constant for | |
4047 | an immediate operand on the target machine. You can assume that | |
4048 | @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact, | |
4049 | @samp{1} is a suitable definition for this macro on machines where | |
4050 | anything @code{CONSTANT_P} is valid.@refill | |
4051 | @end table | |
4052 | ||
4053 | @node Condition Code | |
4054 | @section Condition Code Status | |
4055 | @cindex condition code status | |
4056 | ||
4057 | @c prevent bad page break with this line | |
4058 | This describes the condition code status. | |
4059 | ||
4060 | @findex cc_status | |
4061 | The file @file{conditions.h} defines a variable @code{cc_status} to | |
4062 | describe how the condition code was computed (in case the interpretation of | |
4063 | the condition code depends on the instruction that it was set by). This | |
4064 | variable contains the RTL expressions on which the condition code is | |
4065 | currently based, and several standard flags. | |
4066 | ||
4067 | Sometimes additional machine-specific flags must be defined in the machine | |
4068 | description header file. It can also add additional machine-specific | |
4069 | information by defining @code{CC_STATUS_MDEP}. | |
4070 | ||
4071 | @table @code | |
4072 | @findex CC_STATUS_MDEP | |
4073 | @item CC_STATUS_MDEP | |
4074 | C code for a data type which is used for declaring the @code{mdep} | |
4075 | component of @code{cc_status}. It defaults to @code{int}. | |
4076 | ||
4077 | This macro is not used on machines that do not use @code{cc0}. | |
4078 | ||
4079 | @findex CC_STATUS_MDEP_INIT | |
4080 | @item CC_STATUS_MDEP_INIT | |
4081 | A C expression to initialize the @code{mdep} field to ``empty''. | |
4082 | The default definition does nothing, since most machines don't use | |
4083 | the field anyway. If you want to use the field, you should probably | |
4084 | define this macro to initialize it. | |
4085 | ||
4086 | This macro is not used on machines that do not use @code{cc0}. | |
4087 | ||
4088 | @findex NOTICE_UPDATE_CC | |
4089 | @item NOTICE_UPDATE_CC (@var{exp}, @var{insn}) | |
4090 | A C compound statement to set the components of @code{cc_status} | |
4091 | appropriately for an insn @var{insn} whose body is @var{exp}. It is | |
4092 | this macro's responsibility to recognize insns that set the condition | |
4093 | code as a byproduct of other activity as well as those that explicitly | |
4094 | set @code{(cc0)}. | |
4095 | ||
4096 | This macro is not used on machines that do not use @code{cc0}. | |
4097 | ||
4098 | If there are insns that do not set the condition code but do alter | |
4099 | other machine registers, this macro must check to see whether they | |
4100 | invalidate the expressions that the condition code is recorded as | |
4101 | reflecting. For example, on the 68000, insns that store in address | |
4102 | registers do not set the condition code, which means that usually | |
4103 | @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such | |
4104 | insns. But suppose that the previous insn set the condition code | |
4105 | based on location @samp{a4@@(102)} and the current insn stores a new | |
4106 | value in @samp{a4}. Although the condition code is not changed by | |
4107 | this, it will no longer be true that it reflects the contents of | |
4108 | @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter | |
4109 | @code{cc_status} in this case to say that nothing is known about the | |
4110 | condition code value. | |
4111 | ||
4112 | The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal | |
4113 | with the results of peephole optimization: insns whose patterns are | |
4114 | @code{parallel} RTXs containing various @code{reg}, @code{mem} or | |
4115 | constants which are just the operands. The RTL structure of these | |
4116 | insns is not sufficient to indicate what the insns actually do. What | |
4117 | @code{NOTICE_UPDATE_CC} should do when it sees one is just to run | |
4118 | @code{CC_STATUS_INIT}. | |
4119 | ||
4120 | A possible definition of @code{NOTICE_UPDATE_CC} is to call a function | |
4121 | that looks at an attribute (@pxref{Insn Attributes}) named, for example, | |
4122 | @samp{cc}. This avoids having detailed information about patterns in | |
4123 | two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}. | |
4124 | ||
4125 | @findex EXTRA_CC_MODES | |
4126 | @item EXTRA_CC_MODES | |
4127 | A list of names to be used for additional modes for condition code | |
4128 | values in registers (@pxref{Jump Patterns}). These names are added | |
4129 | to @code{enum machine_mode} and all have class @code{MODE_CC}. By | |
4130 | convention, they should start with @samp{CC} and end with @samp{mode}. | |
4131 | ||
4132 | You should only define this macro if your machine does not use @code{cc0} | |
4133 | and only if additional modes are required. | |
4134 | ||
4135 | @findex EXTRA_CC_NAMES | |
4136 | @item EXTRA_CC_NAMES | |
4137 | A list of C strings giving the names for the modes listed in | |
4138 | @code{EXTRA_CC_MODES}. For example, the Sparc defines this macro and | |
4139 | @code{EXTRA_CC_MODES} as | |
4140 | ||
4141 | @smallexample | |
4142 | #define EXTRA_CC_MODES CC_NOOVmode, CCFPmode, CCFPEmode | |
4143 | #define EXTRA_CC_NAMES "CC_NOOV", "CCFP", "CCFPE" | |
4144 | @end smallexample | |
4145 | ||
4146 | This macro is not required if @code{EXTRA_CC_MODES} is not defined. | |
4147 | ||
4148 | @findex SELECT_CC_MODE | |
4149 | @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y}) | |
4150 | Returns a mode from class @code{MODE_CC} to be used when comparison | |
4151 | operation code @var{op} is applied to rtx @var{x} and @var{y}. For | |
4152 | example, on the Sparc, @code{SELECT_CC_MODE} is defined as (see | |
4153 | @pxref{Jump Patterns} for a description of the reason for this | |
4154 | definition) | |
4155 | ||
4156 | @smallexample | |
4157 | #define SELECT_CC_MODE(OP,X,Y) \ | |
4158 | (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \ | |
4159 | ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \ | |
4160 | : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \ | |
4161 | || GET_CODE (X) == NEG) \ | |
4162 | ? CC_NOOVmode : CCmode)) | |
4163 | @end smallexample | |
4164 | ||
4165 | You need not define this macro if @code{EXTRA_CC_MODES} is not defined. | |
4166 | ||
4167 | @findex CANONICALIZE_COMPARISON | |
4168 | @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1}) | |
4169 | One some machines not all possible comparisons are defined, but you can | |
4170 | convert an invalid comparison into a valid one. For example, the Alpha | |
4171 | does not have a @code{GT} comparison, but you can use an @code{LT} | |
4172 | comparison instead and swap the order of the operands. | |
4173 | ||
4174 | On such machines, define this macro to be a C statement to do any | |
4175 | required conversions. @var{code} is the initial comparison code | |
4176 | and @var{op0} and @var{op1} are the left and right operands of the | |
4177 | comparison, respectively. You should modify @var{code}, @var{op0}, and | |
4178 | @var{op1} as required. | |
4179 | ||
4180 | GNU CC will not assume that the comparison resulting from this macro is | |
4181 | valid but will see if the resulting insn matches a pattern in the | |
4182 | @file{md} file. | |
4183 | ||
4184 | You need not define this macro if it would never change the comparison | |
4185 | code or operands. | |
4186 | ||
4187 | @findex REVERSIBLE_CC_MODE | |
4188 | @item REVERSIBLE_CC_MODE (@var{mode}) | |
4189 | A C expression whose value is one if it is always safe to reverse a | |
4190 | comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE} | |
4191 | can ever return @var{mode} for a floating-point inequality comparison, | |
4192 | then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero. | |
4193 | ||
4194 | You need not define this macro if it would always returns zero or if the | |
4195 | floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}. | |
4196 | For example, here is the definition used on the Sparc, where floating-point | |
4197 | inequality comparisons are always given @code{CCFPEmode}: | |
4198 | ||
4199 | @smallexample | |
4200 | #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode) | |
4201 | @end smallexample | |
4202 | ||
4203 | @end table | |
4204 | ||
4205 | @node Costs | |
4206 | @section Describing Relative Costs of Operations | |
4207 | @cindex costs of instructions | |
4208 | @cindex relative costs | |
4209 | @cindex speed of instructions | |
4210 | ||
4211 | These macros let you describe the relative speed of various operations | |
4212 | on the target machine. | |
4213 | ||
4214 | @table @code | |
4215 | @findex CONST_COSTS | |
4216 | @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code}) | |
4217 | A part of a C @code{switch} statement that describes the relative costs | |
4218 | of constant RTL expressions. It must contain @code{case} labels for | |
4219 | expression codes @code{const_int}, @code{const}, @code{symbol_ref}, | |
4220 | @code{label_ref} and @code{const_double}. Each case must ultimately | |
4221 | reach a @code{return} statement to return the relative cost of the use | |
4222 | of that kind of constant value in an expression. The cost may depend on | |
4223 | the precise value of the constant, which is available for examination in | |
4224 | @var{x}, and the rtx code of the expression in which it is contained, | |
4225 | found in @var{outer_code}. | |
4226 | ||
4227 | @var{code} is the expression code---redundant, since it can be | |
4228 | obtained with @code{GET_CODE (@var{x})}. | |
4229 | ||
4230 | @findex RTX_COSTS | |
4231 | @findex COSTS_N_INSNS | |
4232 | @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code}) | |
4233 | Like @code{CONST_COSTS} but applies to nonconstant RTL expressions. | |
4234 | This can be used, for example, to indicate how costly a multiply | |
4235 | instruction is. In writing this macro, you can use the construct | |
4236 | @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast | |
4237 | instructions. @var{outer_code} is the code of the expression in which | |
4238 | @var{x} is contained. | |
4239 | ||
4240 | This macro is optional; do not define it if the default cost assumptions | |
4241 | are adequate for the target machine. | |
4242 | ||
4243 | @findex ADDRESS_COST | |
4244 | @item ADDRESS_COST (@var{address}) | |
4245 | An expression giving the cost of an addressing mode that contains | |
4246 | @var{address}. If not defined, the cost is computed from | |
4247 | the @var{address} expression and the @code{CONST_COSTS} values. | |
4248 | ||
4249 | For most CISC machines, the default cost is a good approximation of the | |
4250 | true cost of the addressing mode. However, on RISC machines, all | |
4251 | instructions normally have the same length and execution time. Hence | |
4252 | all addresses will have equal costs. | |
4253 | ||
4254 | In cases where more than one form of an address is known, the form with | |
4255 | the lowest cost will be used. If multiple forms have the same, lowest, | |
4256 | cost, the one that is the most complex will be used. | |
4257 | ||
4258 | For example, suppose an address that is equal to the sum of a register | |
4259 | and a constant is used twice in the same basic block. When this macro | |
4260 | is not defined, the address will be computed in a register and memory | |
4261 | references will be indirect through that register. On machines where | |
4262 | the cost of the addressing mode containing the sum is no higher than | |
4263 | that of a simple indirect reference, this will produce an additional | |
4264 | instruction and possibly require an additional register. Proper | |
4265 | specification of this macro eliminates this overhead for such machines. | |
4266 | ||
4267 | Similar use of this macro is made in strength reduction of loops. | |
4268 | ||
4269 | @var{address} need not be valid as an address. In such a case, the cost | |
4270 | is not relevant and can be any value; invalid addresses need not be | |
4271 | assigned a different cost. | |
4272 | ||
4273 | On machines where an address involving more than one register is as | |
4274 | cheap as an address computation involving only one register, defining | |
4275 | @code{ADDRESS_COST} to reflect this can cause two registers to be live | |
4276 | over a region of code where only one would have been if | |
4277 | @code{ADDRESS_COST} were not defined in that manner. This effect should | |
4278 | be considered in the definition of this macro. Equivalent costs should | |
4279 | probably only be given to addresses with different numbers of registers | |
4280 | on machines with lots of registers. | |
4281 | ||
4282 | This macro will normally either not be defined or be defined as a | |
4283 | constant. | |
4284 | ||
4285 | @findex REGISTER_MOVE_COST | |
4286 | @item REGISTER_MOVE_COST (@var{from}, @var{to}) | |
4287 | A C expression for the cost of moving data from a register in class | |
4288 | @var{from} to one in class @var{to}. The classes are expressed using | |
4289 | the enumeration values such as @code{GENERAL_REGS}. A value of 2 is the | |
4290 | default; other values are interpreted relative to that. | |
4291 | ||
4292 | It is not required that the cost always equal 2 when @var{from} is the | |
4293 | same as @var{to}; on some machines it is expensive to move between | |
4294 | registers if they are not general registers. | |
4295 | ||
4296 | If reload sees an insn consisting of a single @code{set} between two | |
4297 | hard registers, and if @code{REGISTER_MOVE_COST} applied to their | |
4298 | classes returns a value of 2, reload does not check to ensure that the | |
4299 | constraints of the insn are met. Setting a cost of other than 2 will | |
4300 | allow reload to verify that the constraints are met. You should do this | |
4301 | if the @samp{mov@var{m}} pattern's constraints do not allow such copying. | |
4302 | ||
4303 | @findex MEMORY_MOVE_COST | |
4304 | @item MEMORY_MOVE_COST (@var{m}) | |
4305 | A C expression for the cost of moving data of mode @var{m} between a | |
4306 | register and memory. A value of 4 is the default; this cost is relative | |
4307 | to those in @code{REGISTER_MOVE_COST}. | |
4308 | ||
4309 | If moving between registers and memory is more expensive than between | |
4310 | two registers, you should define this macro to express the relative cost. | |
4311 | ||
4312 | @findex BRANCH_COST | |
4313 | @item BRANCH_COST | |
4314 | A C expression for the cost of a branch instruction. A value of 1 is | |
4315 | the default; other values are interpreted relative to that. | |
4316 | @end table | |
4317 | ||
4318 | Here are additional macros which do not specify precise relative costs, | |
4319 | but only that certain actions are more expensive than GNU CC would | |
4320 | ordinarily expect. | |
4321 | ||
4322 | @table @code | |
4323 | @findex SLOW_BYTE_ACCESS | |
4324 | @item SLOW_BYTE_ACCESS | |
4325 | Define this macro as a C expression which is nonzero if accessing less | |
4326 | than a word of memory (i.e. a @code{char} or a @code{short}) is no | |
4327 | faster than accessing a word of memory, i.e., if such access | |
4328 | require more than one instruction or if there is no difference in cost | |
4329 | between byte and (aligned) word loads. | |
4330 | ||
4331 | When this macro is not defined, the compiler will access a field by | |
4332 | finding the smallest containing object; when it is defined, a fullword | |
4333 | load will be used if alignment permits. Unless bytes accesses are | |
4334 | faster than word accesses, using word accesses is preferable since it | |
4335 | may eliminate subsequent memory access if subsequent accesses occur to | |
4336 | other fields in the same word of the structure, but to different bytes. | |
4337 | ||
4338 | @findex SLOW_ZERO_EXTEND | |
4339 | @item SLOW_ZERO_EXTEND | |
4340 | Define this macro if zero-extension (of a @code{char} or @code{short} | |
4341 | to an @code{int}) can be done faster if the destination is a register | |
4342 | that is known to be zero. | |
4343 | ||
4344 | If you define this macro, you must have instruction patterns that | |
4345 | recognize RTL structures like this: | |
4346 | ||
4347 | @smallexample | |
4348 | (set (strict_low_part (subreg:QI (reg:SI @dots{}) 0)) @dots{}) | |
4349 | @end smallexample | |
4350 | ||
4351 | @noindent | |
4352 | and likewise for @code{HImode}. | |
4353 | ||
4354 | @findex SLOW_UNALIGNED_ACCESS | |
4355 | @item SLOW_UNALIGNED_ACCESS | |
4356 | Define this macro to be the value 1 if unaligned accesses have a cost | |
4357 | many times greater than aligned accesses, for example if they are | |
4358 | emulated in a trap handler. | |
4359 | ||
4360 | When this macro is non-zero, the compiler will act as if | |
4361 | @code{STRICT_ALIGNMENT} were non-zero when generating code for block | |
4362 | moves. This can cause significantly more instructions to be produced. | |
4363 | Therefore, do not set this macro non-zero if unaligned accesses only add a | |
4364 | cycle or two to the time for a memory access. | |
4365 | ||
4366 | If the value of this macro is always zero, it need not be defined. | |
4367 | ||
4368 | @findex DONT_REDUCE_ADDR | |
4369 | @item DONT_REDUCE_ADDR | |
4370 | Define this macro to inhibit strength reduction of memory addresses. | |
4371 | (On some machines, such strength reduction seems to do harm rather | |
4372 | than good.) | |
4373 | ||
4374 | @findex MOVE_RATIO | |
4375 | @item MOVE_RATIO | |
4376 | The number of scalar move insns which should be generated instead of a | |
4377 | string move insn or a library call. Increasing the value will always | |
4378 | make code faster, but eventually incurs high cost in increased code size. | |
4379 | ||
4380 | If you don't define this, a reasonable default is used. | |
4381 | ||
4382 | @findex NO_FUNCTION_CSE | |
4383 | @item NO_FUNCTION_CSE | |
4384 | Define this macro if it is as good or better to call a constant | |
4385 | function address than to call an address kept in a register. | |
4386 | ||
4387 | @findex NO_RECURSIVE_FUNCTION_CSE | |
4388 | @item NO_RECURSIVE_FUNCTION_CSE | |
4389 | Define this macro if it is as good or better for a function to call | |
4390 | itself with an explicit address than to call an address kept in a | |
4391 | register. | |
4392 | ||
4393 | @findex ADJUST_COST | |
4394 | @item ADJUST_COST (@var{insn}, @var{link}, @var{dep_insn}, @var{cost}) | |
4395 | A C statement (sans semicolon) to update the integer variable @var{cost} | |
4396 | based on the relationship between @var{insn} that is dependent on | |
4397 | @var{dep_insn} through the dependence @var{link}. The default is to | |
4398 | make no adjustment to @var{cost}. This can be used for example to | |
4399 | specify to the scheduler that an output- or anti-dependence does not | |
4400 | incur the same cost as a data-dependence. | |
4401 | ||
4402 | @findex ADJUST_PRIORITY | |
4403 | @item ADJUST_PRIORITY (@var{insn}) | |
4404 | A C statement (sans semicolon) to update the integer scheduling | |
4405 | priority @code{INSN_PRIORITY(@var{insn})}. Reduce the priority | |
4406 | to execute the @var{insn} earlier, increase the priority to execute | |
4407 | @var{insn} later. Do not define this macro if you do not need to | |
4408 | adjust the scheduling priorities of insns. | |
4409 | @end table | |
4410 | ||
4411 | @node Sections | |
4412 | @section Dividing the Output into Sections (Texts, Data, @dots{}) | |
4413 | @c the above section title is WAY too long. maybe cut the part between | |
4414 | @c the (...)? --mew 10feb93 | |
4415 | ||
4416 | An object file is divided into sections containing different types of | |
4417 | data. In the most common case, there are three sections: the @dfn{text | |
4418 | section}, which holds instructions and read-only data; the @dfn{data | |
4419 | section}, which holds initialized writable data; and the @dfn{bss | |
4420 | section}, which holds uninitialized data. Some systems have other kinds | |
4421 | of sections. | |
4422 | ||
4423 | The compiler must tell the assembler when to switch sections. These | |
4424 | macros control what commands to output to tell the assembler this. You | |
4425 | can also define additional sections. | |
4426 | ||
4427 | @table @code | |
4428 | @findex TEXT_SECTION_ASM_OP | |
4429 | @item TEXT_SECTION_ASM_OP | |
4430 | A C expression whose value is a string containing the assembler | |
4431 | operation that should precede instructions and read-only data. Normally | |
4432 | @code{".text"} is right. | |
4433 | ||
4434 | @findex DATA_SECTION_ASM_OP | |
4435 | @item DATA_SECTION_ASM_OP | |
4436 | A C expression whose value is a string containing the assembler | |
4437 | operation to identify the following data as writable initialized data. | |
4438 | Normally @code{".data"} is right. | |
4439 | ||
4440 | @findex SHARED_SECTION_ASM_OP | |
4441 | @item SHARED_SECTION_ASM_OP | |
4442 | If defined, a C expression whose value is a string containing the | |
4443 | assembler operation to identify the following data as shared data. If | |
4444 | not defined, @code{DATA_SECTION_ASM_OP} will be used. | |
4445 | ||
4446 | @findex BSS_SECTION_ASM_OP | |
4447 | @item BSS_SECTION_ASM_OP | |
4448 | If defined, a C expression whose value is a string containing the | |
4449 | assembler operation to identify the following data as uninitialized global | |
4450 | data. If not defined, and neither @code{ASM_OUTPUT_BSS} nor | |
4451 | @code{ASM_OUTPUT_ALIGNED_BSS} are defined, uninitialized global data will be | |
4452 | output in the data section if @samp{-fno-common} is passed, otherwise | |
4453 | @code{ASM_OUTPUT_COMMON} will be used. | |
4454 | ||
4455 | @findex SHARED_BSS_SECTION_ASM_OP | |
4456 | @item SHARED_BSS_SECTION_ASM_OP | |
4457 | If defined, a C expression whose value is a string containing the | |
4458 | assembler operation to identify the following data as uninitialized global | |
4459 | shared data. If not defined, and @code{BSS_SECTION_ASM_OP} is, the latter | |
4460 | will be used. | |
4461 | ||
4462 | @findex INIT_SECTION_ASM_OP | |
4463 | @item INIT_SECTION_ASM_OP | |
4464 | If defined, a C expression whose value is a string containing the | |
4465 | assembler operation to identify the following data as initialization | |
4466 | code. If not defined, GNU CC will assume such a section does not | |
4467 | exist. | |
4468 | ||
4469 | @findex EXTRA_SECTIONS | |
4470 | @findex in_text | |
4471 | @findex in_data | |
4472 | @item EXTRA_SECTIONS | |
4473 | A list of names for sections other than the standard two, which are | |
4474 | @code{in_text} and @code{in_data}. You need not define this macro | |
4475 | on a system with no other sections (that GCC needs to use). | |
4476 | ||
4477 | @findex EXTRA_SECTION_FUNCTIONS | |
4478 | @findex text_section | |
4479 | @findex data_section | |
4480 | @item EXTRA_SECTION_FUNCTIONS | |
4481 | One or more functions to be defined in @file{varasm.c}. These | |
4482 | functions should do jobs analogous to those of @code{text_section} and | |
4483 | @code{data_section}, for your additional sections. Do not define this | |
4484 | macro if you do not define @code{EXTRA_SECTIONS}. | |
4485 | ||
4486 | @findex READONLY_DATA_SECTION | |
4487 | @item READONLY_DATA_SECTION | |
4488 | On most machines, read-only variables, constants, and jump tables are | |
4489 | placed in the text section. If this is not the case on your machine, | |
4490 | this macro should be defined to be the name of a function (either | |
4491 | @code{data_section} or a function defined in @code{EXTRA_SECTIONS}) that | |
4492 | switches to the section to be used for read-only items. | |
4493 | ||
4494 | If these items should be placed in the text section, this macro should | |
4495 | not be defined. | |
4496 | ||
4497 | @findex SELECT_SECTION | |
4498 | @item SELECT_SECTION (@var{exp}, @var{reloc}) | |
4499 | A C statement or statements to switch to the appropriate section for | |
4500 | output of @var{exp}. You can assume that @var{exp} is either a | |
4501 | @code{VAR_DECL} node or a constant of some sort. @var{reloc} | |
4502 | indicates whether the initial value of @var{exp} requires link-time | |
4503 | relocations. Select the section by calling @code{text_section} or one | |
4504 | of the alternatives for other sections. | |
4505 | ||
4506 | Do not define this macro if you put all read-only variables and | |
4507 | constants in the read-only data section (usually the text section). | |
4508 | ||
4509 | @findex SELECT_RTX_SECTION | |
4510 | @item SELECT_RTX_SECTION (@var{mode}, @var{rtx}) | |
4511 | A C statement or statements to switch to the appropriate section for | |
4512 | output of @var{rtx} in mode @var{mode}. You can assume that @var{rtx} | |
4513 | is some kind of constant in RTL. The argument @var{mode} is redundant | |
4514 | except in the case of a @code{const_int} rtx. Select the section by | |
4515 | calling @code{text_section} or one of the alternatives for other | |
4516 | sections. | |
4517 | ||
4518 | Do not define this macro if you put all constants in the read-only | |
4519 | data section. | |
4520 | ||
4521 | @findex JUMP_TABLES_IN_TEXT_SECTION | |
4522 | @item JUMP_TABLES_IN_TEXT_SECTION | |
4523 | Define this macro if jump tables (for @code{tablejump} insns) should be | |
4524 | output in the text section, along with the assembler instructions. | |
4525 | Otherwise, the readonly data section is used. | |
4526 | ||
4527 | This macro is irrelevant if there is no separate readonly data section. | |
4528 | ||
4529 | @findex ENCODE_SECTION_INFO | |
4530 | @item ENCODE_SECTION_INFO (@var{decl}) | |
4531 | Define this macro if references to a symbol must be treated differently | |
4532 | depending on something about the variable or function named by the | |
4533 | symbol (such as what section it is in). | |
4534 | ||
4535 | The macro definition, if any, is executed immediately after the rtl for | |
4536 | @var{decl} has been created and stored in @code{DECL_RTL (@var{decl})}. | |
4537 | The value of the rtl will be a @code{mem} whose address is a | |
4538 | @code{symbol_ref}. | |
4539 | ||
4540 | @cindex @code{SYMBOL_REF_FLAG}, in @code{ENCODE_SECTION_INFO} | |
4541 | The usual thing for this macro to do is to record a flag in the | |
4542 | @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a | |
4543 | modified name string in the @code{symbol_ref} (if one bit is not enough | |
4544 | information). | |
4545 | ||
4546 | @findex STRIP_NAME_ENCODING | |
4547 | @item STRIP_NAME_ENCODING (@var{var}, @var{sym_name}) | |
4548 | Decode @var{sym_name} and store the real name part in @var{var}, sans | |
4549 | the characters that encode section info. Define this macro if | |
4550 | @code{ENCODE_SECTION_INFO} alters the symbol's name string. | |
4551 | ||
4552 | @findex UNIQUE_SECTION | |
4553 | @item UNIQUE_SECTION (@var{decl}) | |
4554 | For objects going into their own sections, a C expression of the name of the | |
4555 | section, expressed as a STRING_CST node, to put @var{decl} into. The | |
4556 | STRING_CST node must be allocated in the saveable obstack. Function | |
4557 | @code{build_string} can be used to do this. If you do not define this macro, | |
4558 | GNU CC will use the symbol name as the section name. | |
4559 | @end table | |
4560 | ||
4561 | @node PIC | |
4562 | @section Position Independent Code | |
4563 | @cindex position independent code | |
4564 | @cindex PIC | |
4565 | ||
4566 | This section describes macros that help implement generation of position | |
4567 | independent code. Simply defining these macros is not enough to | |
4568 | generate valid PIC; you must also add support to the macros | |
4569 | @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as | |
4570 | well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of | |
4571 | @samp{movsi} to do something appropriate when the source operand | |
4572 | contains a symbolic address. You may also need to alter the handling of | |
4573 | switch statements so that they use relative addresses. | |
4574 | @c i rearranged the order of the macros above to try to force one of | |
4575 | @c them to the next line, to eliminate an overfull hbox. --mew 10feb93 | |
4576 | ||
4577 | @table @code | |
4578 | @findex PIC_OFFSET_TABLE_REGNUM | |
4579 | @item PIC_OFFSET_TABLE_REGNUM | |
4580 | The register number of the register used to address a table of static | |
4581 | data addresses in memory. In some cases this register is defined by a | |
4582 | processor's ``application binary interface'' (ABI). When this macro | |
4583 | is defined, RTL is generated for this register once, as with the stack | |
4584 | pointer and frame pointer registers. If this macro is not defined, it | |
4585 | is up to the machine-dependent files to allocate such a register (if | |
4586 | necessary). | |
4587 | ||
4588 | @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED | |
4589 | @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED | |
4590 | Define this macro if the register defined by | |
4591 | @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define | |
4592 | this macro if @code{PPIC_OFFSET_TABLE_REGNUM} is not defined. | |
4593 | ||
4594 | @findex FINALIZE_PIC | |
4595 | @item FINALIZE_PIC | |
4596 | By generating position-independent code, when two different programs (A | |
4597 | and B) share a common library (libC.a), the text of the library can be | |
4598 | shared whether or not the library is linked at the same address for both | |
4599 | programs. In some of these environments, position-independent code | |
4600 | requires not only the use of different addressing modes, but also | |
4601 | special code to enable the use of these addressing modes. | |
4602 | ||
4603 | The @code{FINALIZE_PIC} macro serves as a hook to emit these special | |
4604 | codes once the function is being compiled into assembly code, but not | |
4605 | before. (It is not done before, because in the case of compiling an | |
4606 | inline function, it would lead to multiple PIC prologues being | |
4607 | included in functions which used inline functions and were compiled to | |
4608 | assembly language.) | |
4609 | ||
4610 | @findex LEGITIMATE_PIC_OPERAND_P | |
4611 | @item LEGITIMATE_PIC_OPERAND_P (@var{x}) | |
4612 | A C expression that is nonzero if @var{x} is a legitimate immediate | |
4613 | operand on the target machine when generating position independent code. | |
4614 | You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not | |
4615 | check this. You can also assume @var{flag_pic} is true, so you need not | |
4616 | check it either. You need not define this macro if all constants | |
4617 | (including @code{SYMBOL_REF}) can be immediate operands when generating | |
4618 | position independent code. | |
4619 | @end table | |
4620 | ||
4621 | @node Assembler Format | |
4622 | @section Defining the Output Assembler Language | |
4623 | ||
4624 | This section describes macros whose principal purpose is to describe how | |
4625 | to write instructions in assembler language--rather than what the | |
4626 | instructions do. | |
4627 | ||
4628 | @menu | |
4629 | * File Framework:: Structural information for the assembler file. | |
4630 | * Data Output:: Output of constants (numbers, strings, addresses). | |
4631 | * Uninitialized Data:: Output of uninitialized variables. | |
4632 | * Label Output:: Output and generation of labels. | |
4633 | * Initialization:: General principles of initialization | |
4634 | and termination routines. | |
4635 | * Macros for Initialization:: | |
4636 | Specific macros that control the handling of | |
4637 | initialization and termination routines. | |
4638 | * Instruction Output:: Output of actual instructions. | |
4639 | * Dispatch Tables:: Output of jump tables. | |
4640 | * Exception Region Output:: Output of exception region code. | |
4641 | * Alignment Output:: Pseudo ops for alignment and skipping data. | |
4642 | @end menu | |
4643 | ||
4644 | @node File Framework | |
4645 | @subsection The Overall Framework of an Assembler File | |
4646 | @cindex assembler format | |
4647 | @cindex output of assembler code | |
4648 | ||
4649 | @c prevent bad page break with this line | |
4650 | This describes the overall framework of an assembler file. | |
4651 | ||
4652 | @table @code | |
4653 | @findex ASM_FILE_START | |
4654 | @item ASM_FILE_START (@var{stream}) | |
4655 | A C expression which outputs to the stdio stream @var{stream} | |
4656 | some appropriate text to go at the start of an assembler file. | |
4657 | ||
4658 | Normally this macro is defined to output a line containing | |
4659 | @samp{#NO_APP}, which is a comment that has no effect on most | |
4660 | assemblers but tells the GNU assembler that it can save time by not | |
4661 | checking for certain assembler constructs. | |
4662 | ||
4663 | On systems that use SDB, it is necessary to output certain commands; | |
4664 | see @file{attasm.h}. | |
4665 | ||
4666 | @findex ASM_FILE_END | |
4667 | @item ASM_FILE_END (@var{stream}) | |
4668 | A C expression which outputs to the stdio stream @var{stream} | |
4669 | some appropriate text to go at the end of an assembler file. | |
4670 | ||
4671 | If this macro is not defined, the default is to output nothing | |
4672 | special at the end of the file. Most systems don't require any | |
4673 | definition. | |
4674 | ||
4675 | On systems that use SDB, it is necessary to output certain commands; | |
4676 | see @file{attasm.h}. | |
4677 | ||
4678 | @findex ASM_IDENTIFY_GCC | |
4679 | @item ASM_IDENTIFY_GCC (@var{file}) | |
4680 | A C statement to output assembler commands which will identify | |
4681 | the object file as having been compiled with GNU CC (or another | |
4682 | GNU compiler). | |
4683 | ||
4684 | If you don't define this macro, the string @samp{gcc_compiled.:} | |
4685 | is output. This string is calculated to define a symbol which, | |
4686 | on BSD systems, will never be defined for any other reason. | |
4687 | GDB checks for the presence of this symbol when reading the | |
4688 | symbol table of an executable. | |
4689 | ||
4690 | On non-BSD systems, you must arrange communication with GDB in | |
4691 | some other fashion. If GDB is not used on your system, you can | |
4692 | define this macro with an empty body. | |
4693 | ||
4694 | @findex ASM_COMMENT_START | |
4695 | @item ASM_COMMENT_START | |
4696 | A C string constant describing how to begin a comment in the target | |
4697 | assembler language. The compiler assumes that the comment will end at | |
4698 | the end of the line. | |
4699 | ||
4700 | @findex ASM_APP_ON | |
4701 | @item ASM_APP_ON | |
4702 | A C string constant for text to be output before each @code{asm} | |
4703 | statement or group of consecutive ones. Normally this is | |
4704 | @code{"#APP"}, which is a comment that has no effect on most | |
4705 | assemblers but tells the GNU assembler that it must check the lines | |
4706 | that follow for all valid assembler constructs. | |
4707 | ||
4708 | @findex ASM_APP_OFF | |
4709 | @item ASM_APP_OFF | |
4710 | A C string constant for text to be output after each @code{asm} | |
4711 | statement or group of consecutive ones. Normally this is | |
4712 | @code{"#NO_APP"}, which tells the GNU assembler to resume making the | |
4713 | time-saving assumptions that are valid for ordinary compiler output. | |
4714 | ||
4715 | @findex ASM_OUTPUT_SOURCE_FILENAME | |
4716 | @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name}) | |
4717 | A C statement to output COFF information or DWARF debugging information | |
4718 | which indicates that filename @var{name} is the current source file to | |
4719 | the stdio stream @var{stream}. | |
4720 | ||
4721 | This macro need not be defined if the standard form of output | |
4722 | for the file format in use is appropriate. | |
4723 | ||
4724 | @findex ASM_OUTPUT_SOURCE_LINE | |
4725 | @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}) | |
4726 | A C statement to output DBX or SDB debugging information before code | |
4727 | for line number @var{line} of the current source file to the | |
4728 | stdio stream @var{stream}. | |
4729 | ||
4730 | This macro need not be defined if the standard form of debugging | |
4731 | information for the debugger in use is appropriate. | |
4732 | ||
4733 | @findex ASM_OUTPUT_IDENT | |
4734 | @item ASM_OUTPUT_IDENT (@var{stream}, @var{string}) | |
4735 | A C statement to output something to the assembler file to handle a | |
4736 | @samp{#ident} directive containing the text @var{string}. If this | |
4737 | macro is not defined, nothing is output for a @samp{#ident} directive. | |
4738 | ||
4739 | @findex ASM_OUTPUT_SECTION_NAME | |
4740 | @item ASM_OUTPUT_SECTION_NAME (@var{stream}, @var{decl}, @var{name}) | |
4741 | A C statement to output something to the assembler file to switch to section | |
4742 | @var{name} for object @var{decl} which is either a @code{FUNCTION_DECL}, a | |
4743 | @code{VAR_DECL} or @code{NULL_TREE}. Some target formats do not support | |
4744 | arbitrary sections. Do not define this macro in such cases. | |
4745 | ||
4746 | At present this macro is only used to support section attributes. | |
4747 | When this macro is undefined, section attributes are disabled. | |
4748 | ||
4749 | @findex OBJC_PROLOGUE | |
4750 | @item OBJC_PROLOGUE | |
4751 | A C statement to output any assembler statements which are required to | |
4752 | precede any Objective C object definitions or message sending. The | |
4753 | statement is executed only when compiling an Objective C program. | |
4754 | @end table | |
4755 | ||
4756 | @need 2000 | |
4757 | @node Data Output | |
4758 | @subsection Output of Data | |
4759 | ||
4760 | @c prevent bad page break with this line | |
4761 | This describes data output. | |
4762 | ||
4763 | @table @code | |
4764 | @findex ASM_OUTPUT_LONG_DOUBLE | |
4765 | @findex ASM_OUTPUT_DOUBLE | |
4766 | @findex ASM_OUTPUT_FLOAT | |
4767 | @item ASM_OUTPUT_LONG_DOUBLE (@var{stream}, @var{value}) | |
4768 | @itemx ASM_OUTPUT_DOUBLE (@var{stream}, @var{value}) | |
4769 | @itemx ASM_OUTPUT_FLOAT (@var{stream}, @var{value}) | |
4770 | @itemx ASM_OUTPUT_THREE_QUARTER_FLOAT (@var{stream}, @var{value}) | |
4771 | @itemx ASM_OUTPUT_SHORT_FLOAT (@var{stream}, @var{value}) | |
4772 | @itemx ASM_OUTPUT_BYTE_FLOAT (@var{stream}, @var{value}) | |
4773 | A C statement to output to the stdio stream @var{stream} an assembler | |
4774 | instruction to assemble a floating-point constant of @code{TFmode}, | |
4775 | @code{DFmode}, @code{SFmode}, @code{TQFmode}, @code{HFmode}, or | |
4776 | @code{QFmode}, respectively, whose value is @var{value}. @var{value} | |
4777 | will be a C expression of type @code{REAL_VALUE_TYPE}. Macros such as | |
4778 | @code{REAL_VALUE_TO_TARGET_DOUBLE} are useful for writing these | |
4779 | definitions. | |
4780 | ||
4781 | @findex ASM_OUTPUT_QUADRUPLE_INT | |
4782 | @findex ASM_OUTPUT_DOUBLE_INT | |
4783 | @findex ASM_OUTPUT_INT | |
4784 | @findex ASM_OUTPUT_SHORT | |
4785 | @findex ASM_OUTPUT_CHAR | |
4786 | @findex output_addr_const | |
4787 | @item ASM_OUTPUT_QUADRUPLE_INT (@var{stream}, @var{exp}) | |
4788 | @itemx ASM_OUTPUT_DOUBLE_INT (@var{stream}, @var{exp}) | |
4789 | @itemx ASM_OUTPUT_INT (@var{stream}, @var{exp}) | |
4790 | @itemx ASM_OUTPUT_SHORT (@var{stream}, @var{exp}) | |
4791 | @itemx ASM_OUTPUT_CHAR (@var{stream}, @var{exp}) | |
4792 | A C statement to output to the stdio stream @var{stream} an assembler | |
4793 | instruction to assemble an integer of 16, 8, 4, 2 or 1 bytes, | |
4794 | respectively, whose value is @var{value}. The argument @var{exp} will | |
4795 | be an RTL expression which represents a constant value. Use | |
4796 | @samp{output_addr_const (@var{stream}, @var{exp})} to output this value | |
4797 | as an assembler expression.@refill | |
4798 | ||
4799 | For sizes larger than @code{UNITS_PER_WORD}, if the action of a macro | |
4800 | would be identical to repeatedly calling the macro corresponding to | |
4801 | a size of @code{UNITS_PER_WORD}, once for each word, you need not define | |
4802 | the macro. | |
4803 | ||
4804 | @findex ASM_OUTPUT_BYTE | |
4805 | @item ASM_OUTPUT_BYTE (@var{stream}, @var{value}) | |
4806 | A C statement to output to the stdio stream @var{stream} an assembler | |
4807 | instruction to assemble a single byte containing the number @var{value}. | |
4808 | ||
4809 | @findex ASM_BYTE_OP | |
4810 | @item ASM_BYTE_OP | |
4811 | A C string constant giving the pseudo-op to use for a sequence of | |
4812 | single-byte constants. If this macro is not defined, the default is | |
4813 | @code{"byte"}. | |
4814 | ||
4815 | @findex ASM_OUTPUT_ASCII | |
4816 | @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len}) | |
4817 | A C statement to output to the stdio stream @var{stream} an assembler | |
4818 | instruction to assemble a string constant containing the @var{len} | |
4819 | bytes at @var{ptr}. @var{ptr} will be a C expression of type | |
4820 | @code{char *} and @var{len} a C expression of type @code{int}. | |
4821 | ||
4822 | If the assembler has a @code{.ascii} pseudo-op as found in the | |
4823 | Berkeley Unix assembler, do not define the macro | |
4824 | @code{ASM_OUTPUT_ASCII}. | |
4825 | ||
4826 | @findex ASM_OUTPUT_POOL_PROLOGUE | |
4827 | @item ASM_OUTPUT_POOL_PROLOGUE (@var{file} @var{funname} @var{fundecl} @var{size}) | |
4828 | A C statement to output assembler commands to define the start of the | |
4829 | constant pool for a function. @var{funname} is a string giving | |
4830 | the name of the function. Should the return type of the function | |
4831 | be required, it can be obtained via @var{fundecl}. @var{size} | |
4832 | is the size, in bytes, of the constant pool that will be written | |
4833 | immediately after this call. | |
4834 | ||
4835 | If no constant-pool prefix is required, the usual case, this macro need | |
4836 | not be defined. | |
4837 | ||
4838 | @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY | |
4839 | @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto}) | |
4840 | A C statement (with or without semicolon) to output a constant in the | |
4841 | constant pool, if it needs special treatment. (This macro need not do | |
4842 | anything for RTL expressions that can be output normally.) | |
4843 | ||
4844 | The argument @var{file} is the standard I/O stream to output the | |
4845 | assembler code on. @var{x} is the RTL expression for the constant to | |
4846 | output, and @var{mode} is the machine mode (in case @var{x} is a | |
4847 | @samp{const_int}). @var{align} is the required alignment for the value | |
4848 | @var{x}; you should output an assembler directive to force this much | |
4849 | alignment. | |
4850 | ||
4851 | The argument @var{labelno} is a number to use in an internal label for | |
4852 | the address of this pool entry. The definition of this macro is | |
4853 | responsible for outputting the label definition at the proper place. | |
4854 | Here is how to do this: | |
4855 | ||
4856 | @example | |
4857 | ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno}); | |
4858 | @end example | |
4859 | ||
4860 | When you output a pool entry specially, you should end with a | |
4861 | @code{goto} to the label @var{jumpto}. This will prevent the same pool | |
4862 | entry from being output a second time in the usual manner. | |
4863 | ||
4864 | You need not define this macro if it would do nothing. | |
4865 | ||
4866 | @findex IS_ASM_LOGICAL_LINE_SEPARATOR | |
4867 | @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}) | |
4868 | Define this macro as a C expression which is nonzero if @var{C} is | |
4869 | used as a logical line separator by the assembler. | |
4870 | ||
4871 | If you do not define this macro, the default is that only | |
4872 | the character @samp{;} is treated as a logical line separator. | |
4873 | ||
4874 | ||
4875 | @findex ASM_OPEN_PAREN | |
4876 | @findex ASM_CLOSE_PAREN | |
4877 | @item ASM_OPEN_PAREN | |
4878 | @itemx ASM_CLOSE_PAREN | |
4879 | These macros are defined as C string constant, describing the syntax | |
4880 | in the assembler for grouping arithmetic expressions. The following | |
4881 | definitions are correct for most assemblers: | |
4882 | ||
4883 | @example | |
4884 | #define ASM_OPEN_PAREN "(" | |
4885 | #define ASM_CLOSE_PAREN ")" | |
4886 | @end example | |
4887 | @end table | |
4888 | ||
4889 | These macros are provided by @file{real.h} for writing the definitions | |
4890 | of @code{ASM_OUTPUT_DOUBLE} and the like: | |
4891 | ||
4892 | @table @code | |
4893 | @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l}) | |
4894 | @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l}) | |
4895 | @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l}) | |
4896 | @findex REAL_VALUE_TO_TARGET_SINGLE | |
4897 | @findex REAL_VALUE_TO_TARGET_DOUBLE | |
4898 | @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE | |
4899 | These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's | |
4900 | floating point representation, and store its bit pattern in the array of | |
4901 | @code{long int} whose address is @var{l}. The number of elements in the | |
4902 | output array is determined by the size of the desired target floating | |
4903 | point data type: 32 bits of it go in each @code{long int} array | |
4904 | element. Each array element holds 32 bits of the result, even if | |
4905 | @code{long int} is wider than 32 bits on the host machine. | |
4906 | ||
4907 | The array element values are designed so that you can print them out | |
4908 | using @code{fprintf} in the order they should appear in the target | |
4909 | machine's memory. | |
4910 | ||
4911 | @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{format}, @var{string}) | |
4912 | @findex REAL_VALUE_TO_DECIMAL | |
4913 | This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a | |
4914 | decimal number and stores it as a string into @var{string}. | |
4915 | You must pass, as @var{string}, the address of a long enough block | |
4916 | of space to hold the result. | |
4917 | ||
4918 | The argument @var{format} is a @code{printf}-specification that serves | |
4919 | as a suggestion for how to format the output string. | |
4920 | @end table | |
4921 | ||
4922 | @node Uninitialized Data | |
4923 | @subsection Output of Uninitialized Variables | |
4924 | ||
4925 | Each of the macros in this section is used to do the whole job of | |
4926 | outputting a single uninitialized variable. | |
4927 | ||
4928 | @table @code | |
4929 | @findex ASM_OUTPUT_COMMON | |
4930 | @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
4931 | A C statement (sans semicolon) to output to the stdio stream | |
4932 | @var{stream} the assembler definition of a common-label named | |
4933 | @var{name} whose size is @var{size} bytes. The variable @var{rounded} | |
4934 | is the size rounded up to whatever alignment the caller wants. | |
4935 | ||
4936 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
4937 | output the name itself; before and after that, output the additional | |
4938 | assembler syntax for defining the name, and a newline. | |
4939 | ||
4940 | This macro controls how the assembler definitions of uninitialized | |
4941 | common global variables are output. | |
4942 | ||
4943 | @findex ASM_OUTPUT_ALIGNED_COMMON | |
4944 | @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment}) | |
4945 | Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a | |
4946 | separate, explicit argument. If you define this macro, it is used in | |
4947 | place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in | |
4948 | handling the required alignment of the variable. The alignment is specified | |
4949 | as the number of bits. | |
4950 | ||
4951 | @findex ASM_OUTPUT_SHARED_COMMON | |
4952 | @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
4953 | If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it | |
4954 | is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON} | |
4955 | will be used. | |
4956 | ||
4957 | @findex ASM_OUTPUT_BSS | |
4958 | @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded}) | |
4959 | A C statement (sans semicolon) to output to the stdio stream | |
4960 | @var{stream} the assembler definition of uninitialized global @var{decl} named | |
4961 | @var{name} whose size is @var{size} bytes. The variable @var{rounded} | |
4962 | is the size rounded up to whatever alignment the caller wants. | |
4963 | ||
4964 | Try to use function @code{asm_output_bss} defined in @file{varasm.c} when | |
4965 | defining this macro. If unable, use the expression | |
4966 | @code{assemble_name (@var{stream}, @var{name})} to output the name itself; | |
4967 | before and after that, output the additional assembler syntax for defining | |
4968 | the name, and a newline. | |
4969 | ||
4970 | This macro controls how the assembler definitions of uninitialized global | |
4971 | variables are output. This macro exists to properly support languages like | |
4972 | @code{c++} which do not have @code{common} data. However, this macro currently | |
4973 | is not defined for all targets. If this macro and | |
4974 | @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON} | |
4975 | or @code{ASM_OUTPUT_ALIGNED_COMMON} is used. | |
4976 | ||
4977 | @findex ASM_OUTPUT_ALIGNED_BSS | |
4978 | @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) | |
4979 | Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a | |
4980 | separate, explicit argument. If you define this macro, it is used in | |
4981 | place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in | |
4982 | handling the required alignment of the variable. The alignment is specified | |
4983 | as the number of bits. | |
4984 | ||
4985 | Try to use function @code{asm_output_aligned_bss} defined in file | |
4986 | @file{varasm.c} when defining this macro. | |
4987 | ||
4988 | @findex ASM_OUTPUT_SHARED_BSS | |
4989 | @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded}) | |
4990 | If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it | |
4991 | is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS} | |
4992 | will be used. | |
4993 | ||
4994 | @findex ASM_OUTPUT_LOCAL | |
4995 | @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
4996 | A C statement (sans semicolon) to output to the stdio stream | |
4997 | @var{stream} the assembler definition of a local-common-label named | |
4998 | @var{name} whose size is @var{size} bytes. The variable @var{rounded} | |
4999 | is the size rounded up to whatever alignment the caller wants. | |
5000 | ||
5001 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
5002 | output the name itself; before and after that, output the additional | |
5003 | assembler syntax for defining the name, and a newline. | |
5004 | ||
5005 | This macro controls how the assembler definitions of uninitialized | |
5006 | static variables are output. | |
5007 | ||
5008 | @findex ASM_OUTPUT_ALIGNED_LOCAL | |
5009 | @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment}) | |
5010 | Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a | |
5011 | separate, explicit argument. If you define this macro, it is used in | |
5012 | place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in | |
5013 | handling the required alignment of the variable. The alignment is specified | |
5014 | as the number of bits. | |
5015 | ||
5016 | @findex ASM_OUTPUT_SHARED_LOCAL | |
5017 | @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
5018 | If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it | |
5019 | is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL} | |
5020 | will be used. | |
5021 | @end table | |
5022 | ||
5023 | @node Label Output | |
5024 | @subsection Output and Generation of Labels | |
5025 | ||
5026 | @c prevent bad page break with this line | |
5027 | This is about outputting labels. | |
5028 | ||
5029 | @table @code | |
5030 | @findex ASM_OUTPUT_LABEL | |
5031 | @findex assemble_name | |
5032 | @item ASM_OUTPUT_LABEL (@var{stream}, @var{name}) | |
5033 | A C statement (sans semicolon) to output to the stdio stream | |
5034 | @var{stream} the assembler definition of a label named @var{name}. | |
5035 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
5036 | output the name itself; before and after that, output the additional | |
5037 | assembler syntax for defining the name, and a newline. | |
5038 | ||
5039 | @findex ASM_DECLARE_FUNCTION_NAME | |
5040 | @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl}) | |
5041 | A C statement (sans semicolon) to output to the stdio stream | |
5042 | @var{stream} any text necessary for declaring the name @var{name} of a | |
5043 | function which is being defined. This macro is responsible for | |
5044 | outputting the label definition (perhaps using | |
5045 | @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the | |
5046 | @code{FUNCTION_DECL} tree node representing the function. | |
5047 | ||
5048 | If this macro is not defined, then the function name is defined in the | |
5049 | usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). | |
5050 | ||
5051 | @findex ASM_DECLARE_FUNCTION_SIZE | |
5052 | @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl}) | |
5053 | A C statement (sans semicolon) to output to the stdio stream | |
5054 | @var{stream} any text necessary for declaring the size of a function | |
5055 | which is being defined. The argument @var{name} is the name of the | |
5056 | function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node | |
5057 | representing the function. | |
5058 | ||
5059 | If this macro is not defined, then the function size is not defined. | |
5060 | ||
5061 | @findex ASM_DECLARE_OBJECT_NAME | |
5062 | @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl}) | |
5063 | A C statement (sans semicolon) to output to the stdio stream | |
5064 | @var{stream} any text necessary for declaring the name @var{name} of an | |
5065 | initialized variable which is being defined. This macro must output the | |
5066 | label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument | |
5067 | @var{decl} is the @code{VAR_DECL} tree node representing the variable. | |
5068 | ||
5069 | If this macro is not defined, then the variable name is defined in the | |
5070 | usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). | |
5071 | ||
5072 | @findex ASM_FINISH_DECLARE_OBJECT | |
5073 | @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend}) | |
5074 | A C statement (sans semicolon) to finish up declaring a variable name | |
5075 | once the compiler has processed its initializer fully and thus has had a | |
5076 | chance to determine the size of an array when controlled by an | |
5077 | initializer. This is used on systems where it's necessary to declare | |
5078 | something about the size of the object. | |
5079 | ||
5080 | If you don't define this macro, that is equivalent to defining it to do | |
5081 | nothing. | |
5082 | ||
5083 | @findex ASM_GLOBALIZE_LABEL | |
5084 | @item ASM_GLOBALIZE_LABEL (@var{stream}, @var{name}) | |
5085 | A C statement (sans semicolon) to output to the stdio stream | |
5086 | @var{stream} some commands that will make the label @var{name} global; | |
5087 | that is, available for reference from other files. Use the expression | |
5088 | @code{assemble_name (@var{stream}, @var{name})} to output the name | |
5089 | itself; before and after that, output the additional assembler syntax | |
5090 | for making that name global, and a newline. | |
5091 | ||
5092 | @findex ASM_WEAKEN_LABEL | |
5093 | @item ASM_WEAKEN_LABEL | |
5094 | A C statement (sans semicolon) to output to the stdio stream | |
5095 | @var{stream} some commands that will make the label @var{name} weak; | |
5096 | that is, available for reference from other files but only used if | |
5097 | no other definition is available. Use the expression | |
5098 | @code{assemble_name (@var{stream}, @var{name})} to output the name | |
5099 | itself; before and after that, output the additional assembler syntax | |
5100 | for making that name weak, and a newline. | |
5101 | ||
5102 | If you don't define this macro, GNU CC will not support weak | |
5103 | symbols and you should not define the @code{SUPPORTS_WEAK} macro. | |
5104 | ||
5105 | @findex SUPPORTS_WEAK | |
5106 | @item SUPPORTS_WEAK | |
5107 | A C expression which evaluates to true if the target supports weak symbols. | |
5108 | ||
5109 | If you don't define this macro, @file{defaults.h} provides a default | |
5110 | definition. If @code{ASM_WEAKEN_LABEL} is defined, the default | |
5111 | definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if | |
5112 | you want to control weak symbol support with a compiler flag such as | |
5113 | @samp{-melf}. | |
5114 | ||
5115 | @findex MAKE_DECL_ONE_ONLY (@var{decl}) | |
5116 | @item MAKE_DECL_ONE_ONLY | |
5117 | A C statement (sans semicolon) to mark @var{decl} to be emitted as a | |
5118 | public symbol such that extra copies in multiple translation units will | |
5119 | be discarded by the linker. Define this macro if your object file | |
5120 | format provides support for this concept, such as the @samp{COMDAT} | |
5121 | section flags in the Microsoft Windows PE/COFF format, and this support | |
5122 | requires changes to @var{decl}, such as putting it in a separate section. | |
5123 | ||
5124 | @findex SUPPORTS_ONE_ONLY | |
5125 | @item SUPPORTS_WEAK | |
5126 | A C expression which evaluates to true if the target supports one-only | |
5127 | semantics. | |
5128 | ||
5129 | If you don't define this macro, @file{varasm.c} provides a default | |
5130 | definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default | |
5131 | definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if | |
5132 | you want to control weak symbol support with a compiler flag, or if | |
5133 | setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to | |
5134 | be emitted as one-only. | |
5135 | ||
5136 | @findex ASM_OUTPUT_EXTERNAL | |
5137 | @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name}) | |
5138 | A C statement (sans semicolon) to output to the stdio stream | |
5139 | @var{stream} any text necessary for declaring the name of an external | |
5140 | symbol named @var{name} which is referenced in this compilation but | |
5141 | not defined. The value of @var{decl} is the tree node for the | |
5142 | declaration. | |
5143 | ||
5144 | This macro need not be defined if it does not need to output anything. | |
5145 | The GNU assembler and most Unix assemblers don't require anything. | |
5146 | ||
5147 | @findex ASM_OUTPUT_EXTERNAL_LIBCALL | |
5148 | @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref}) | |
5149 | A C statement (sans semicolon) to output on @var{stream} an assembler | |
5150 | pseudo-op to declare a library function name external. The name of the | |
5151 | library function is given by @var{symref}, which has type @code{rtx} and | |
5152 | is a @code{symbol_ref}. | |
5153 | ||
5154 | This macro need not be defined if it does not need to output anything. | |
5155 | The GNU assembler and most Unix assemblers don't require anything. | |
5156 | ||
5157 | @findex ASM_OUTPUT_LABELREF | |
5158 | @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name}) | |
5159 | A C statement (sans semicolon) to output to the stdio stream | |
5160 | @var{stream} a reference in assembler syntax to a label named | |
5161 | @var{name}. This should add @samp{_} to the front of the name, if that | |
5162 | is customary on your operating system, as it is in most Berkeley Unix | |
5163 | systems. This macro is used in @code{assemble_name}. | |
5164 | ||
5165 | @ignore @c Seems not to exist anymore. | |
5166 | @findex ASM_OUTPUT_LABELREF_AS_INT | |
5167 | @item ASM_OUTPUT_LABELREF_AS_INT (@var{file}, @var{label}) | |
5168 | Define this macro for systems that use the program @code{collect2}. | |
5169 | The definition should be a C statement to output a word containing | |
5170 | a reference to the label @var{label}. | |
5171 | @end ignore | |
5172 | ||
5173 | @findex ASM_OUTPUT_INTERNAL_LABEL | |
5174 | @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num}) | |
5175 | A C statement to output to the stdio stream @var{stream} a label whose | |
5176 | name is made from the string @var{prefix} and the number @var{num}. | |
5177 | ||
5178 | It is absolutely essential that these labels be distinct from the labels | |
5179 | used for user-level functions and variables. Otherwise, certain programs | |
5180 | will have name conflicts with internal labels. | |
5181 | ||
5182 | It is desirable to exclude internal labels from the symbol table of the | |
5183 | object file. Most assemblers have a naming convention for labels that | |
5184 | should be excluded; on many systems, the letter @samp{L} at the | |
5185 | beginning of a label has this effect. You should find out what | |
5186 | convention your system uses, and follow it. | |
5187 | ||
5188 | The usual definition of this macro is as follows: | |
5189 | ||
5190 | @example | |
5191 | fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num}) | |
5192 | @end example | |
5193 | ||
5194 | @findex ASM_GENERATE_INTERNAL_LABEL | |
5195 | @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num}) | |
5196 | A C statement to store into the string @var{string} a label whose name | |
5197 | is made from the string @var{prefix} and the number @var{num}. | |
5198 | ||
5199 | This string, when output subsequently by @code{assemble_name}, should | |
5200 | produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce | |
5201 | with the same @var{prefix} and @var{num}. | |
5202 | ||
5203 | If the string begins with @samp{*}, then @code{assemble_name} will | |
5204 | output the rest of the string unchanged. It is often convenient for | |
5205 | @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the | |
5206 | string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets | |
5207 | to output the string, and may change it. (Of course, | |
5208 | @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so | |
5209 | you should know what it does on your machine.) | |
5210 | ||
5211 | @findex ASM_FORMAT_PRIVATE_NAME | |
5212 | @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number}) | |
5213 | A C expression to assign to @var{outvar} (which is a variable of type | |
5214 | @code{char *}) a newly allocated string made from the string | |
5215 | @var{name} and the number @var{number}, with some suitable punctuation | |
5216 | added. Use @code{alloca} to get space for the string. | |
5217 | ||
5218 | The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to | |
5219 | produce an assembler label for an internal static variable whose name is | |
5220 | @var{name}. Therefore, the string must be such as to result in valid | |
5221 | assembler code. The argument @var{number} is different each time this | |
5222 | macro is executed; it prevents conflicts between similarly-named | |
5223 | internal static variables in different scopes. | |
5224 | ||
5225 | Ideally this string should not be a valid C identifier, to prevent any | |
5226 | conflict with the user's own symbols. Most assemblers allow periods | |
5227 | or percent signs in assembler symbols; putting at least one of these | |
5228 | between the name and the number will suffice. | |
5229 | ||
5230 | @findex ASM_OUTPUT_DEF | |
5231 | @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value}) | |
5232 | A C statement to output to the stdio stream @var{stream} assembler code | |
5233 | which defines (equates) the symbol @var{name} to have the value @var{value}. | |
5234 | ||
5235 | If SET_ASM_OP is defined, a default definition is provided which is | |
5236 | correct for most systems. | |
5237 | @findex OBJC_GEN_METHOD_LABEL | |
5238 | @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name}) | |
5239 | Define this macro to override the default assembler names used for | |
5240 | Objective C methods. | |
5241 | ||
5242 | The default name is a unique method number followed by the name of the | |
5243 | class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of | |
5244 | the category is also included in the assembler name (e.g.@: | |
5245 | @samp{_1_Foo_Bar}). | |
5246 | ||
5247 | These names are safe on most systems, but make debugging difficult since | |
5248 | the method's selector is not present in the name. Therefore, particular | |
5249 | systems define other ways of computing names. | |
5250 | ||
5251 | @var{buf} is an expression of type @code{char *} which gives you a | |
5252 | buffer in which to store the name; its length is as long as | |
5253 | @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus | |
5254 | 50 characters extra. | |
5255 | ||
5256 | The argument @var{is_inst} specifies whether the method is an instance | |
5257 | method or a class method; @var{class_name} is the name of the class; | |
5258 | @var{cat_name} is the name of the category (or NULL if the method is not | |
5259 | in a category); and @var{sel_name} is the name of the selector. | |
5260 | ||
5261 | On systems where the assembler can handle quoted names, you can use this | |
5262 | macro to provide more human-readable names. | |
5263 | @end table | |
5264 | ||
5265 | @node Initialization | |
5266 | @subsection How Initialization Functions Are Handled | |
5267 | @cindex initialization routines | |
5268 | @cindex termination routines | |
5269 | @cindex constructors, output of | |
5270 | @cindex destructors, output of | |
5271 | ||
5272 | The compiled code for certain languages includes @dfn{constructors} | |
5273 | (also called @dfn{initialization routines})---functions to initialize | |
5274 | data in the program when the program is started. These functions need | |
5275 | to be called before the program is ``started''---that is to say, before | |
5276 | @code{main} is called. | |
5277 | ||
5278 | Compiling some languages generates @dfn{destructors} (also called | |
5279 | @dfn{termination routines}) that should be called when the program | |
5280 | terminates. | |
5281 | ||
5282 | To make the initialization and termination functions work, the compiler | |
5283 | must output something in the assembler code to cause those functions to | |
5284 | be called at the appropriate time. When you port the compiler to a new | |
5285 | system, you need to specify how to do this. | |
5286 | ||
5287 | There are two major ways that GCC currently supports the execution of | |
5288 | initialization and termination functions. Each way has two variants. | |
5289 | Much of the structure is common to all four variations. | |
5290 | ||
5291 | @findex __CTOR_LIST__ | |
5292 | @findex __DTOR_LIST__ | |
5293 | The linker must build two lists of these functions---a list of | |
5294 | initialization functions, called @code{__CTOR_LIST__}, and a list of | |
5295 | termination functions, called @code{__DTOR_LIST__}. | |
5296 | ||
5297 | Each list always begins with an ignored function pointer (which may hold | |
5298 | 0, @minus{}1, or a count of the function pointers after it, depending on | |
5299 | the environment). This is followed by a series of zero or more function | |
5300 | pointers to constructors (or destructors), followed by a function | |
5301 | pointer containing zero. | |
5302 | ||
5303 | Depending on the operating system and its executable file format, either | |
5304 | @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup | |
5305 | time and exit time. Constructors are called in reverse order of the | |
5306 | list; destructors in forward order. | |
5307 | ||
5308 | The best way to handle static constructors works only for object file | |
5309 | formats which provide arbitrarily-named sections. A section is set | |
5310 | aside for a list of constructors, and another for a list of destructors. | |
5311 | Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each | |
5312 | object file that defines an initialization function also puts a word in | |
5313 | the constructor section to point to that function. The linker | |
5314 | accumulates all these words into one contiguous @samp{.ctors} section. | |
5315 | Termination functions are handled similarly. | |
5316 | ||
5317 | To use this method, you need appropriate definitions of the macros | |
5318 | @code{ASM_OUTPUT_CONSTRUCTOR} and @code{ASM_OUTPUT_DESTRUCTOR}. Usually | |
5319 | you can get them by including @file{svr4.h}. | |
5320 | ||
5321 | When arbitrary sections are available, there are two variants, depending | |
5322 | upon how the code in @file{crtstuff.c} is called. On systems that | |
5323 | support an @dfn{init} section which is executed at program startup, | |
5324 | parts of @file{crtstuff.c} are compiled into that section. The | |
5325 | program is linked by the @code{gcc} driver like this: | |
5326 | ||
5327 | @example | |
5328 | ld -o @var{output_file} crtbegin.o @dots{} crtend.o -lgcc | |
5329 | @end example | |
5330 | ||
5331 | The head of a function (@code{__do_global_ctors}) appears in the init | |
5332 | section of @file{crtbegin.o}; the remainder of the function appears in | |
5333 | the init section of @file{crtend.o}. The linker will pull these two | |
5334 | parts of the section together, making a whole function. If any of the | |
5335 | user's object files linked into the middle of it contribute code, then that | |
5336 | code will be executed as part of the body of @code{__do_global_ctors}. | |
5337 | ||
5338 | To use this variant, you must define the @code{INIT_SECTION_ASM_OP} | |
5339 | macro properly. | |
5340 | ||
5341 | If no init section is available, do not define | |
5342 | @code{INIT_SECTION_ASM_OP}. Then @code{__do_global_ctors} is built into | |
5343 | the text section like all other functions, and resides in | |
5344 | @file{libgcc.a}. When GCC compiles any function called @code{main}, it | |
5345 | inserts a procedure call to @code{__main} as the first executable code | |
5346 | after the function prologue. The @code{__main} function, also defined | |
5347 | in @file{libgcc2.c}, simply calls @file{__do_global_ctors}. | |
5348 | ||
5349 | In file formats that don't support arbitrary sections, there are again | |
5350 | two variants. In the simplest variant, the GNU linker (GNU @code{ld}) | |
5351 | and an `a.out' format must be used. In this case, | |
5352 | @code{ASM_OUTPUT_CONSTRUCTOR} is defined to produce a @code{.stabs} | |
5353 | entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__}, | |
5354 | and with the address of the void function containing the initialization | |
5355 | code as its value. The GNU linker recognizes this as a request to add | |
5356 | the value to a ``set''; the values are accumulated, and are eventually | |
5357 | placed in the executable as a vector in the format described above, with | |
5358 | a leading (ignored) count and a trailing zero element. | |
5359 | @code{ASM_OUTPUT_DESTRUCTOR} is handled similarly. Since no init | |
5360 | section is available, the absence of @code{INIT_SECTION_ASM_OP} causes | |
5361 | the compilation of @code{main} to call @code{__main} as above, starting | |
5362 | the initialization process. | |
5363 | ||
5364 | The last variant uses neither arbitrary sections nor the GNU linker. | |
5365 | This is preferable when you want to do dynamic linking and when using | |
5366 | file formats which the GNU linker does not support, such as `ECOFF'. In | |
5367 | this case, @code{ASM_OUTPUT_CONSTRUCTOR} does not produce an | |
5368 | @code{N_SETT} symbol; initialization and termination functions are | |
5369 | recognized simply by their names. This requires an extra program in the | |
5370 | linkage step, called @code{collect2}. This program pretends to be the | |
5371 | linker, for use with GNU CC; it does its job by running the ordinary | |
5372 | linker, but also arranges to include the vectors of initialization and | |
5373 | termination functions. These functions are called via @code{__main} as | |
5374 | described above. | |
5375 | ||
5376 | Choosing among these configuration options has been simplified by a set | |
5377 | of operating-system-dependent files in the @file{config} subdirectory. | |
5378 | These files define all of the relevant parameters. Usually it is | |
5379 | sufficient to include one into your specific machine-dependent | |
5380 | configuration file. These files are: | |
5381 | ||
5382 | @table @file | |
5383 | @item aoutos.h | |
5384 | For operating systems using the `a.out' format. | |
5385 | ||
5386 | @item next.h | |
5387 | For operating systems using the `MachO' format. | |
5388 | ||
5389 | @item svr3.h | |
5390 | For System V Release 3 and similar systems using `COFF' format. | |
5391 | ||
5392 | @item svr4.h | |
5393 | For System V Release 4 and similar systems using `ELF' format. | |
5394 | ||
5395 | @item vms.h | |
5396 | For the VMS operating system. | |
5397 | @end table | |
5398 | ||
5399 | @ifinfo | |
5400 | The following section describes the specific macros that control and | |
5401 | customize the handling of initialization and termination functions. | |
5402 | @end ifinfo | |
5403 | ||
5404 | @node Macros for Initialization | |
5405 | @subsection Macros Controlling Initialization Routines | |
5406 | ||
5407 | Here are the macros that control how the compiler handles initialization | |
5408 | and termination functions: | |
5409 | ||
5410 | @table @code | |
5411 | @findex INIT_SECTION_ASM_OP | |
5412 | @item INIT_SECTION_ASM_OP | |
5413 | If defined, a C string constant for the assembler operation to identify | |
5414 | the following data as initialization code. If not defined, GNU CC will | |
5415 | assume such a section does not exist. When you are using special | |
5416 | sections for initialization and termination functions, this macro also | |
5417 | controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to run the | |
5418 | initialization functions. | |
5419 | ||
5420 | @item HAS_INIT_SECTION | |
5421 | @findex HAS_INIT_SECTION | |
5422 | If defined, @code{main} will not call @code{__main} as described above. | |
5423 | This macro should be defined for systems that control the contents of the | |
5424 | init section on a symbol-by-symbol basis, such as OSF/1, and should not | |
5425 | be defined explicitly for systems that support | |
5426 | @code{INIT_SECTION_ASM_OP}. | |
5427 | ||
5428 | @item LD_INIT_SWITCH | |
5429 | @findex LD_INIT_SWITCH | |
5430 | If defined, a C string constant for a switch that tells the linker that | |
5431 | the following symbol is an initialization routine. | |
5432 | ||
5433 | @item LD_FINI_SWITCH | |
5434 | @findex LD_FINI_SWITCH | |
5435 | If defined, a C string constant for a switch that tells the linker that | |
5436 | the following symbol is a finalization routine. | |
5437 | ||
5438 | @item INVOKE__main | |
5439 | @findex INVOKE__main | |
5440 | If defined, @code{main} will call @code{__main} despite the presence of | |
5441 | @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems | |
5442 | where the init section is not actually run automatically, but is still | |
5443 | useful for collecting the lists of constructors and destructors. | |
5444 | ||
5445 | @item ASM_OUTPUT_CONSTRUCTOR (@var{stream}, @var{name}) | |
5446 | @findex ASM_OUTPUT_CONSTRUCTOR | |
5447 | Define this macro as a C statement to output on the stream @var{stream} | |
5448 | the assembler code to arrange to call the function named @var{name} at | |
5449 | initialization time. | |
5450 | ||
5451 | Assume that @var{name} is the name of a C function generated | |
5452 | automatically by the compiler. This function takes no arguments. Use | |
5453 | the function @code{assemble_name} to output the name @var{name}; this | |
5454 | performs any system-specific syntactic transformations such as adding an | |
5455 | underscore. | |
5456 | ||
5457 | If you don't define this macro, nothing special is output to arrange to | |
5458 | call the function. This is correct when the function will be called in | |
5459 | some other manner---for example, by means of the @code{collect2} program, | |
5460 | which looks through the symbol table to find these functions by their | |
5461 | names. | |
5462 | ||
5463 | @item ASM_OUTPUT_DESTRUCTOR (@var{stream}, @var{name}) | |
5464 | @findex ASM_OUTPUT_DESTRUCTOR | |
5465 | This is like @code{ASM_OUTPUT_CONSTRUCTOR} but used for termination | |
5466 | functions rather than initialization functions. | |
5467 | @end table | |
5468 | ||
5469 | If your system uses @code{collect2} as the means of processing | |
5470 | constructors, then that program normally uses @code{nm} to scan an | |
5471 | object file for constructor functions to be called. On certain kinds of | |
5472 | systems, you can define these macros to make @code{collect2} work faster | |
5473 | (and, in some cases, make it work at all): | |
5474 | ||
5475 | @table @code | |
5476 | @findex OBJECT_FORMAT_COFF | |
5477 | @item OBJECT_FORMAT_COFF | |
5478 | Define this macro if the system uses COFF (Common Object File Format) | |
5479 | object files, so that @code{collect2} can assume this format and scan | |
5480 | object files directly for dynamic constructor/destructor functions. | |
5481 | ||
5482 | @findex OBJECT_FORMAT_ROSE | |
5483 | @item OBJECT_FORMAT_ROSE | |
5484 | Define this macro if the system uses ROSE format object files, so that | |
5485 | @code{collect2} can assume this format and scan object files directly | |
5486 | for dynamic constructor/destructor functions. | |
5487 | ||
5488 | These macros are effective only in a native compiler; @code{collect2} as | |
5489 | part of a cross compiler always uses @code{nm} for the target machine. | |
5490 | ||
5491 | @findex REAL_NM_FILE_NAME | |
5492 | @item REAL_NM_FILE_NAME | |
5493 | Define this macro as a C string constant containing the file name to use | |
5494 | to execute @code{nm}. The default is to search the path normally for | |
5495 | @code{nm}. | |
5496 | ||
5497 | If your system supports shared libraries and has a program to list the | |
5498 | dynamic dependencies of a given library or executable, you can define | |
5499 | these macros to enable support for running initialization and | |
5500 | termination functions in shared libraries: | |
5501 | ||
5502 | @findex LDD_SUFFIX | |
5503 | @item LDD_SUFFIX | |
5504 | Define this macro to a C string constant containing the name of the | |
5505 | program which lists dynamic dependencies, like @code{"ldd"} under SunOS 4. | |
5506 | ||
5507 | @findex PARSE_LDD_OUTPUT | |
5508 | @item PARSE_LDD_OUTPUT (@var{PTR}) | |
5509 | Define this macro to be C code that extracts filenames from the output | |
5510 | of the program denoted by @code{LDD_SUFFIX}. @var{PTR} is a variable | |
5511 | of type @code{char *} that points to the beginning of a line of output | |
5512 | from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the | |
5513 | code must advance @var{PTR} to the beginning of the filename on that | |
5514 | line. Otherwise, it must set @var{PTR} to @code{NULL}. | |
5515 | ||
5516 | @end table | |
5517 | ||
5518 | @node Instruction Output | |
5519 | @subsection Output of Assembler Instructions | |
5520 | ||
5521 | @c prevent bad page break with this line | |
5522 | This describes assembler instruction output. | |
5523 | ||
5524 | @table @code | |
5525 | @findex REGISTER_NAMES | |
5526 | @item REGISTER_NAMES | |
5527 | A C initializer containing the assembler's names for the machine | |
5528 | registers, each one as a C string constant. This is what translates | |
5529 | register numbers in the compiler into assembler language. | |
5530 | ||
5531 | @findex ADDITIONAL_REGISTER_NAMES | |
5532 | @item ADDITIONAL_REGISTER_NAMES | |
5533 | If defined, a C initializer for an array of structures containing a name | |
5534 | and a register number. This macro defines additional names for hard | |
5535 | registers, thus allowing the @code{asm} option in declarations to refer | |
5536 | to registers using alternate names. | |
5537 | ||
5538 | @findex ASM_OUTPUT_OPCODE | |
5539 | @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr}) | |
5540 | Define this macro if you are using an unusual assembler that | |
5541 | requires different names for the machine instructions. | |
5542 | ||
5543 | The definition is a C statement or statements which output an | |
5544 | assembler instruction opcode to the stdio stream @var{stream}. The | |
5545 | macro-operand @var{ptr} is a variable of type @code{char *} which | |
5546 | points to the opcode name in its ``internal'' form---the form that is | |
5547 | written in the machine description. The definition should output the | |
5548 | opcode name to @var{stream}, performing any translation you desire, and | |
5549 | increment the variable @var{ptr} to point at the end of the opcode | |
5550 | so that it will not be output twice. | |
5551 | ||
5552 | In fact, your macro definition may process less than the entire opcode | |
5553 | name, or more than the opcode name; but if you want to process text | |
5554 | that includes @samp{%}-sequences to substitute operands, you must take | |
5555 | care of the substitution yourself. Just be sure to increment | |
5556 | @var{ptr} over whatever text should not be output normally. | |
5557 | ||
5558 | @findex recog_operand | |
5559 | If you need to look at the operand values, they can be found as the | |
5560 | elements of @code{recog_operand}. | |
5561 | ||
5562 | If the macro definition does nothing, the instruction is output | |
5563 | in the usual way. | |
5564 | ||
5565 | @findex FINAL_PRESCAN_INSN | |
5566 | @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands}) | |
5567 | If defined, a C statement to be executed just prior to the output of | |
5568 | assembler code for @var{insn}, to modify the extracted operands so | |
5569 | they will be output differently. | |
5570 | ||
5571 | Here the argument @var{opvec} is the vector containing the operands | |
5572 | extracted from @var{insn}, and @var{noperands} is the number of | |
5573 | elements of the vector which contain meaningful data for this insn. | |
5574 | The contents of this vector are what will be used to convert the insn | |
5575 | template into assembler code, so you can change the assembler output | |
5576 | by changing the contents of the vector. | |
5577 | ||
5578 | This macro is useful when various assembler syntaxes share a single | |
5579 | file of instruction patterns; by defining this macro differently, you | |
5580 | can cause a large class of instructions to be output differently (such | |
5581 | as with rearranged operands). Naturally, variations in assembler | |
5582 | syntax affecting individual insn patterns ought to be handled by | |
5583 | writing conditional output routines in those patterns. | |
5584 | ||
5585 | If this macro is not defined, it is equivalent to a null statement. | |
5586 | ||
5587 | @findex FINAL_PRESCAN_LABEL | |
5588 | @item FINAL_PRESCAN_LABEL | |
5589 | If defined, @code{FINAL_PRESCAN_INSN} will be called on each | |
5590 | @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and | |
5591 | @var{noperands} will be zero. | |
5592 | ||
5593 | @findex PRINT_OPERAND | |
5594 | @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code}) | |
5595 | A C compound statement to output to stdio stream @var{stream} the | |
5596 | assembler syntax for an instruction operand @var{x}. @var{x} is an | |
5597 | RTL expression. | |
5598 | ||
5599 | @var{code} is a value that can be used to specify one of several ways | |
5600 | of printing the operand. It is used when identical operands must be | |
5601 | printed differently depending on the context. @var{code} comes from | |
5602 | the @samp{%} specification that was used to request printing of the | |
5603 | operand. If the specification was just @samp{%@var{digit}} then | |
5604 | @var{code} is 0; if the specification was @samp{%@var{ltr} | |
5605 | @var{digit}} then @var{code} is the ASCII code for @var{ltr}. | |
5606 | ||
5607 | @findex reg_names | |
5608 | If @var{x} is a register, this macro should print the register's name. | |
5609 | The names can be found in an array @code{reg_names} whose type is | |
5610 | @code{char *[]}. @code{reg_names} is initialized from | |
5611 | @code{REGISTER_NAMES}. | |
5612 | ||
5613 | When the machine description has a specification @samp{%@var{punct}} | |
5614 | (a @samp{%} followed by a punctuation character), this macro is called | |
5615 | with a null pointer for @var{x} and the punctuation character for | |
5616 | @var{code}. | |
5617 | ||
5618 | @findex PRINT_OPERAND_PUNCT_VALID_P | |
5619 | @item PRINT_OPERAND_PUNCT_VALID_P (@var{code}) | |
5620 | A C expression which evaluates to true if @var{code} is a valid | |
5621 | punctuation character for use in the @code{PRINT_OPERAND} macro. If | |
5622 | @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no | |
5623 | punctuation characters (except for the standard one, @samp{%}) are used | |
5624 | in this way. | |
5625 | ||
5626 | @findex PRINT_OPERAND_ADDRESS | |
5627 | @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x}) | |
5628 | A C compound statement to output to stdio stream @var{stream} the | |
5629 | assembler syntax for an instruction operand that is a memory reference | |
5630 | whose address is @var{x}. @var{x} is an RTL expression. | |
5631 | ||
5632 | @cindex @code{ENCODE_SECTION_INFO} usage | |
5633 | On some machines, the syntax for a symbolic address depends on the | |
5634 | section that the address refers to. On these machines, define the macro | |
5635 | @code{ENCODE_SECTION_INFO} to store the information into the | |
5636 | @code{symbol_ref}, and then check for it here. @xref{Assembler Format}. | |
5637 | ||
5638 | @findex DBR_OUTPUT_SEQEND | |
5639 | @findex dbr_sequence_length | |
5640 | @item DBR_OUTPUT_SEQEND(@var{file}) | |
5641 | A C statement, to be executed after all slot-filler instructions have | |
5642 | been output. If necessary, call @code{dbr_sequence_length} to | |
5643 | determine the number of slots filled in a sequence (zero if not | |
5644 | currently outputting a sequence), to decide how many no-ops to output, | |
5645 | or whatever. | |
5646 | ||
5647 | Don't define this macro if it has nothing to do, but it is helpful in | |
5648 | reading assembly output if the extent of the delay sequence is made | |
5649 | explicit (e.g. with white space). | |
5650 | ||
5651 | @findex final_sequence | |
5652 | Note that output routines for instructions with delay slots must be | |
5653 | prepared to deal with not being output as part of a sequence (i.e. | |
5654 | when the scheduling pass is not run, or when no slot fillers could be | |
5655 | found.) The variable @code{final_sequence} is null when not | |
5656 | processing a sequence, otherwise it contains the @code{sequence} rtx | |
5657 | being output. | |
5658 | ||
5659 | @findex REGISTER_PREFIX | |
5660 | @findex LOCAL_LABEL_PREFIX | |
5661 | @findex USER_LABEL_PREFIX | |
5662 | @findex IMMEDIATE_PREFIX | |
5663 | @findex asm_fprintf | |
5664 | @item REGISTER_PREFIX | |
5665 | @itemx LOCAL_LABEL_PREFIX | |
5666 | @itemx USER_LABEL_PREFIX | |
5667 | @itemx IMMEDIATE_PREFIX | |
5668 | If defined, C string expressions to be used for the @samp{%R}, @samp{%L}, | |
5669 | @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see | |
5670 | @file{final.c}). These are useful when a single @file{md} file must | |
5671 | support multiple assembler formats. In that case, the various @file{tm.h} | |
5672 | files can define these macros differently. | |
5673 | ||
5674 | @findex ASSEMBLER_DIALECT | |
5675 | @item ASSEMBLER_DIALECT | |
5676 | If your target supports multiple dialects of assembler language (such as | |
5677 | different opcodes), define this macro as a C expression that gives the | |
5678 | numeric index of the assembler language dialect to use, with zero as the | |
5679 | first variant. | |
5680 | ||
5681 | If this macro is defined, you may use constructs of the form | |
5682 | @samp{@{option0|option1|option2@dots{}@}} in the output | |
5683 | templates of patterns (@pxref{Output Template}) or in the first argument | |
5684 | of @code{asm_fprintf}. This construct outputs @samp{option0}, | |
5685 | @samp{option1} or @samp{option2}, etc., if the value of | |
5686 | @code{ASSEMBLER_DIALECT} is zero, one or two, etc. Any special | |
5687 | characters within these strings retain their usual meaning. | |
5688 | ||
5689 | If you do not define this macro, the characters @samp{@{}, @samp{|} and | |
5690 | @samp{@}} do not have any special meaning when used in templates or | |
5691 | operands to @code{asm_fprintf}. | |
5692 | ||
5693 | Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX}, | |
5694 | @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express | |
5695 | the variations in assemble language syntax with that mechanism. Define | |
5696 | @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax | |
5697 | if the syntax variant are larger and involve such things as different | |
5698 | opcodes or operand order. | |
5699 | ||
5700 | @findex ASM_OUTPUT_REG_PUSH | |
5701 | @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno}) | |
5702 | A C expression to output to @var{stream} some assembler code | |
5703 | which will push hard register number @var{regno} onto the stack. | |
5704 | The code need not be optimal, since this macro is used only when | |
5705 | profiling. | |
5706 | ||
5707 | @findex ASM_OUTPUT_REG_POP | |
5708 | @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno}) | |
5709 | A C expression to output to @var{stream} some assembler code | |
5710 | which will pop hard register number @var{regno} off of the stack. | |
5711 | The code need not be optimal, since this macro is used only when | |
5712 | profiling. | |
5713 | @end table | |
5714 | ||
5715 | @node Dispatch Tables | |
5716 | @subsection Output of Dispatch Tables | |
5717 | ||
5718 | @c prevent bad page break with this line | |
5719 | This concerns dispatch tables. | |
5720 | ||
5721 | @table @code | |
5722 | @cindex dispatch table | |
5723 | @findex ASM_OUTPUT_ADDR_DIFF_ELT | |
5724 | @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{value}, @var{rel}) | |
5725 | A C statement to output to the stdio stream @var{stream} an assembler | |
5726 | pseudo-instruction to generate a difference between two labels. | |
5727 | @var{value} and @var{rel} are the numbers of two internal labels. The | |
5728 | definitions of these labels are output using | |
5729 | @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same | |
5730 | way here. For example, | |
5731 | ||
5732 | @example | |
5733 | fprintf (@var{stream}, "\t.word L%d-L%d\n", | |
5734 | @var{value}, @var{rel}) | |
5735 | @end example | |
5736 | ||
5737 | You must provide this macro on machines where the addresses in a | |
5738 | dispatch table are relative to the table's own address. If defined, GNU | |
5739 | CC will also use this macro on all machines when producing PIC. | |
5740 | ||
5741 | @findex ASM_OUTPUT_ADDR_VEC_ELT | |
5742 | @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value}) | |
5743 | This macro should be provided on machines where the addresses | |
5744 | in a dispatch table are absolute. | |
5745 | ||
5746 | The definition should be a C statement to output to the stdio stream | |
5747 | @var{stream} an assembler pseudo-instruction to generate a reference to | |
5748 | a label. @var{value} is the number of an internal label whose | |
5749 | definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}. | |
5750 | For example, | |
5751 | ||
5752 | @example | |
5753 | fprintf (@var{stream}, "\t.word L%d\n", @var{value}) | |
5754 | @end example | |
5755 | ||
5756 | @findex ASM_OUTPUT_CASE_LABEL | |
5757 | @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table}) | |
5758 | Define this if the label before a jump-table needs to be output | |
5759 | specially. The first three arguments are the same as for | |
5760 | @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the | |
5761 | jump-table which follows (a @code{jump_insn} containing an | |
5762 | @code{addr_vec} or @code{addr_diff_vec}). | |
5763 | ||
5764 | This feature is used on system V to output a @code{swbeg} statement | |
5765 | for the table. | |
5766 | ||
5767 | If this macro is not defined, these labels are output with | |
5768 | @code{ASM_OUTPUT_INTERNAL_LABEL}. | |
5769 | ||
5770 | @findex ASM_OUTPUT_CASE_END | |
5771 | @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table}) | |
5772 | Define this if something special must be output at the end of a | |
5773 | jump-table. The definition should be a C statement to be executed | |
5774 | after the assembler code for the table is written. It should write | |
5775 | the appropriate code to stdio stream @var{stream}. The argument | |
5776 | @var{table} is the jump-table insn, and @var{num} is the label-number | |
5777 | of the preceding label. | |
5778 | ||
5779 | If this macro is not defined, nothing special is output at the end of | |
5780 | the jump-table. | |
5781 | @end table | |
5782 | ||
5783 | @node Exception Region Output | |
5784 | @subsection Assembler Commands for Exception Regions | |
5785 | ||
5786 | @c prevent bad page break with this line | |
5787 | ||
5788 | This describes commands marking the start and the end of an exception | |
5789 | region. | |
5790 | ||
5791 | @table @code | |
5792 | @findex ASM_OUTPUT_EH_REGION_BEG | |
5793 | @item ASM_OUTPUT_EH_REGION_BEG () | |
5794 | A C expression to output text to mark the start of an exception region. | |
5795 | ||
5796 | This macro need not be defined on most platforms. | |
5797 | ||
5798 | @findex ASM_OUTPUT_EH_REGION_END | |
5799 | @item ASM_OUTPUT_EH_REGION_END () | |
5800 | A C expression to output text to mark the end of an exception region. | |
5801 | ||
5802 | This macro need not be defined on most platforms. | |
5803 | ||
5804 | @findex OMIT_EH_TABLE | |
5805 | @item OMIT_EH_TABLE () | |
5806 | A C expression that is nonzero if the normal exception table output | |
5807 | should be omitted. | |
5808 | ||
5809 | This macro need not be defined on most platforms. | |
5810 | ||
5811 | @findex EH_TABLE_LOOKUP | |
5812 | @item EH_TABLE_LOOKUP () | |
5813 | Alternate runtime support for looking up an exception at runtime and | |
5814 | finding the associated handler, if the default method won't work. | |
5815 | ||
5816 | This macro need not be defined on most platforms. | |
5817 | ||
5818 | @findex DOESNT_NEED_UNWINDER | |
5819 | @item DOESNT_NEED_UNWINDER | |
5820 | A C expression that decides whether or not the current function needs to | |
5821 | have a function unwinder generated for it. See the file @code{except.c} | |
5822 | for details on when to define this, and how. | |
5823 | ||
5824 | @findex MASK_RETURN_ADDR | |
5825 | @item MASK_RETURN_ADDR | |
5826 | An rtx used to mask the return address found via RETURN_ADDR_RTX, so | |
5827 | that it does not contain any extraneous set bits in it. | |
5828 | @end table | |
5829 | ||
5830 | @node Alignment Output | |
5831 | @subsection Assembler Commands for Alignment | |
5832 | ||
5833 | @c prevent bad page break with this line | |
5834 | This describes commands for alignment. | |
5835 | ||
5836 | @table @code | |
5837 | @findex ASM_OUTPUT_ALIGN_CODE | |
5838 | @item ASM_OUTPUT_ALIGN_CODE (@var{file}) | |
5839 | A C expression to output text to align the location counter in the way | |
5840 | that is desirable at a point in the code that is reached only by | |
5841 | jumping. | |
5842 | ||
5843 | This macro need not be defined if you don't want any special alignment | |
5844 | to be done at such a time. Most machine descriptions do not currently | |
5845 | define the macro. | |
5846 | ||
5847 | @findex ASM_OUTPUT_LOOP_ALIGN | |
5848 | @item ASM_OUTPUT_LOOP_ALIGN (@var{file}) | |
5849 | A C expression to output text to align the location counter in the way | |
5850 | that is desirable at the beginning of a loop. | |
5851 | ||
5852 | This macro need not be defined if you don't want any special alignment | |
5853 | to be done at such a time. Most machine descriptions do not currently | |
5854 | define the macro. | |
5855 | ||
5856 | @findex ASM_OUTPUT_SKIP | |
5857 | @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes}) | |
5858 | A C statement to output to the stdio stream @var{stream} an assembler | |
5859 | instruction to advance the location counter by @var{nbytes} bytes. | |
5860 | Those bytes should be zero when loaded. @var{nbytes} will be a C | |
5861 | expression of type @code{int}. | |
5862 | ||
5863 | @findex ASM_NO_SKIP_IN_TEXT | |
5864 | @item ASM_NO_SKIP_IN_TEXT | |
5865 | Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the | |
5866 | text section because it fails put zeros in the bytes that are skipped. | |
5867 | This is true on many Unix systems, where the pseudo--op to skip bytes | |
5868 | produces no-op instructions rather than zeros when used in the text | |
5869 | section. | |
5870 | ||
5871 | @findex ASM_OUTPUT_ALIGN | |
5872 | @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power}) | |
5873 | A C statement to output to the stdio stream @var{stream} an assembler | |
5874 | command to advance the location counter to a multiple of 2 to the | |
5875 | @var{power} bytes. @var{power} will be a C expression of type @code{int}. | |
5876 | @end table | |
5877 | ||
5878 | @need 3000 | |
5879 | @node Debugging Info | |
5880 | @section Controlling Debugging Information Format | |
5881 | ||
5882 | @c prevent bad page break with this line | |
5883 | This describes how to specify debugging information. | |
5884 | ||
5885 | @menu | |
5886 | * All Debuggers:: Macros that affect all debugging formats uniformly. | |
5887 | * DBX Options:: Macros enabling specific options in DBX format. | |
5888 | * DBX Hooks:: Hook macros for varying DBX format. | |
5889 | * File Names and DBX:: Macros controlling output of file names in DBX format. | |
5890 | * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats. | |
5891 | @end menu | |
5892 | ||
5893 | @node All Debuggers | |
5894 | @subsection Macros Affecting All Debugging Formats | |
5895 | ||
5896 | @c prevent bad page break with this line | |
5897 | These macros affect all debugging formats. | |
5898 | ||
5899 | @table @code | |
5900 | @findex DBX_REGISTER_NUMBER | |
5901 | @item DBX_REGISTER_NUMBER (@var{regno}) | |
5902 | A C expression that returns the DBX register number for the compiler | |
5903 | register number @var{regno}. In simple cases, the value of this | |
5904 | expression may be @var{regno} itself. But sometimes there are some | |
5905 | registers that the compiler knows about and DBX does not, or vice | |
5906 | versa. In such cases, some register may need to have one number in | |
5907 | the compiler and another for DBX. | |
5908 | ||
5909 | If two registers have consecutive numbers inside GNU CC, and they can be | |
5910 | used as a pair to hold a multiword value, then they @emph{must} have | |
5911 | consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}. | |
5912 | Otherwise, debuggers will be unable to access such a pair, because they | |
5913 | expect register pairs to be consecutive in their own numbering scheme. | |
5914 | ||
5915 | If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that | |
5916 | does not preserve register pairs, then what you must do instead is | |
5917 | redefine the actual register numbering scheme. | |
5918 | ||
5919 | @findex DEBUGGER_AUTO_OFFSET | |
5920 | @item DEBUGGER_AUTO_OFFSET (@var{x}) | |
5921 | A C expression that returns the integer offset value for an automatic | |
5922 | variable having address @var{x} (an RTL expression). The default | |
5923 | computation assumes that @var{x} is based on the frame-pointer and | |
5924 | gives the offset from the frame-pointer. This is required for targets | |
5925 | that produce debugging output for DBX or COFF-style debugging output | |
5926 | for SDB and allow the frame-pointer to be eliminated when the | |
5927 | @samp{-g} options is used. | |
5928 | ||
5929 | @findex DEBUGGER_ARG_OFFSET | |
5930 | @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x}) | |
5931 | A C expression that returns the integer offset value for an argument | |
5932 | having address @var{x} (an RTL expression). The nominal offset is | |
5933 | @var{offset}. | |
5934 | ||
5935 | @findex PREFERRED_DEBUGGING_TYPE | |
5936 | @item PREFERRED_DEBUGGING_TYPE | |
5937 | A C expression that returns the type of debugging output GNU CC produces | |
5938 | when the user specifies @samp{-g} or @samp{-ggdb}. Define this if you | |
5939 | have arranged for GNU CC to support more than one format of debugging | |
5940 | output. Currently, the allowable values are @code{DBX_DEBUG}, | |
5941 | @code{SDB_DEBUG}, @code{DWARF_DEBUG}, and @code{XCOFF_DEBUG}. | |
5942 | ||
5943 | The value of this macro only affects the default debugging output; the | |
5944 | user can always get a specific type of output by using @samp{-gstabs}, | |
5945 | @samp{-gcoff}, @samp{-gdwarf}, or @samp{-gxcoff}. | |
5946 | @end table | |
5947 | ||
5948 | @node DBX Options | |
5949 | @subsection Specific Options for DBX Output | |
5950 | ||
5951 | @c prevent bad page break with this line | |
5952 | These are specific options for DBX output. | |
5953 | ||
5954 | @table @code | |
5955 | @findex DBX_DEBUGGING_INFO | |
5956 | @item DBX_DEBUGGING_INFO | |
5957 | Define this macro if GNU CC should produce debugging output for DBX | |
5958 | in response to the @samp{-g} option. | |
5959 | ||
5960 | @findex XCOFF_DEBUGGING_INFO | |
5961 | @item XCOFF_DEBUGGING_INFO | |
5962 | Define this macro if GNU CC should produce XCOFF format debugging output | |
5963 | in response to the @samp{-g} option. This is a variant of DBX format. | |
5964 | ||
5965 | @findex DEFAULT_GDB_EXTENSIONS | |
5966 | @item DEFAULT_GDB_EXTENSIONS | |
5967 | Define this macro to control whether GNU CC should by default generate | |
5968 | GDB's extended version of DBX debugging information (assuming DBX-format | |
5969 | debugging information is enabled at all). If you don't define the | |
5970 | macro, the default is 1: always generate the extended information | |
5971 | if there is any occasion to. | |
5972 | ||
5973 | @findex DEBUG_SYMS_TEXT | |
5974 | @item DEBUG_SYMS_TEXT | |
5975 | Define this macro if all @code{.stabs} commands should be output while | |
5976 | in the text section. | |
5977 | ||
5978 | @findex ASM_STABS_OP | |
5979 | @item ASM_STABS_OP | |
5980 | A C string constant naming the assembler pseudo op to use instead of | |
5981 | @code{.stabs} to define an ordinary debugging symbol. If you don't | |
5982 | define this macro, @code{.stabs} is used. This macro applies only to | |
5983 | DBX debugging information format. | |
5984 | ||
5985 | @findex ASM_STABD_OP | |
5986 | @item ASM_STABD_OP | |
5987 | A C string constant naming the assembler pseudo op to use instead of | |
5988 | @code{.stabd} to define a debugging symbol whose value is the current | |
5989 | location. If you don't define this macro, @code{.stabd} is used. | |
5990 | This macro applies only to DBX debugging information format. | |
5991 | ||
5992 | @findex ASM_STABN_OP | |
5993 | @item ASM_STABN_OP | |
5994 | A C string constant naming the assembler pseudo op to use instead of | |
5995 | @code{.stabn} to define a debugging symbol with no name. If you don't | |
5996 | define this macro, @code{.stabn} is used. This macro applies only to | |
5997 | DBX debugging information format. | |
5998 | ||
5999 | @findex DBX_NO_XREFS | |
6000 | @item DBX_NO_XREFS | |
6001 | Define this macro if DBX on your system does not support the construct | |
6002 | @samp{xs@var{tagname}}. On some systems, this construct is used to | |
6003 | describe a forward reference to a structure named @var{tagname}. | |
6004 | On other systems, this construct is not supported at all. | |
6005 | ||
6006 | @findex DBX_CONTIN_LENGTH | |
6007 | @item DBX_CONTIN_LENGTH | |
6008 | A symbol name in DBX-format debugging information is normally | |
6009 | continued (split into two separate @code{.stabs} directives) when it | |
6010 | exceeds a certain length (by default, 80 characters). On some | |
6011 | operating systems, DBX requires this splitting; on others, splitting | |
6012 | must not be done. You can inhibit splitting by defining this macro | |
6013 | with the value zero. You can override the default splitting-length by | |
6014 | defining this macro as an expression for the length you desire. | |
6015 | ||
6016 | @findex DBX_CONTIN_CHAR | |
6017 | @item DBX_CONTIN_CHAR | |
6018 | Normally continuation is indicated by adding a @samp{\} character to | |
6019 | the end of a @code{.stabs} string when a continuation follows. To use | |
6020 | a different character instead, define this macro as a character | |
6021 | constant for the character you want to use. Do not define this macro | |
6022 | if backslash is correct for your system. | |
6023 | ||
6024 | @findex DBX_STATIC_STAB_DATA_SECTION | |
6025 | @item DBX_STATIC_STAB_DATA_SECTION | |
6026 | Define this macro if it is necessary to go to the data section before | |
6027 | outputting the @samp{.stabs} pseudo-op for a non-global static | |
6028 | variable. | |
6029 | ||
6030 | @findex DBX_TYPE_DECL_STABS_CODE | |
6031 | @item DBX_TYPE_DECL_STABS_CODE | |
6032 | The value to use in the ``code'' field of the @code{.stabs} directive | |
6033 | for a typedef. The default is @code{N_LSYM}. | |
6034 | ||
6035 | @findex DBX_STATIC_CONST_VAR_CODE | |
6036 | @item DBX_STATIC_CONST_VAR_CODE | |
6037 | The value to use in the ``code'' field of the @code{.stabs} directive | |
6038 | for a static variable located in the text section. DBX format does not | |
6039 | provide any ``right'' way to do this. The default is @code{N_FUN}. | |
6040 | ||
6041 | @findex DBX_REGPARM_STABS_CODE | |
6042 | @item DBX_REGPARM_STABS_CODE | |
6043 | The value to use in the ``code'' field of the @code{.stabs} directive | |
6044 | for a parameter passed in registers. DBX format does not provide any | |
6045 | ``right'' way to do this. The default is @code{N_RSYM}. | |
6046 | ||
6047 | @findex DBX_REGPARM_STABS_LETTER | |
6048 | @item DBX_REGPARM_STABS_LETTER | |
6049 | The letter to use in DBX symbol data to identify a symbol as a parameter | |
6050 | passed in registers. DBX format does not customarily provide any way to | |
6051 | do this. The default is @code{'P'}. | |
6052 | ||
6053 | @findex DBX_MEMPARM_STABS_LETTER | |
6054 | @item DBX_MEMPARM_STABS_LETTER | |
6055 | The letter to use in DBX symbol data to identify a symbol as a stack | |
6056 | parameter. The default is @code{'p'}. | |
6057 | ||
6058 | @findex DBX_FUNCTION_FIRST | |
6059 | @item DBX_FUNCTION_FIRST | |
6060 | Define this macro if the DBX information for a function and its | |
6061 | arguments should precede the assembler code for the function. Normally, | |
6062 | in DBX format, the debugging information entirely follows the assembler | |
6063 | code. | |
6064 | ||
6065 | @findex DBX_LBRAC_FIRST | |
6066 | @item DBX_LBRAC_FIRST | |
6067 | Define this macro if the @code{N_LBRAC} symbol for a block should | |
6068 | precede the debugging information for variables and functions defined in | |
6069 | that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes | |
6070 | first. | |
6071 | ||
6072 | @findex DBX_BLOCKS_FUNCTION_RELATIVE | |
6073 | @item DBX_BLOCKS_FUNCTION_RELATIVE | |
6074 | Define this macro if the value of a symbol describing the scope of a | |
6075 | block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start | |
6076 | of the enclosing function. Normally, GNU C uses an absolute address. | |
6077 | ||
6078 | @findex DBX_USE_BINCL | |
6079 | @item DBX_USE_BINCL | |
6080 | Define this macro if GNU C should generate @code{N_BINCL} and | |
6081 | @code{N_EINCL} stabs for included header files, as on Sun systems. This | |
6082 | macro also directs GNU C to output a type number as a pair of a file | |
6083 | number and a type number within the file. Normally, GNU C does not | |
6084 | generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single | |
6085 | number for a type number. | |
6086 | @end table | |
6087 | ||
6088 | @node DBX Hooks | |
6089 | @subsection Open-Ended Hooks for DBX Format | |
6090 | ||
6091 | @c prevent bad page break with this line | |
6092 | These are hooks for DBX format. | |
6093 | ||
6094 | @table @code | |
6095 | @findex DBX_OUTPUT_LBRAC | |
6096 | @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name}) | |
6097 | Define this macro to say how to output to @var{stream} the debugging | |
6098 | information for the start of a scope level for variable names. The | |
6099 | argument @var{name} is the name of an assembler symbol (for use with | |
6100 | @code{assemble_name}) whose value is the address where the scope begins. | |
6101 | ||
6102 | @findex DBX_OUTPUT_RBRAC | |
6103 | @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name}) | |
6104 | Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level. | |
6105 | ||
6106 | @findex DBX_OUTPUT_ENUM | |
6107 | @item DBX_OUTPUT_ENUM (@var{stream}, @var{type}) | |
6108 | Define this macro if the target machine requires special handling to | |
6109 | output an enumeration type. The definition should be a C statement | |
6110 | (sans semicolon) to output the appropriate information to @var{stream} | |
6111 | for the type @var{type}. | |
6112 | ||
6113 | @findex DBX_OUTPUT_FUNCTION_END | |
6114 | @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function}) | |
6115 | Define this macro if the target machine requires special output at the | |
6116 | end of the debugging information for a function. The definition should | |
6117 | be a C statement (sans semicolon) to output the appropriate information | |
6118 | to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for | |
6119 | the function. | |
6120 | ||
6121 | @findex DBX_OUTPUT_STANDARD_TYPES | |
6122 | @item DBX_OUTPUT_STANDARD_TYPES (@var{syms}) | |
6123 | Define this macro if you need to control the order of output of the | |
6124 | standard data types at the beginning of compilation. The argument | |
6125 | @var{syms} is a @code{tree} which is a chain of all the predefined | |
6126 | global symbols, including names of data types. | |
6127 | ||
6128 | Normally, DBX output starts with definitions of the types for integers | |
6129 | and characters, followed by all the other predefined types of the | |
6130 | particular language in no particular order. | |
6131 | ||
6132 | On some machines, it is necessary to output different particular types | |
6133 | first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output | |
6134 | those symbols in the necessary order. Any predefined types that you | |
6135 | don't explicitly output will be output afterward in no particular order. | |
6136 | ||
6137 | Be careful not to define this macro so that it works only for C. There | |
6138 | are no global variables to access most of the built-in types, because | |
6139 | another language may have another set of types. The way to output a | |
6140 | particular type is to look through @var{syms} to see if you can find it. | |
6141 | Here is an example: | |
6142 | ||
6143 | @smallexample | |
6144 | @{ | |
6145 | tree decl; | |
6146 | for (decl = syms; decl; decl = TREE_CHAIN (decl)) | |
6147 | if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)), | |
6148 | "long int")) | |
6149 | dbxout_symbol (decl); | |
6150 | @dots{} | |
6151 | @} | |
6152 | @end smallexample | |
6153 | ||
6154 | @noindent | |
6155 | This does nothing if the expected type does not exist. | |
6156 | ||
6157 | See the function @code{init_decl_processing} in @file{c-decl.c} to find | |
6158 | the names to use for all the built-in C types. | |
6159 | ||
6160 | Here is another way of finding a particular type: | |
6161 | ||
6162 | @c this is still overfull. --mew 10feb93 | |
6163 | @smallexample | |
6164 | @{ | |
6165 | tree decl; | |
6166 | for (decl = syms; decl; decl = TREE_CHAIN (decl)) | |
6167 | if (TREE_CODE (decl) == TYPE_DECL | |
6168 | && (TREE_CODE (TREE_TYPE (decl)) | |
6169 | == INTEGER_CST) | |
6170 | && TYPE_PRECISION (TREE_TYPE (decl)) == 16 | |
6171 | && TYPE_UNSIGNED (TREE_TYPE (decl))) | |
6172 | @group | |
6173 | /* @r{This must be @code{unsigned short}.} */ | |
6174 | dbxout_symbol (decl); | |
6175 | @dots{} | |
6176 | @} | |
6177 | @end group | |
6178 | @end smallexample | |
6179 | ||
6180 | @findex NO_DBX_FUNCTION_END | |
6181 | @item NO_DBX_FUNCTION_END | |
6182 | Some stabs encapsulation formats (in particular ECOFF), cannot handle the | |
6183 | @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extention construct. | |
6184 | On those machines, define this macro to turn this feature off without | |
6185 | disturbing the rest of the gdb extensions. | |
6186 | ||
6187 | @end table | |
6188 | ||
6189 | @node File Names and DBX | |
6190 | @subsection File Names in DBX Format | |
6191 | ||
6192 | @c prevent bad page break with this line | |
6193 | This describes file names in DBX format. | |
6194 | ||
6195 | @table @code | |
6196 | @findex DBX_WORKING_DIRECTORY | |
6197 | @item DBX_WORKING_DIRECTORY | |
6198 | Define this if DBX wants to have the current directory recorded in each | |
6199 | object file. | |
6200 | ||
6201 | Note that the working directory is always recorded if GDB extensions are | |
6202 | enabled. | |
6203 | ||
6204 | @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME | |
6205 | @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name}) | |
6206 | A C statement to output DBX debugging information to the stdio stream | |
6207 | @var{stream} which indicates that file @var{name} is the main source | |
6208 | file---the file specified as the input file for compilation. | |
6209 | This macro is called only once, at the beginning of compilation. | |
6210 | ||
6211 | This macro need not be defined if the standard form of output | |
6212 | for DBX debugging information is appropriate. | |
6213 | ||
6214 | @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY | |
6215 | @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name}) | |
6216 | A C statement to output DBX debugging information to the stdio stream | |
6217 | @var{stream} which indicates that the current directory during | |
6218 | compilation is named @var{name}. | |
6219 | ||
6220 | This macro need not be defined if the standard form of output | |
6221 | for DBX debugging information is appropriate. | |
6222 | ||
6223 | @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END | |
6224 | @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name}) | |
6225 | A C statement to output DBX debugging information at the end of | |
6226 | compilation of the main source file @var{name}. | |
6227 | ||
6228 | If you don't define this macro, nothing special is output at the end | |
6229 | of compilation, which is correct for most machines. | |
6230 | ||
6231 | @findex DBX_OUTPUT_SOURCE_FILENAME | |
6232 | @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name}) | |
6233 | A C statement to output DBX debugging information to the stdio stream | |
6234 | @var{stream} which indicates that file @var{name} is the current source | |
6235 | file. This output is generated each time input shifts to a different | |
6236 | source file as a result of @samp{#include}, the end of an included file, | |
6237 | or a @samp{#line} command. | |
6238 | ||
6239 | This macro need not be defined if the standard form of output | |
6240 | for DBX debugging information is appropriate. | |
6241 | @end table | |
6242 | ||
6243 | @need 2000 | |
6244 | @node SDB and DWARF | |
6245 | @subsection Macros for SDB and DWARF Output | |
6246 | ||
6247 | @c prevent bad page break with this line | |
6248 | Here are macros for SDB and DWARF output. | |
6249 | ||
6250 | @table @code | |
6251 | @findex SDB_DEBUGGING_INFO | |
6252 | @item SDB_DEBUGGING_INFO | |
6253 | Define this macro if GNU CC should produce COFF-style debugging output | |
6254 | for SDB in response to the @samp{-g} option. | |
6255 | ||
6256 | @findex DWARF_DEBUGGING_INFO | |
6257 | @item DWARF_DEBUGGING_INFO | |
6258 | Define this macro if GNU CC should produce dwarf format debugging output | |
6259 | in response to the @samp{-g} option. | |
6260 | ||
6261 | @findex PUT_SDB_@dots{} | |
6262 | @item PUT_SDB_@dots{} | |
6263 | Define these macros to override the assembler syntax for the special | |
6264 | SDB assembler directives. See @file{sdbout.c} for a list of these | |
6265 | macros and their arguments. If the standard syntax is used, you need | |
6266 | not define them yourself. | |
6267 | ||
6268 | @findex SDB_DELIM | |
6269 | @item SDB_DELIM | |
6270 | Some assemblers do not support a semicolon as a delimiter, even between | |
6271 | SDB assembler directives. In that case, define this macro to be the | |
6272 | delimiter to use (usually @samp{\n}). It is not necessary to define | |
6273 | a new set of @code{PUT_SDB_@var{op}} macros if this is the only change | |
6274 | required. | |
6275 | ||
6276 | @findex SDB_GENERATE_FAKE | |
6277 | @item SDB_GENERATE_FAKE | |
6278 | Define this macro to override the usual method of constructing a dummy | |
6279 | name for anonymous structure and union types. See @file{sdbout.c} for | |
6280 | more information. | |
6281 | ||
6282 | @findex SDB_ALLOW_UNKNOWN_REFERENCES | |
6283 | @item SDB_ALLOW_UNKNOWN_REFERENCES | |
6284 | Define this macro to allow references to unknown structure, | |
6285 | union, or enumeration tags to be emitted. Standard COFF does not | |
6286 | allow handling of unknown references, MIPS ECOFF has support for | |
6287 | it. | |
6288 | ||
6289 | @findex SDB_ALLOW_FORWARD_REFERENCES | |
6290 | @item SDB_ALLOW_FORWARD_REFERENCES | |
6291 | Define this macro to allow references to structure, union, or | |
6292 | enumeration tags that have not yet been seen to be handled. Some | |
6293 | assemblers choke if forward tags are used, while some require it. | |
6294 | @end table | |
6295 | ||
6296 | @node Cross-compilation | |
6297 | @section Cross Compilation and Floating Point | |
6298 | @cindex cross compilation and floating point | |
6299 | @cindex floating point and cross compilation | |
6300 | ||
6301 | While all modern machines use 2's complement representation for integers, | |
6302 | there are a variety of representations for floating point numbers. This | |
6303 | means that in a cross-compiler the representation of floating point numbers | |
6304 | in the compiled program may be different from that used in the machine | |
6305 | doing the compilation. | |
6306 | ||
6307 | @findex atof | |
6308 | Because different representation systems may offer different amounts of | |
6309 | range and precision, the cross compiler cannot safely use the host | |
6310 | machine's floating point arithmetic. Therefore, floating point constants | |
6311 | must be represented in the target machine's format. This means that the | |
6312 | cross compiler cannot use @code{atof} to parse a floating point constant; | |
6313 | it must have its own special routine to use instead. Also, constant | |
6314 | folding must emulate the target machine's arithmetic (or must not be done | |
6315 | at all). | |
6316 | ||
6317 | The macros in the following table should be defined only if you are cross | |
6318 | compiling between different floating point formats. | |
6319 | ||
6320 | Otherwise, don't define them. Then default definitions will be set up which | |
6321 | use @code{double} as the data type, @code{==} to test for equality, etc. | |
6322 | ||
6323 | You don't need to worry about how many times you use an operand of any | |
6324 | of these macros. The compiler never uses operands which have side effects. | |
6325 | ||
6326 | @table @code | |
6327 | @findex REAL_VALUE_TYPE | |
6328 | @item REAL_VALUE_TYPE | |
6329 | A macro for the C data type to be used to hold a floating point value | |
6330 | in the target machine's format. Typically this would be a | |
6331 | @code{struct} containing an array of @code{int}. | |
6332 | ||
6333 | @findex REAL_VALUES_EQUAL | |
6334 | @item REAL_VALUES_EQUAL (@var{x}, @var{y}) | |
6335 | A macro for a C expression which compares for equality the two values, | |
6336 | @var{x} and @var{y}, both of type @code{REAL_VALUE_TYPE}. | |
6337 | ||
6338 | @findex REAL_VALUES_LESS | |
6339 | @item REAL_VALUES_LESS (@var{x}, @var{y}) | |
6340 | A macro for a C expression which tests whether @var{x} is less than | |
6341 | @var{y}, both values being of type @code{REAL_VALUE_TYPE} and | |
6342 | interpreted as floating point numbers in the target machine's | |
6343 | representation. | |
6344 | ||
6345 | @findex REAL_VALUE_LDEXP | |
6346 | @findex ldexp | |
6347 | @item REAL_VALUE_LDEXP (@var{x}, @var{scale}) | |
6348 | A macro for a C expression which performs the standard library | |
6349 | function @code{ldexp}, but using the target machine's floating point | |
6350 | representation. Both @var{x} and the value of the expression have | |
6351 | type @code{REAL_VALUE_TYPE}. The second argument, @var{scale}, is an | |
6352 | integer. | |
6353 | ||
6354 | @findex REAL_VALUE_FIX | |
6355 | @item REAL_VALUE_FIX (@var{x}) | |
6356 | A macro whose definition is a C expression to convert the target-machine | |
6357 | floating point value @var{x} to a signed integer. @var{x} has type | |
6358 | @code{REAL_VALUE_TYPE}. | |
6359 | ||
6360 | @findex REAL_VALUE_UNSIGNED_FIX | |
6361 | @item REAL_VALUE_UNSIGNED_FIX (@var{x}) | |
6362 | A macro whose definition is a C expression to convert the target-machine | |
6363 | floating point value @var{x} to an unsigned integer. @var{x} has type | |
6364 | @code{REAL_VALUE_TYPE}. | |
6365 | ||
6366 | @findex REAL_VALUE_RNDZINT | |
6367 | @item REAL_VALUE_RNDZINT (@var{x}) | |
6368 | A macro whose definition is a C expression to round the target-machine | |
6369 | floating point value @var{x} towards zero to an integer value (but still | |
6370 | as a floating point number). @var{x} has type @code{REAL_VALUE_TYPE}, | |
6371 | and so does the value. | |
6372 | ||
6373 | @findex REAL_VALUE_UNSIGNED_RNDZINT | |
6374 | @item REAL_VALUE_UNSIGNED_RNDZINT (@var{x}) | |
6375 | A macro whose definition is a C expression to round the target-machine | |
6376 | floating point value @var{x} towards zero to an unsigned integer value | |
6377 | (but still represented as a floating point number). @var{x} has type | |
6378 | @code{REAL_VALUE_TYPE}, and so does the value. | |
6379 | ||
6380 | @findex REAL_VALUE_ATOF | |
6381 | @item REAL_VALUE_ATOF (@var{string}, @var{mode}) | |
6382 | A macro for a C expression which converts @var{string}, an expression of | |
6383 | type @code{char *}, into a floating point number in the target machine's | |
6384 | representation for mode @var{mode}. The value has type | |
6385 | @code{REAL_VALUE_TYPE}. | |
6386 | ||
6387 | @findex REAL_INFINITY | |
6388 | @item REAL_INFINITY | |
6389 | Define this macro if infinity is a possible floating point value, and | |
6390 | therefore division by 0 is legitimate. | |
6391 | ||
6392 | @findex REAL_VALUE_ISINF | |
6393 | @findex isinf | |
6394 | @item REAL_VALUE_ISINF (@var{x}) | |
6395 | A macro for a C expression which determines whether @var{x}, a floating | |
6396 | point value, is infinity. The value has type @code{int}. | |
6397 | By default, this is defined to call @code{isinf}. | |
6398 | ||
6399 | @findex REAL_VALUE_ISNAN | |
6400 | @findex isnan | |
6401 | @item REAL_VALUE_ISNAN (@var{x}) | |
6402 | A macro for a C expression which determines whether @var{x}, a floating | |
6403 | point value, is a ``nan'' (not-a-number). The value has type | |
6404 | @code{int}. By default, this is defined to call @code{isnan}. | |
6405 | @end table | |
6406 | ||
6407 | @cindex constant folding and floating point | |
6408 | Define the following additional macros if you want to make floating | |
6409 | point constant folding work while cross compiling. If you don't | |
6410 | define them, cross compilation is still possible, but constant folding | |
6411 | will not happen for floating point values. | |
6412 | ||
6413 | @table @code | |
6414 | @findex REAL_ARITHMETIC | |
6415 | @item REAL_ARITHMETIC (@var{output}, @var{code}, @var{x}, @var{y}) | |
6416 | A macro for a C statement which calculates an arithmetic operation of | |
6417 | the two floating point values @var{x} and @var{y}, both of type | |
6418 | @code{REAL_VALUE_TYPE} in the target machine's representation, to | |
6419 | produce a result of the same type and representation which is stored | |
6420 | in @var{output} (which will be a variable). | |
6421 | ||
6422 | The operation to be performed is specified by @var{code}, a tree code | |
6423 | which will always be one of the following: @code{PLUS_EXPR}, | |
6424 | @code{MINUS_EXPR}, @code{MULT_EXPR}, @code{RDIV_EXPR}, | |
6425 | @code{MAX_EXPR}, @code{MIN_EXPR}.@refill | |
6426 | ||
6427 | @cindex overflow while constant folding | |
6428 | The expansion of this macro is responsible for checking for overflow. | |
6429 | If overflow happens, the macro expansion should execute the statement | |
6430 | @code{return 0;}, which indicates the inability to perform the | |
6431 | arithmetic operation requested. | |
6432 | ||
6433 | @findex REAL_VALUE_NEGATE | |
6434 | @item REAL_VALUE_NEGATE (@var{x}) | |
6435 | A macro for a C expression which returns the negative of the floating | |
6436 | point value @var{x}. Both @var{x} and the value of the expression | |
6437 | have type @code{REAL_VALUE_TYPE} and are in the target machine's | |
6438 | floating point representation. | |
6439 | ||
6440 | There is no way for this macro to report overflow, since overflow | |
6441 | can't happen in the negation operation. | |
6442 | ||
6443 | @findex REAL_VALUE_TRUNCATE | |
6444 | @item REAL_VALUE_TRUNCATE (@var{mode}, @var{x}) | |
6445 | A macro for a C expression which converts the floating point value | |
6446 | @var{x} to mode @var{mode}. | |
6447 | ||
6448 | Both @var{x} and the value of the expression are in the target machine's | |
6449 | floating point representation and have type @code{REAL_VALUE_TYPE}. | |
6450 | However, the value should have an appropriate bit pattern to be output | |
6451 | properly as a floating constant whose precision accords with mode | |
6452 | @var{mode}. | |
6453 | ||
6454 | There is no way for this macro to report overflow. | |
6455 | ||
6456 | @findex REAL_VALUE_TO_INT | |
6457 | @item REAL_VALUE_TO_INT (@var{low}, @var{high}, @var{x}) | |
6458 | A macro for a C expression which converts a floating point value | |
6459 | @var{x} into a double-precision integer which is then stored into | |
6460 | @var{low} and @var{high}, two variables of type @var{int}. | |
6461 | ||
6462 | @item REAL_VALUE_FROM_INT (@var{x}, @var{low}, @var{high}, @var{mode}) | |
6463 | @findex REAL_VALUE_FROM_INT | |
6464 | A macro for a C expression which converts a double-precision integer | |
6465 | found in @var{low} and @var{high}, two variables of type @var{int}, | |
6466 | into a floating point value which is then stored into @var{x}. | |
6467 | The value is in the target machine's representation for mode @var{mode} | |
6468 | and has the type @code{REAL_VALUE_TYPE}. | |
6469 | @end table | |
6470 | ||
6471 | @node Misc | |
6472 | @section Miscellaneous Parameters | |
6473 | @cindex parameters, miscellaneous | |
6474 | ||
6475 | @c prevent bad page break with this line | |
6476 | Here are several miscellaneous parameters. | |
6477 | ||
6478 | @table @code | |
6479 | @item PREDICATE_CODES | |
6480 | @findex PREDICATE_CODES | |
6481 | Define this if you have defined special-purpose predicates in the file | |
6482 | @file{@var{machine}.c}. This macro is called within an initializer of an | |
6483 | array of structures. The first field in the structure is the name of a | |
6484 | predicate and the second field is an array of rtl codes. For each | |
6485 | predicate, list all rtl codes that can be in expressions matched by the | |
6486 | predicate. The list should have a trailing comma. Here is an example | |
6487 | of two entries in the list for a typical RISC machine: | |
6488 | ||
6489 | @smallexample | |
6490 | #define PREDICATE_CODES \ | |
6491 | @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \ | |
6492 | @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@}, | |
6493 | @end smallexample | |
6494 | ||
6495 | Defining this macro does not affect the generated code (however, | |
6496 | incorrect definitions that omit an rtl code that may be matched by the | |
6497 | predicate can cause the compiler to malfunction). Instead, it allows | |
6498 | the table built by @file{genrecog} to be more compact and efficient, | |
6499 | thus speeding up the compiler. The most important predicates to include | |
6500 | in the list specified by this macro are thoses used in the most insn | |
6501 | patterns. | |
6502 | ||
6503 | @findex CASE_VECTOR_MODE | |
6504 | @item CASE_VECTOR_MODE | |
6505 | An alias for a machine mode name. This is the machine mode that | |
6506 | elements of a jump-table should have. | |
6507 | ||
6508 | @findex CASE_VECTOR_PC_RELATIVE | |
6509 | @item CASE_VECTOR_PC_RELATIVE | |
6510 | Define this macro if jump-tables should contain relative addresses. | |
6511 | ||
6512 | @findex CASE_DROPS_THROUGH | |
6513 | @item CASE_DROPS_THROUGH | |
6514 | Define this if control falls through a @code{case} insn when the index | |
6515 | value is out of range. This means the specified default-label is | |
6516 | actually ignored by the @code{case} insn proper. | |
6517 | ||
6518 | @findex CASE_VALUES_THRESHOLD | |
6519 | @item CASE_VALUES_THRESHOLD | |
6520 | Define this to be the smallest number of different values for which it | |
6521 | is best to use a jump-table instead of a tree of conditional branches. | |
6522 | The default is four for machines with a @code{casesi} instruction and | |
6523 | five otherwise. This is best for most machines. | |
6524 | ||
6525 | @findex WORD_REGISTER_OPERATIONS | |
6526 | @item WORD_REGISTER_OPERATIONS | |
6527 | Define this macro if operations between registers with integral mode | |
6528 | smaller than a word are always performed on the entire register. | |
6529 | Most RISC machines have this property and most CISC machines do not. | |
6530 | ||
6531 | @findex LOAD_EXTEND_OP | |
6532 | @item LOAD_EXTEND_OP (@var{mode}) | |
6533 | Define this macro to be a C expression indicating when insns that read | |
6534 | memory in @var{mode}, an integral mode narrower than a word, set the | |
6535 | bits outside of @var{mode} to be either the sign-extension or the | |
6536 | zero-extension of the data read. Return @code{SIGN_EXTEND} for values | |
6537 | of @var{mode} for which the | |
6538 | insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and | |
6539 | @code{NIL} for other modes. | |
6540 | ||
6541 | This macro is not called with @var{mode} non-integral or with a width | |
6542 | greater than or equal to @code{BITS_PER_WORD}, so you may return any | |
6543 | value in this case. Do not define this macro if it would always return | |
6544 | @code{NIL}. On machines where this macro is defined, you will normally | |
6545 | define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}. | |
6546 | ||
6547 | @findex IMPLICIT_FIX_EXPR | |
6548 | @item IMPLICIT_FIX_EXPR | |
6549 | An alias for a tree code that should be used by default for conversion | |
6550 | of floating point values to fixed point. Normally, | |
6551 | @code{FIX_ROUND_EXPR} is used.@refill | |
6552 | ||
6553 | @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC | |
6554 | @item FIXUNS_TRUNC_LIKE_FIX_TRUNC | |
6555 | Define this macro if the same instructions that convert a floating | |
6556 | point number to a signed fixed point number also convert validly to an | |
6557 | unsigned one. | |
6558 | ||
6559 | @findex EASY_DIV_EXPR | |
6560 | @item EASY_DIV_EXPR | |
6561 | An alias for a tree code that is the easiest kind of division to | |
6562 | compile code for in the general case. It may be | |
6563 | @code{TRUNC_DIV_EXPR}, @code{FLOOR_DIV_EXPR}, @code{CEIL_DIV_EXPR} or | |
6564 | @code{ROUND_DIV_EXPR}. These four division operators differ in how | |
6565 | they round the result to an integer. @code{EASY_DIV_EXPR} is used | |
6566 | when it is permissible to use any of those kinds of division and the | |
6567 | choice should be made on the basis of efficiency.@refill | |
6568 | ||
6569 | @findex MOVE_MAX | |
6570 | @item MOVE_MAX | |
6571 | The maximum number of bytes that a single instruction can move quickly | |
6572 | between memory and registers or between two memory locations. | |
6573 | ||
6574 | @findex MAX_MOVE_MAX | |
6575 | @item MAX_MOVE_MAX | |
6576 | The maximum number of bytes that a single instruction can move quickly | |
6577 | between memory and registers or between two memory locations. If this | |
6578 | is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the | |
6579 | constant value that is the largest value that @code{MOVE_MAX} can have | |
6580 | at run-time. | |
6581 | ||
6582 | @findex SHIFT_COUNT_TRUNCATED | |
6583 | @item SHIFT_COUNT_TRUNCATED | |
6584 | A C expression that is nonzero if on this machine the number of bits | |
6585 | actually used for the count of a shift operation is equal to the number | |
6586 | of bits needed to represent the size of the object being shifted. When | |
6587 | this macro is non-zero, the compiler will assume that it is safe to omit | |
6588 | a sign-extend, zero-extend, and certain bitwise `and' instructions that | |
6589 | truncates the count of a shift operation. On machines that have | |
6590 | instructions that act on bitfields at variable positions, which may | |
6591 | include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED} | |
6592 | also enables deletion of truncations of the values that serve as | |
6593 | arguments to bitfield instructions. | |
6594 | ||
6595 | If both types of instructions truncate the count (for shifts) and | |
6596 | position (for bitfield operations), or if no variable-position bitfield | |
6597 | instructions exist, you should define this macro. | |
6598 | ||
6599 | However, on some machines, such as the 80386 and the 680x0, truncation | |
6600 | only applies to shift operations and not the (real or pretended) | |
6601 | bitfield operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on | |
6602 | such machines. Instead, add patterns to the @file{md} file that include | |
6603 | the implied truncation of the shift instructions. | |
6604 | ||
6605 | You need not define this macro if it would always have the value of zero. | |
6606 | ||
6607 | @findex TRULY_NOOP_TRUNCATION | |
6608 | @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec}) | |
6609 | A C expression which is nonzero if on this machine it is safe to | |
6610 | ``convert'' an integer of @var{inprec} bits to one of @var{outprec} | |
6611 | bits (where @var{outprec} is smaller than @var{inprec}) by merely | |
6612 | operating on it as if it had only @var{outprec} bits. | |
6613 | ||
6614 | On many machines, this expression can be 1. | |
6615 | ||
6616 | @c rearranged this, removed the phrase "it is reported that". this was | |
6617 | @c to fix an overfull hbox. --mew 10feb93 | |
6618 | When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for | |
6619 | modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result. | |
6620 | If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in | |
6621 | such cases may improve things. | |
6622 | ||
6623 | @findex STORE_FLAG_VALUE | |
6624 | @item STORE_FLAG_VALUE | |
6625 | A C expression describing the value returned by a comparison operator | |
6626 | with an integral mode and stored by a store-flag instruction | |
6627 | (@samp{s@var{cond}}) when the condition is true. This description must | |
6628 | apply to @emph{all} the @samp{s@var{cond}} patterns and all the | |
6629 | comparison operators whose results have a @code{MODE_INT} mode. | |
6630 | ||
6631 | A value of 1 or -1 means that the instruction implementing the | |
6632 | comparison operator returns exactly 1 or -1 when the comparison is true | |
6633 | and 0 when the comparison is false. Otherwise, the value indicates | |
6634 | which bits of the result are guaranteed to be 1 when the comparison is | |
6635 | true. This value is interpreted in the mode of the comparison | |
6636 | operation, which is given by the mode of the first operand in the | |
6637 | @samp{s@var{cond}} pattern. Either the low bit or the sign bit of | |
6638 | @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by | |
6639 | the compiler. | |
6640 | ||
6641 | If @code{STORE_FLAG_VALUE} is neither 1 or -1, the compiler will | |
6642 | generate code that depends only on the specified bits. It can also | |
6643 | replace comparison operators with equivalent operations if they cause | |
6644 | the required bits to be set, even if the remaining bits are undefined. | |
6645 | For example, on a machine whose comparison operators return an | |
6646 | @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as | |
6647 | @samp{0x80000000}, saying that just the sign bit is relevant, the | |
6648 | expression | |
6649 | ||
6650 | @smallexample | |
6651 | (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0)) | |
6652 | @end smallexample | |
6653 | ||
6654 | @noindent | |
6655 | can be converted to | |
6656 | ||
6657 | @smallexample | |
6658 | (ashift:SI @var{x} (const_int @var{n})) | |
6659 | @end smallexample | |
6660 | ||
6661 | @noindent | |
6662 | where @var{n} is the appropriate shift count to move the bit being | |
6663 | tested into the sign bit. | |
6664 | ||
6665 | There is no way to describe a machine that always sets the low-order bit | |
6666 | for a true value, but does not guarantee the value of any other bits, | |
6667 | but we do not know of any machine that has such an instruction. If you | |
6668 | are trying to port GNU CC to such a machine, include an instruction to | |
6669 | perform a logical-and of the result with 1 in the pattern for the | |
6670 | comparison operators and let us know | |
6671 | @ifset USING | |
6672 | (@pxref{Bug Reporting,,How to Report Bugs}). | |
6673 | @end ifset | |
6674 | @ifclear USING | |
6675 | (@pxref{Bug Reporting,,How to Report Bugs,gcc.info,Using GCC}). | |
6676 | @end ifclear | |
6677 | ||
6678 | Often, a machine will have multiple instructions that obtain a value | |
6679 | from a comparison (or the condition codes). Here are rules to guide the | |
6680 | choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions | |
6681 | to be used: | |
6682 | ||
6683 | @itemize @bullet | |
6684 | @item | |
6685 | Use the shortest sequence that yields a valid definition for | |
6686 | @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to | |
6687 | ``normalize'' the value (convert it to, e.g., 1 or 0) than for the | |
6688 | comparison operators to do so because there may be opportunities to | |
6689 | combine the normalization with other operations. | |
6690 | ||
6691 | @item | |
6692 | For equal-length sequences, use a value of 1 or -1, with -1 being | |
6693 | slightly preferred on machines with expensive jumps and 1 preferred on | |
6694 | other machines. | |
6695 | ||
6696 | @item | |
6697 | As a second choice, choose a value of @samp{0x80000001} if instructions | |
6698 | exist that set both the sign and low-order bits but do not define the | |
6699 | others. | |
6700 | ||
6701 | @item | |
6702 | Otherwise, use a value of @samp{0x80000000}. | |
6703 | @end itemize | |
6704 | ||
6705 | Many machines can produce both the value chosen for | |
6706 | @code{STORE_FLAG_VALUE} and its negation in the same number of | |
6707 | instructions. On those machines, you should also define a pattern for | |
6708 | those cases, e.g., one matching | |
6709 | ||
6710 | @smallexample | |
6711 | (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C}))) | |
6712 | @end smallexample | |
6713 | ||
6714 | Some machines can also perform @code{and} or @code{plus} operations on | |
6715 | condition code values with less instructions than the corresponding | |
6716 | @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those | |
6717 | machines, define the appropriate patterns. Use the names @code{incscc} | |
6718 | and @code{decscc}, respectively, for the patterns which perform | |
6719 | @code{plus} or @code{minus} operations on condition code values. See | |
6720 | @file{rs6000.md} for some examples. The GNU Superoptizer can be used to | |
6721 | find such instruction sequences on other machines. | |
6722 | ||
6723 | You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag | |
6724 | instructions. | |
6725 | ||
6726 | @findex FLOAT_STORE_FLAG_VALUE | |
6727 | @item FLOAT_STORE_FLAG_VALUE | |
6728 | A C expression that gives a non-zero floating point value that is | |
6729 | returned when comparison operators with floating-point results are true. | |
6730 | Define this macro on machine that have comparison operations that return | |
6731 | floating-point values. If there are no such operations, do not define | |
6732 | this macro. | |
6733 | ||
6734 | @findex Pmode | |
6735 | @item Pmode | |
6736 | An alias for the machine mode for pointers. On most machines, define | |
6737 | this to be the integer mode corresponding to the width of a hardware | |
6738 | pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines. | |
6739 | On some machines you must define this to be one of the partial integer | |
6740 | modes, such as @code{PSImode}. | |
6741 | ||
6742 | The width of @code{Pmode} must be at least as large as the value of | |
6743 | @code{POINTER_SIZE}. If it is not equal, you must define the macro | |
6744 | @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended | |
6745 | to @code{Pmode}. | |
6746 | ||
6747 | @findex FUNCTION_MODE | |
6748 | @item FUNCTION_MODE | |
6749 | An alias for the machine mode used for memory references to functions | |
6750 | being called, in @code{call} RTL expressions. On most machines this | |
6751 | should be @code{QImode}. | |
6752 | ||
6753 | @findex INTEGRATE_THRESHOLD | |
6754 | @item INTEGRATE_THRESHOLD (@var{decl}) | |
6755 | A C expression for the maximum number of instructions above which the | |
6756 | function @var{decl} should not be inlined. @var{decl} is a | |
6757 | @code{FUNCTION_DECL} node. | |
6758 | ||
6759 | The default definition of this macro is 64 plus 8 times the number of | |
6760 | arguments that the function accepts. Some people think a larger | |
6761 | threshold should be used on RISC machines. | |
6762 | ||
6763 | @findex SCCS_DIRECTIVE | |
6764 | @item SCCS_DIRECTIVE | |
6765 | Define this if the preprocessor should ignore @code{#sccs} directives | |
6766 | and print no error message. | |
6767 | ||
6768 | @findex NO_IMPLICIT_EXTERN_C | |
6769 | @item NO_IMPLICIT_EXTERN_C | |
6770 | Define this macro if the system header files support C++ as well as C. | |
6771 | This macro inhibits the usual method of using system header files in | |
6772 | C++, which is to pretend that the file's contents are enclosed in | |
6773 | @samp{extern "C" @{@dots{}@}}. | |
6774 | ||
6775 | @findex HANDLE_PRAGMA | |
6776 | @findex #pragma | |
6777 | @findex pragma | |
6778 | @item HANDLE_PRAGMA (@var{stream}, @var{node}) | |
6779 | Define this macro if you want to implement any pragmas. If defined, it | |
6780 | is a C expression whose value is 1 if the pragma was handled by the function. | |
6781 | The argument @var{stream} is the stdio input stream from which the source text | |
6782 | can be read. @var{node} is the tree node for the identifier after the | |
6783 | @code{#pragma}. | |
6784 | ||
6785 | It is generally a bad idea to implement new uses of @code{#pragma}. The | |
6786 | only reason to define this macro is for compatibility with other | |
6787 | compilers that do support @code{#pragma} for the sake of any user | |
6788 | programs which already use it. | |
6789 | ||
6790 | @findex VALID_MACHINE_DECL_ATTRIBUTE | |
6791 | @item VALID_MACHINE_DECL_ATTRIBUTE (@var{decl}, @var{attributes}, @var{identifier}, @var{args}) | |
6792 | If defined, a C expression whose value is nonzero if @var{identifier} with | |
6793 | arguments @var{args} is a valid machine specific attribute for @var{decl}. | |
6794 | The attributes in @var{attributes} have previously been assigned to @var{decl}. | |
6795 | ||
6796 | @findex VALID_MACHINE_TYPE_ATTRIBUTE | |
6797 | @item VALID_MACHINE_TYPE_ATTRIBUTE (@var{type}, @var{attributes}, @var{identifier}, @var{args}) | |
6798 | If defined, a C expression whose value is nonzero if @var{identifier} with | |
6799 | arguments @var{args} is a valid machine specific attribute for @var{type}. | |
6800 | The attributes in @var{attributes} have previously been assigned to @var{type}. | |
6801 | ||
6802 | @findex COMP_TYPE_ATTRIBUTES | |
6803 | @item COMP_TYPE_ATTRIBUTES (@var{type1}, @var{type2}) | |
6804 | If defined, a C expression whose value is zero if the attributes on | |
6805 | @var{type1} and @var{type2} are incompatible, one if they are compatible, | |
6806 | and two if they are nearly compatible (which causes a warning to be | |
6807 | generated). | |
6808 | ||
6809 | @findex SET_DEFAULT_TYPE_ATTRIBUTES | |
6810 | @item SET_DEFAULT_TYPE_ATTRIBUTES (@var{type}) | |
6811 | If defined, a C statement that assigns default attributes to | |
6812 | newly defined @var{type}. | |
6813 | ||
6814 | @findex DOLLARS_IN_IDENTIFIERS | |
6815 | @item DOLLARS_IN_IDENTIFIERS | |
6816 | Define this macro to control use of the character @samp{$} in identifier | |
37d13a29 | 6817 | names. 0 means @samp{$} is not allowed by default; 1 means it is allowed. |
feca2ed3 | 6818 | 1 is the default; there is no need to define this macro in that case. |
37d13a29 | 6819 | This macro controls the compiler proper; it does not affect the preprocessor. |
feca2ed3 JW |
6820 | |
6821 | @findex NO_DOLLAR_IN_LABEL | |
6822 | @item NO_DOLLAR_IN_LABEL | |
6823 | Define this macro if the assembler does not accept the character | |
6824 | @samp{$} in label names. By default constructors and destructors in | |
6825 | G++ have @samp{$} in the identifiers. If this macro is defined, | |
6826 | @samp{.} is used instead. | |
6827 | ||
6828 | @findex NO_DOT_IN_LABEL | |
6829 | @item NO_DOT_IN_LABEL | |
6830 | Define this macro if the assembler does not accept the character | |
6831 | @samp{.} in label names. By default constructors and destructors in G++ | |
6832 | have names that use @samp{.}. If this macro is defined, these names | |
6833 | are rewritten to avoid @samp{.}. | |
6834 | ||
6835 | @findex DEFAULT_MAIN_RETURN | |
6836 | @item DEFAULT_MAIN_RETURN | |
6837 | Define this macro if the target system expects every program's @code{main} | |
6838 | function to return a standard ``success'' value by default (if no other | |
6839 | value is explicitly returned). | |
6840 | ||
6841 | The definition should be a C statement (sans semicolon) to generate the | |
6842 | appropriate rtl instructions. It is used only when compiling the end of | |
6843 | @code{main}. | |
6844 | ||
6845 | @item HAVE_ATEXIT | |
6846 | @findex HAVE_ATEXIT | |
6847 | Define this if the target system supports the function | |
6848 | @code{atexit} from the ANSI C standard. If this is not defined, | |
6849 | and @code{INIT_SECTION_ASM_OP} is not defined, a default | |
6850 | @code{exit} function will be provided to support C++. | |
6851 | ||
6852 | @item EXIT_BODY | |
6853 | @findex EXIT_BODY | |
6854 | Define this if your @code{exit} function needs to do something | |
6855 | besides calling an external function @code{_cleanup} before | |
6856 | terminating with @code{_exit}. The @code{EXIT_BODY} macro is | |
6857 | only needed if netiher @code{HAVE_ATEXIT} nor | |
6858 | @code{INIT_SECTION_ASM_OP} are defined. | |
6859 | ||
6860 | @findex INSN_SETS_ARE_DELAYED | |
6861 | @item INSN_SETS_ARE_DELAYED (@var{insn}) | |
6862 | Define this macro as a C expression that is nonzero if it is safe for the | |
6863 | delay slot scheduler to place instructions in the delay slot of @var{insn}, | |
6864 | even if they appear to use a resource set or clobbered in @var{insn}. | |
6865 | @var{insn} is always a @code{jump_insn} or an @code{insn}; GNU CC knows that | |
6866 | every @code{call_insn} has this behavior. On machines where some @code{insn} | |
6867 | or @code{jump_insn} is really a function call and hence has this behavior, | |
6868 | you should define this macro. | |
6869 | ||
6870 | You need not define this macro if it would always return zero. | |
6871 | ||
6872 | @findex INSN_REFERENCES_ARE_DELAYED | |
6873 | @item INSN_REFERENCES_ARE_DELAYED (@var{insn}) | |
6874 | Define this macro as a C expression that is nonzero if it is safe for the | |
6875 | delay slot scheduler to place instructions in the delay slot of @var{insn}, | |
6876 | even if they appear to set or clobber a resource referenced in @var{insn}. | |
6877 | @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where | |
6878 | some @code{insn} or @code{jump_insn} is really a function call and its operands | |
6879 | are registers whose use is actually in the subroutine it calls, you should | |
6880 | define this macro. Doing so allows the delay slot scheduler to move | |
6881 | instructions which copy arguments into the argument registers into the delay | |
6882 | slot of @var{insn}. | |
6883 | ||
6884 | You need not define this macro if it would always return zero. | |
6885 | ||
6886 | @findex MACHINE_DEPENDENT_REORG | |
6887 | @item MACHINE_DEPENDENT_REORG (@var{insn}) | |
6888 | In rare cases, correct code generation requires extra machine | |
6889 | dependent processing between the second jump optimization pass and | |
6890 | delayed branch scheduling. On those machines, define this macro as a C | |
6891 | statement to act on the code starting at @var{insn}. | |
6892 | ||
6893 | @findex GIV_SORT_CRITERION | |
6894 | @item GIV_SORT_CRITERION (@var{giv1}, @var{giv2}) | |
6895 | In some cases, the strength reduction optimization pass can produce better | |
6896 | code if this is defined. This macro controls the order that induction | |
6897 | variables are combined. This macro is particularly useful if the target has | |
6898 | limited addressing modes. For instance, the SH target has only positive | |
6899 | offsets in addresses. Thus sorting to put the smallest address first | |
6900 | allows the most combinations to be found. | |
6901 | ||
6902 | @end table |