]>
Commit | Line | Data |
---|---|---|
a8154559 | 1 | @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,2002 |
fed3cef0 | 2 | @c Free Software Foundation, Inc. |
feca2ed3 JW |
3 | @c This is part of the GCC manual. |
4 | @c For copying conditions, see the file gcc.texi. | |
5 | ||
6 | @node Target Macros | |
672a6f42 | 7 | @chapter Target Description Macros and Functions |
feca2ed3 JW |
8 | @cindex machine description macros |
9 | @cindex target description macros | |
10 | @cindex macros, target description | |
11 | @cindex @file{tm.h} macros | |
12 | ||
13 | In addition to the file @file{@var{machine}.md}, a machine description | |
14 | includes a C header file conventionally given the name | |
672a6f42 NB |
15 | @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}. |
16 | The header file defines numerous macros that convey the information | |
17 | about the target machine that does not fit into the scheme of the | |
18 | @file{.md} file. The file @file{tm.h} should be a link to | |
19 | @file{@var{machine}.h}. The header file @file{config.h} includes | |
20 | @file{tm.h} and most compiler source files include @file{config.h}. The | |
648c546a | 21 | source file defines a variable @code{targetm}, which is a structure |
672a6f42 NB |
22 | containing pointers to functions and data relating to the target |
23 | machine. @file{@var{machine}.c} should also contain their definitions, | |
24 | if they are not defined elsewhere in GCC, and other functions called | |
25 | through the macros defined in the @file{.h} file. | |
feca2ed3 JW |
26 | |
27 | @menu | |
648c546a | 28 | * Target Structure:: The @code{targetm} variable. |
feca2ed3 | 29 | * Driver:: Controlling how the driver runs the compilation passes. |
630d3d5a | 30 | * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}. |
414c4dc4 | 31 | * Per-Function Data:: Defining data structures for per-function information. |
feca2ed3 JW |
32 | * Storage Layout:: Defining sizes and alignments of data. |
33 | * Type Layout:: Defining sizes and properties of basic user data types. | |
b2b263e1 | 34 | * Escape Sequences:: Defining the value of target character escape sequences |
feca2ed3 JW |
35 | * Registers:: Naming and describing the hardware registers. |
36 | * Register Classes:: Defining the classes of hardware registers. | |
37 | * Stack and Calling:: Defining which way the stack grows and by how much. | |
38 | * Varargs:: Defining the varargs macros. | |
39 | * Trampolines:: Code set up at run time to enter a nested function. | |
40 | * Library Calls:: Controlling how library routines are implicitly called. | |
41 | * Addressing Modes:: Defining addressing modes valid for memory operands. | |
42 | * Condition Code:: Defining how insns update the condition code. | |
43 | * Costs:: Defining relative costs of different operations. | |
c237e94a | 44 | * Scheduling:: Adjusting the behavior of the instruction scheduler. |
feca2ed3 JW |
45 | * Sections:: Dividing storage into text, data, and other sections. |
46 | * PIC:: Macros for position independent code. | |
47 | * Assembler Format:: Defining how to write insns and pseudo-ops to output. | |
48 | * Debugging Info:: Defining the format of debugging output. | |
b216cd4a | 49 | * Floating Point:: Handling floating point for cross-compilers. |
9f09b1f2 | 50 | * Mode Switching:: Insertion of mode-switching instructions. |
91d231cb | 51 | * Target Attributes:: Defining target-specific uses of @code{__attribute__}. |
d604bca3 | 52 | * MIPS Coprocessors:: MIPS coprocessor support and how to customize it. |
feca2ed3 JW |
53 | * Misc:: Everything else. |
54 | @end menu | |
55 | ||
672a6f42 | 56 | @node Target Structure |
648c546a | 57 | @section The Global @code{targetm} Variable |
672a6f42 NB |
58 | @cindex target hooks |
59 | @cindex target functions | |
60 | ||
f6897b10 SS |
61 | @deftypevar {struct gcc_target} targetm |
62 | The target @file{.c} file must define the global @code{targetm} variable | |
672a6f42 NB |
63 | which contains pointers to functions and data relating to the target |
64 | machine. The variable is declared in @file{target.h}; | |
65 | @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is | |
66 | used to initialize the variable, and macros for the default initializers | |
67 | for elements of the structure. The @file{.c} file should override those | |
68 | macros for which the default definition is inappropriate. For example: | |
69 | @smallexample | |
70 | #include "target.h" | |
71 | #include "target-def.h" | |
72 | ||
73 | /* @r{Initialize the GCC target structure.} */ | |
74 | ||
91d231cb JM |
75 | #undef TARGET_COMP_TYPE_ATTRIBUTES |
76 | #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes | |
672a6f42 | 77 | |
f6897b10 | 78 | struct gcc_target targetm = TARGET_INITIALIZER; |
672a6f42 NB |
79 | @end smallexample |
80 | @end deftypevar | |
81 | ||
82 | Where a macro should be defined in the @file{.c} file in this manner to | |
648c546a | 83 | form part of the @code{targetm} structure, it is documented below as a |
672a6f42 NB |
84 | ``Target Hook'' with a prototype. Many macros will change in future |
85 | from being defined in the @file{.h} file to being part of the | |
648c546a | 86 | @code{targetm} structure. |
672a6f42 | 87 | |
feca2ed3 JW |
88 | @node Driver |
89 | @section Controlling the Compilation Driver, @file{gcc} | |
90 | @cindex driver | |
91 | @cindex controlling the compilation driver | |
92 | ||
93 | @c prevent bad page break with this line | |
94 | You can control the compilation driver. | |
95 | ||
96 | @table @code | |
97 | @findex SWITCH_TAKES_ARG | |
98 | @item SWITCH_TAKES_ARG (@var{char}) | |
630d3d5a | 99 | A C expression which determines whether the option @option{-@var{char}} |
feca2ed3 JW |
100 | takes arguments. The value should be the number of arguments that |
101 | option takes--zero, for many options. | |
102 | ||
103 | By default, this macro is defined as | |
104 | @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options | |
105 | properly. You need not define @code{SWITCH_TAKES_ARG} unless you | |
106 | wish to add additional options which take arguments. Any redefinition | |
107 | should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for | |
108 | additional options. | |
109 | ||
110 | @findex WORD_SWITCH_TAKES_ARG | |
111 | @item WORD_SWITCH_TAKES_ARG (@var{name}) | |
630d3d5a | 112 | A C expression which determines whether the option @option{-@var{name}} |
feca2ed3 JW |
113 | takes arguments. The value should be the number of arguments that |
114 | option takes--zero, for many options. This macro rather than | |
115 | @code{SWITCH_TAKES_ARG} is used for multi-character option names. | |
116 | ||
117 | By default, this macro is defined as | |
118 | @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options | |
119 | properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you | |
120 | wish to add additional options which take arguments. Any redefinition | |
121 | should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for | |
122 | additional options. | |
123 | ||
88117d44 NC |
124 | @findex SWITCH_CURTAILS_COMPILATION |
125 | @item SWITCH_CURTAILS_COMPILATION (@var{char}) | |
630d3d5a | 126 | A C expression which determines whether the option @option{-@var{char}} |
88117d44 | 127 | stops compilation before the generation of an executable. The value is |
df2a54e9 | 128 | boolean, nonzero if the option does stop an executable from being |
88117d44 NC |
129 | generated, zero otherwise. |
130 | ||
131 | By default, this macro is defined as | |
132 | @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard | |
133 | options properly. You need not define | |
134 | @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional | |
135 | options which affect the generation of an executable. Any redefinition | |
136 | should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check | |
137 | for additional options. | |
138 | ||
feca2ed3 JW |
139 | @findex SWITCHES_NEED_SPACES |
140 | @item SWITCHES_NEED_SPACES | |
141 | A string-valued C expression which enumerates the options for which | |
142 | the linker needs a space between the option and its argument. | |
143 | ||
144 | If this macro is not defined, the default value is @code{""}. | |
145 | ||
0259b07a DD |
146 | @findex TARGET_OPTION_TRANSLATE_TABLE |
147 | @item TARGET_OPTION_TRANSLATE_TABLE | |
148 | If defined, a list of pairs of strings, the first of which is a | |
149 | potential command line target to the @file{gcc} driver program, and the | |
150 | second of which is a space-separated (tabs and other whitespace are not | |
151 | supported) list of options with which to replace the first option. The | |
152 | target defining this list is responsible for assuring that the results | |
153 | are valid. Replacement options may not be the @code{--opt} style, they | |
154 | must be the @code{-opt} style. It is the intention of this macro to | |
155 | provide a mechanism for substitution that affects the multilibs chosen, | |
156 | such as one option that enables many options, some of which select | |
157 | multilibs. Example nonsensical definition, where @code{-malt-abi}, | |
158 | @code{-EB}, and @code{-mspoo} cause different multilibs to be chosen: | |
159 | ||
478c9e72 | 160 | @smallexample |
0259b07a DD |
161 | #define TARGET_OPTION_TRANSLATE_TABLE \ |
162 | @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \ | |
163 | @{ "-compat", "-EB -malign=4 -mspoo" @} | |
478c9e72 | 164 | @end smallexample |
0259b07a | 165 | |
feca2ed3 JW |
166 | @findex CPP_SPEC |
167 | @item CPP_SPEC | |
a3a15b4d | 168 | A C string constant that tells the GCC driver program options to |
161d7b59 JM |
169 | pass to CPP@. It can also specify how to translate options you |
170 | give to GCC into options for GCC to pass to the CPP@. | |
feca2ed3 JW |
171 | |
172 | Do not define this macro if it does not need to do anything. | |
173 | ||
a9374841 MM |
174 | @findex CPLUSPLUS_CPP_SPEC |
175 | @item CPLUSPLUS_CPP_SPEC | |
176 | This macro is just like @code{CPP_SPEC}, but is used for C++, rather | |
161d7b59 | 177 | than C@. If you do not define this macro, then the value of |
a9374841 MM |
178 | @code{CPP_SPEC} (if any) will be used instead. |
179 | ||
feca2ed3 JW |
180 | @findex CC1_SPEC |
181 | @item CC1_SPEC | |
a3a15b4d | 182 | A C string constant that tells the GCC driver program options to |
66519c70 JL |
183 | pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language |
184 | front ends. | |
a3a15b4d | 185 | It can also specify how to translate options you give to GCC into options |
630d3d5a | 186 | for GCC to pass to front ends. |
feca2ed3 JW |
187 | |
188 | Do not define this macro if it does not need to do anything. | |
189 | ||
190 | @findex CC1PLUS_SPEC | |
191 | @item CC1PLUS_SPEC | |
a3a15b4d | 192 | A C string constant that tells the GCC driver program options to |
feca2ed3 | 193 | pass to @code{cc1plus}. It can also specify how to translate options you |
a3a15b4d | 194 | give to GCC into options for GCC to pass to the @code{cc1plus}. |
feca2ed3 JW |
195 | |
196 | Do not define this macro if it does not need to do anything. | |
1d96e5b4 FF |
197 | Note that everything defined in CC1_SPEC is already passed to |
198 | @code{cc1plus} so there is no need to duplicate the contents of | |
161d7b59 | 199 | CC1_SPEC in CC1PLUS_SPEC@. |
feca2ed3 JW |
200 | |
201 | @findex ASM_SPEC | |
202 | @item ASM_SPEC | |
a3a15b4d | 203 | A C string constant that tells the GCC driver program options to |
feca2ed3 | 204 | pass to the assembler. It can also specify how to translate options |
a3a15b4d | 205 | you give to GCC into options for GCC to pass to the assembler. |
feca2ed3 JW |
206 | See the file @file{sun3.h} for an example of this. |
207 | ||
208 | Do not define this macro if it does not need to do anything. | |
209 | ||
210 | @findex ASM_FINAL_SPEC | |
211 | @item ASM_FINAL_SPEC | |
a3a15b4d | 212 | A C string constant that tells the GCC driver program how to |
feca2ed3 JW |
213 | run any programs which cleanup after the normal assembler. |
214 | Normally, this is not needed. See the file @file{mips.h} for | |
215 | an example of this. | |
216 | ||
217 | Do not define this macro if it does not need to do anything. | |
218 | ||
219 | @findex LINK_SPEC | |
220 | @item LINK_SPEC | |
a3a15b4d | 221 | A C string constant that tells the GCC driver program options to |
feca2ed3 | 222 | pass to the linker. It can also specify how to translate options you |
a3a15b4d | 223 | give to GCC into options for GCC to pass to the linker. |
feca2ed3 JW |
224 | |
225 | Do not define this macro if it does not need to do anything. | |
226 | ||
227 | @findex LIB_SPEC | |
228 | @item LIB_SPEC | |
229 | Another C string constant used much like @code{LINK_SPEC}. The difference | |
230 | between the two is that @code{LIB_SPEC} is used at the end of the | |
231 | command given to the linker. | |
232 | ||
233 | If this macro is not defined, a default is provided that | |
234 | loads the standard C library from the usual place. See @file{gcc.c}. | |
235 | ||
236 | @findex LIBGCC_SPEC | |
237 | @item LIBGCC_SPEC | |
a3a15b4d | 238 | Another C string constant that tells the GCC driver program |
feca2ed3 JW |
239 | how and when to place a reference to @file{libgcc.a} into the |
240 | linker command line. This constant is placed both before and after | |
241 | the value of @code{LIB_SPEC}. | |
242 | ||
a3a15b4d | 243 | If this macro is not defined, the GCC driver provides a default that |
630d3d5a | 244 | passes the string @option{-lgcc} to the linker. |
feca2ed3 JW |
245 | |
246 | @findex STARTFILE_SPEC | |
247 | @item STARTFILE_SPEC | |
248 | Another C string constant used much like @code{LINK_SPEC}. The | |
249 | difference between the two is that @code{STARTFILE_SPEC} is used at | |
250 | the very beginning of the command given to the linker. | |
251 | ||
252 | If this macro is not defined, a default is provided that loads the | |
253 | standard C startup file from the usual place. See @file{gcc.c}. | |
254 | ||
255 | @findex ENDFILE_SPEC | |
256 | @item ENDFILE_SPEC | |
257 | Another C string constant used much like @code{LINK_SPEC}. The | |
258 | difference between the two is that @code{ENDFILE_SPEC} is used at | |
259 | the very end of the command given to the linker. | |
260 | ||
261 | Do not define this macro if it does not need to do anything. | |
262 | ||
008355a6 AO |
263 | @findex THREAD_MODEL_SPEC |
264 | @item THREAD_MODEL_SPEC | |
265 | GCC @code{-v} will print the thread model GCC was configured to use. | |
266 | However, this doesn't work on platforms that are multilibbed on thread | |
267 | models, such as AIX 4.3. On such platforms, define | |
268 | @code{THREAD_MODEL_SPEC} such that it evaluates to a string without | |
269 | blanks that names one of the recognized thread models. @code{%*}, the | |
270 | default value of this macro, will expand to the value of | |
271 | @code{thread_file} set in @file{config.gcc}. | |
272 | ||
feca2ed3 JW |
273 | @findex EXTRA_SPECS |
274 | @item EXTRA_SPECS | |
275 | Define this macro to provide additional specifications to put in the | |
276 | @file{specs} file that can be used in various specifications like | |
277 | @code{CC1_SPEC}. | |
278 | ||
279 | The definition should be an initializer for an array of structures, | |
280 | containing a string constant, that defines the specification name, and a | |
281 | string constant that provides the specification. | |
282 | ||
283 | Do not define this macro if it does not need to do anything. | |
284 | ||
285 | @code{EXTRA_SPECS} is useful when an architecture contains several | |
630d3d5a | 286 | related targets, which have various @code{@dots{}_SPECS} which are similar |
feca2ed3 JW |
287 | to each other, and the maintainer would like one central place to keep |
288 | these definitions. | |
289 | ||
290 | For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to | |
291 | define either @code{_CALL_SYSV} when the System V calling sequence is | |
292 | used or @code{_CALL_AIX} when the older AIX-based calling sequence is | |
293 | used. | |
294 | ||
295 | The @file{config/rs6000/rs6000.h} target file defines: | |
296 | ||
297 | @example | |
298 | #define EXTRA_SPECS \ | |
299 | @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @}, | |
300 | ||
301 | #define CPP_SYS_DEFAULT "" | |
302 | @end example | |
303 | ||
304 | The @file{config/rs6000/sysv.h} target file defines: | |
305 | @smallexample | |
306 | #undef CPP_SPEC | |
307 | #define CPP_SPEC \ | |
308 | "%@{posix: -D_POSIX_SOURCE @} \ | |
309 | %@{mcall-sysv: -D_CALL_SYSV @} %@{mcall-aix: -D_CALL_AIX @} \ | |
310 | %@{!mcall-sysv: %@{!mcall-aix: %(cpp_sysv_default) @}@} \ | |
311 | %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}" | |
312 | ||
313 | #undef CPP_SYSV_DEFAULT | |
314 | #define CPP_SYSV_DEFAULT "-D_CALL_SYSV" | |
315 | @end smallexample | |
316 | ||
317 | while the @file{config/rs6000/eabiaix.h} target file defines | |
318 | @code{CPP_SYSV_DEFAULT} as: | |
319 | ||
320 | @smallexample | |
321 | #undef CPP_SYSV_DEFAULT | |
322 | #define CPP_SYSV_DEFAULT "-D_CALL_AIX" | |
323 | @end smallexample | |
324 | ||
325 | @findex LINK_LIBGCC_SPECIAL | |
326 | @item LINK_LIBGCC_SPECIAL | |
327 | Define this macro if the driver program should find the library | |
630d3d5a | 328 | @file{libgcc.a} itself and should not pass @option{-L} options to the |
feca2ed3 | 329 | linker. If you do not define this macro, the driver program will pass |
630d3d5a JM |
330 | the argument @option{-lgcc} to tell the linker to do the search and will |
331 | pass @option{-L} options to it. | |
feca2ed3 JW |
332 | |
333 | @findex LINK_LIBGCC_SPECIAL_1 | |
334 | @item LINK_LIBGCC_SPECIAL_1 | |
335 | Define this macro if the driver program should find the library | |
336 | @file{libgcc.a}. If you do not define this macro, the driver program will pass | |
630d3d5a | 337 | the argument @option{-lgcc} to tell the linker to do the search. |
feca2ed3 | 338 | This macro is similar to @code{LINK_LIBGCC_SPECIAL}, except that it does |
630d3d5a | 339 | not affect @option{-L} options. |
feca2ed3 | 340 | |
bbd7687d DM |
341 | @findex LINK_GCC_C_SEQUENCE_SPEC |
342 | @item LINK_GCC_C_SEQUENCE_SPEC | |
343 | The sequence in which libgcc and libc are specified to the linker. | |
344 | By default this is @code{%G %L %G}. | |
345 | ||
9ec36da5 JL |
346 | @findex LINK_COMMAND_SPEC |
347 | @item LINK_COMMAND_SPEC | |
348 | A C string constant giving the complete command line need to execute the | |
349 | linker. When you do this, you will need to update your port each time a | |
350 | change is made to the link command line within @file{gcc.c}. Therefore, | |
351 | define this macro only if you need to completely redefine the command | |
352 | line for invoking the linker and there is no other way to accomplish | |
bbd7687d DM |
353 | the effect you need. Overriding this macro may be avoidable by overriding |
354 | @code{LINK_GCC_C_SEQUENCE_SPEC} instead. | |
9ec36da5 | 355 | |
5897739e JO |
356 | @findex LINK_ELIMINATE_DUPLICATE_LDIRECTORIES |
357 | @item LINK_ELIMINATE_DUPLICATE_LDIRECTORIES | |
2cc07db4 | 358 | A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search |
5897739e JO |
359 | directories from linking commands. Do not give it a nonzero value if |
360 | removing duplicate search directories changes the linker's semantics. | |
361 | ||
feca2ed3 JW |
362 | @findex MULTILIB_DEFAULTS |
363 | @item MULTILIB_DEFAULTS | |
364 | Define this macro as a C expression for the initializer of an array of | |
365 | string to tell the driver program which options are defaults for this | |
366 | target and thus do not need to be handled specially when using | |
367 | @code{MULTILIB_OPTIONS}. | |
368 | ||
369 | Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in | |
370 | the target makefile fragment or if none of the options listed in | |
371 | @code{MULTILIB_OPTIONS} are set by default. | |
372 | @xref{Target Fragment}. | |
373 | ||
374 | @findex RELATIVE_PREFIX_NOT_LINKDIR | |
375 | @item RELATIVE_PREFIX_NOT_LINKDIR | |
376 | Define this macro to tell @code{gcc} that it should only translate | |
630d3d5a | 377 | a @option{-B} prefix into a @option{-L} linker option if the prefix |
feca2ed3 JW |
378 | indicates an absolute file name. |
379 | ||
380 | @findex STANDARD_EXEC_PREFIX | |
381 | @item STANDARD_EXEC_PREFIX | |
382 | Define this macro as a C string constant if you wish to override the | |
383 | standard choice of @file{/usr/local/lib/gcc-lib/} as the default prefix to | |
384 | try when searching for the executable files of the compiler. | |
385 | ||
386 | @findex MD_EXEC_PREFIX | |
387 | @item MD_EXEC_PREFIX | |
388 | If defined, this macro is an additional prefix to try after | |
389 | @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched | |
630d3d5a | 390 | when the @option{-b} option is used, or the compiler is built as a cross |
5505263f JL |
391 | compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it |
392 | to the list of directories used to find the assembler in @file{configure.in}. | |
feca2ed3 JW |
393 | |
394 | @findex STANDARD_STARTFILE_PREFIX | |
395 | @item STANDARD_STARTFILE_PREFIX | |
396 | Define this macro as a C string constant if you wish to override the | |
397 | standard choice of @file{/usr/local/lib/} as the default prefix to | |
398 | try when searching for startup files such as @file{crt0.o}. | |
399 | ||
400 | @findex MD_STARTFILE_PREFIX | |
401 | @item MD_STARTFILE_PREFIX | |
402 | If defined, this macro supplies an additional prefix to try after the | |
403 | standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the | |
630d3d5a | 404 | @option{-b} option is used, or when the compiler is built as a cross |
feca2ed3 JW |
405 | compiler. |
406 | ||
407 | @findex MD_STARTFILE_PREFIX_1 | |
408 | @item MD_STARTFILE_PREFIX_1 | |
409 | If defined, this macro supplies yet another prefix to try after the | |
630d3d5a | 410 | standard prefixes. It is not searched when the @option{-b} option is |
feca2ed3 JW |
411 | used, or when the compiler is built as a cross compiler. |
412 | ||
413 | @findex INIT_ENVIRONMENT | |
414 | @item INIT_ENVIRONMENT | |
e9a25f70 | 415 | Define this macro as a C string constant if you wish to set environment |
feca2ed3 JW |
416 | variables for programs called by the driver, such as the assembler and |
417 | loader. The driver passes the value of this macro to @code{putenv} to | |
418 | initialize the necessary environment variables. | |
419 | ||
420 | @findex LOCAL_INCLUDE_DIR | |
421 | @item LOCAL_INCLUDE_DIR | |
422 | Define this macro as a C string constant if you wish to override the | |
423 | standard choice of @file{/usr/local/include} as the default prefix to | |
424 | try when searching for local header files. @code{LOCAL_INCLUDE_DIR} | |
425 | comes before @code{SYSTEM_INCLUDE_DIR} in the search order. | |
426 | ||
c237e94a ZW |
427 | Cross compilers do not search either @file{/usr/local/include} or its |
428 | replacement. | |
feca2ed3 | 429 | |
dc36ec2c RK |
430 | @findex MODIFY_TARGET_NAME |
431 | @item MODIFY_TARGET_NAME | |
432 | Define this macro if you with to define command-line switches that modify the | |
433 | default target name | |
434 | ||
435 | For each switch, you can include a string to be appended to the first | |
436 | part of the configuration name or a string to be deleted from the | |
437 | configuration name, if present. The definition should be an initializer | |
438 | for an array of structures. Each array element should have three | |
439 | elements: the switch name (a string constant, including the initial | |
440 | dash), one of the enumeration codes @code{ADD} or @code{DELETE} to | |
441 | indicate whether the string should be inserted or deleted, and the string | |
442 | to be inserted or deleted (a string constant). | |
443 | ||
444 | For example, on a machine where @samp{64} at the end of the | |
630d3d5a JM |
445 | configuration name denotes a 64-bit target and you want the @option{-32} |
446 | and @option{-64} switches to select between 32- and 64-bit targets, you would | |
dc36ec2c RK |
447 | code |
448 | ||
449 | @smallexample | |
450 | #define MODIFY_TARGET_NAME \ | |
451 | @{ @{ "-32", DELETE, "64"@}, \ | |
452 | @{"-64", ADD, "64"@}@} | |
453 | @end smallexample | |
454 | ||
455 | ||
feca2ed3 JW |
456 | @findex SYSTEM_INCLUDE_DIR |
457 | @item SYSTEM_INCLUDE_DIR | |
458 | Define this macro as a C string constant if you wish to specify a | |
459 | system-specific directory to search for header files before the standard | |
460 | directory. @code{SYSTEM_INCLUDE_DIR} comes before | |
461 | @code{STANDARD_INCLUDE_DIR} in the search order. | |
462 | ||
463 | Cross compilers do not use this macro and do not search the directory | |
464 | specified. | |
465 | ||
466 | @findex STANDARD_INCLUDE_DIR | |
467 | @item STANDARD_INCLUDE_DIR | |
468 | Define this macro as a C string constant if you wish to override the | |
469 | standard choice of @file{/usr/include} as the default prefix to | |
470 | try when searching for header files. | |
471 | ||
472 | Cross compilers do not use this macro and do not search either | |
473 | @file{/usr/include} or its replacement. | |
474 | ||
e9a25f70 JL |
475 | @findex STANDARD_INCLUDE_COMPONENT |
476 | @item STANDARD_INCLUDE_COMPONENT | |
477 | The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}. | |
478 | See @code{INCLUDE_DEFAULTS}, below, for the description of components. | |
479 | If you do not define this macro, no component is used. | |
480 | ||
feca2ed3 JW |
481 | @findex INCLUDE_DEFAULTS |
482 | @item INCLUDE_DEFAULTS | |
483 | Define this macro if you wish to override the entire default search path | |
e9a25f70 JL |
484 | for include files. For a native compiler, the default search path |
485 | usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR}, | |
feca2ed3 JW |
486 | @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and |
487 | @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR} | |
488 | and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile}, | |
161d7b59 | 489 | and specify private search areas for GCC@. The directory |
feca2ed3 JW |
490 | @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs. |
491 | ||
492 | The definition should be an initializer for an array of structures. | |
e9a25f70 | 493 | Each array element should have four elements: the directory name (a |
9f6dc500 HPN |
494 | string constant), the component name (also a string constant), a flag |
495 | for C++-only directories, | |
e9a25f70 JL |
496 | and a flag showing that the includes in the directory don't need to be |
497 | wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of | |
498 | the array with a null element. | |
499 | ||
500 | The component name denotes what GNU package the include file is part of, | |
501 | if any, in all upper-case letters. For example, it might be @samp{GCC} | |
9f6dc500 | 502 | or @samp{BINUTILS}. If the package is part of a vendor-supplied |
e9a25f70 JL |
503 | operating system, code the component name as @samp{0}. |
504 | ||
e9a25f70 | 505 | For example, here is the definition used for VAX/VMS: |
feca2ed3 JW |
506 | |
507 | @example | |
508 | #define INCLUDE_DEFAULTS \ | |
509 | @{ \ | |
e9a25f70 JL |
510 | @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \ |
511 | @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \ | |
512 | @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \ | |
513 | @{ ".", 0, 0, 0@}, \ | |
514 | @{ 0, 0, 0, 0@} \ | |
feca2ed3 JW |
515 | @} |
516 | @end example | |
517 | @end table | |
518 | ||
519 | Here is the order of prefixes tried for exec files: | |
520 | ||
521 | @enumerate | |
522 | @item | |
630d3d5a | 523 | Any prefixes specified by the user with @option{-B}. |
feca2ed3 JW |
524 | |
525 | @item | |
526 | The environment variable @code{GCC_EXEC_PREFIX}, if any. | |
527 | ||
528 | @item | |
529 | The directories specified by the environment variable @code{COMPILER_PATH}. | |
530 | ||
531 | @item | |
532 | The macro @code{STANDARD_EXEC_PREFIX}. | |
533 | ||
534 | @item | |
535 | @file{/usr/lib/gcc/}. | |
536 | ||
537 | @item | |
538 | The macro @code{MD_EXEC_PREFIX}, if any. | |
539 | @end enumerate | |
540 | ||
541 | Here is the order of prefixes tried for startfiles: | |
542 | ||
543 | @enumerate | |
544 | @item | |
630d3d5a | 545 | Any prefixes specified by the user with @option{-B}. |
feca2ed3 JW |
546 | |
547 | @item | |
548 | The environment variable @code{GCC_EXEC_PREFIX}, if any. | |
549 | ||
550 | @item | |
551 | The directories specified by the environment variable @code{LIBRARY_PATH} | |
512b62fb | 552 | (or port-specific name; native only, cross compilers do not use this). |
feca2ed3 JW |
553 | |
554 | @item | |
555 | The macro @code{STANDARD_EXEC_PREFIX}. | |
556 | ||
557 | @item | |
558 | @file{/usr/lib/gcc/}. | |
559 | ||
560 | @item | |
561 | The macro @code{MD_EXEC_PREFIX}, if any. | |
562 | ||
563 | @item | |
564 | The macro @code{MD_STARTFILE_PREFIX}, if any. | |
565 | ||
566 | @item | |
567 | The macro @code{STANDARD_STARTFILE_PREFIX}. | |
568 | ||
569 | @item | |
570 | @file{/lib/}. | |
571 | ||
572 | @item | |
573 | @file{/usr/lib/}. | |
574 | @end enumerate | |
575 | ||
576 | @node Run-time Target | |
577 | @section Run-time Target Specification | |
578 | @cindex run-time target specification | |
579 | @cindex predefined macros | |
580 | @cindex target specifications | |
581 | ||
582 | @c prevent bad page break with this line | |
583 | Here are run-time target specifications. | |
584 | ||
585 | @table @code | |
12a41c22 NB |
586 | @findex TARGET_CPU_CPP_BUILTINS |
587 | @item TARGET_CPU_CPP_BUILTINS() | |
588 | This function-like macro expands to a block of code that defines | |
589 | built-in preprocessor macros and assertions for the target cpu, using | |
1f95326c | 590 | the functions @code{builtin_define}, @code{builtin_define_std} and |
81a75f0f | 591 | @code{builtin_assert} defined in @file{c-common.c}. When the front end |
12a41c22 NB |
592 | calls this macro it provides a trailing semicolon, and since it has |
593 | finished command line option processing your code can use those | |
594 | results freely. | |
3df89291 NB |
595 | |
596 | @code{builtin_assert} takes a string in the form you pass to the | |
597 | command-line option @option{-A}, such as @code{cpu=mips}, and creates | |
d90a95fb | 598 | the assertion. @code{builtin_define} takes a string in the form |
3df89291 NB |
599 | accepted by option @option{-D} and unconditionally defines the macro. |
600 | ||
d90a95fb | 601 | @code{builtin_define_std} takes a string representing the name of an |
3df89291 | 602 | object-like macro. If it doesn't lie in the user's namespace, |
d90a95fb | 603 | @code{builtin_define_std} defines it unconditionally. Otherwise, it |
3df89291 NB |
604 | defines a version with two leading underscores, and another version |
605 | with two leading and trailing underscores, and defines the original | |
606 | only if an ISO standard was not requested on the command line. For | |
607 | example, passing @code{unix} defines @code{__unix}, @code{__unix__} | |
608 | and possibly @code{unix}; passing @code{_mips} defines @code{__mips}, | |
609 | @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64} | |
610 | defines only @code{_ABI64}. | |
611 | ||
e0322d5c NB |
612 | You can also test for the C dialect being compiled. The variable |
613 | @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus} | |
614 | or @code{clk_objective_c}. Note that if we are preprocessing | |
615 | assembler, this variable will be @code{clk_c} but the function-like | |
616 | macro @code{preprocessing_asm_p()} will return true, so you might want | |
ce3649d2 EC |
617 | to check for that first. If you need to check for strict ANSI, the |
618 | variable @code{flag_iso} can be used. | |
e0322d5c | 619 | |
12a41c22 NB |
620 | With @code{TARGET_OS_CPP_BUILTINS} this macro obsoletes the |
621 | @code{CPP_PREDEFINES} target macro. | |
622 | ||
623 | @findex TARGET_OS_CPP_BUILTINS | |
624 | @item TARGET_OS_CPP_BUILTINS() | |
625 | Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional | |
626 | and is used for the target operating system instead. | |
627 | ||
628 | With @code{TARGET_CPU_CPP_BUILTINS} this macro obsoletes the | |
629 | @code{CPP_PREDEFINES} target macro. | |
3df89291 | 630 | |
feca2ed3 JW |
631 | @findex CPP_PREDEFINES |
632 | @item CPP_PREDEFINES | |
630d3d5a | 633 | Define this to be a string constant containing @option{-D} options to |
feca2ed3 | 634 | define the predefined macros that identify this machine and system. |
5490d604 JM |
635 | These macros will be predefined unless the @option{-ansi} option (or a |
636 | @option{-std} option for strict ISO C conformance) is specified. | |
feca2ed3 JW |
637 | |
638 | In addition, a parallel set of macros are predefined, whose names are | |
639 | made by appending @samp{__} at the beginning and at the end. These | |
5490d604 JM |
640 | @samp{__} macros are permitted by the ISO standard, so they are |
641 | predefined regardless of whether @option{-ansi} or a @option{-std} option | |
642 | is specified. | |
feca2ed3 JW |
643 | |
644 | For example, on the Sun, one can use the following value: | |
645 | ||
646 | @smallexample | |
647 | "-Dmc68000 -Dsun -Dunix" | |
648 | @end smallexample | |
649 | ||
650 | The result is to define the macros @code{__mc68000__}, @code{__sun__} | |
651 | and @code{__unix__} unconditionally, and the macros @code{mc68000}, | |
630d3d5a | 652 | @code{sun} and @code{unix} provided @option{-ansi} is not specified. |
feca2ed3 JW |
653 | |
654 | @findex extern int target_flags | |
655 | @item extern int target_flags; | |
656 | This declaration should be present. | |
657 | ||
658 | @cindex optional hardware or system features | |
659 | @cindex features, optional, in system conventions | |
660 | @item TARGET_@dots{} | |
661 | This series of macros is to allow compiler command arguments to | |
662 | enable or disable the use of optional features of the target machine. | |
663 | For example, one machine description serves both the 68000 and | |
664 | the 68020; a command argument tells the compiler whether it should | |
665 | use 68020-only instructions or not. This command argument works | |
666 | by means of a macro @code{TARGET_68020} that tests a bit in | |
667 | @code{target_flags}. | |
668 | ||
669 | Define a macro @code{TARGET_@var{featurename}} for each such option. | |
9f6dc500 HPN |
670 | Its definition should test a bit in @code{target_flags}. It is |
671 | recommended that a helper macro @code{TARGET_MASK_@var{featurename}} | |
672 | is defined for each bit-value to test, and used in | |
673 | @code{TARGET_@var{featurename}} and @code{TARGET_SWITCHES}. For | |
674 | example: | |
feca2ed3 JW |
675 | |
676 | @smallexample | |
9f6dc500 HPN |
677 | #define TARGET_MASK_68020 1 |
678 | #define TARGET_68020 (target_flags & TARGET_MASK_68020) | |
feca2ed3 JW |
679 | @end smallexample |
680 | ||
681 | One place where these macros are used is in the condition-expressions | |
682 | of instruction patterns. Note how @code{TARGET_68020} appears | |
683 | frequently in the 68000 machine description file, @file{m68k.md}. | |
684 | Another place they are used is in the definitions of the other | |
685 | macros in the @file{@var{machine}.h} file. | |
686 | ||
687 | @findex TARGET_SWITCHES | |
688 | @item TARGET_SWITCHES | |
689 | This macro defines names of command options to set and clear | |
690 | bits in @code{target_flags}. Its definition is an initializer | |
691 | with a subgrouping for each command option. | |
692 | ||
693 | Each subgrouping contains a string constant, that defines the option | |
b8468bc7 NC |
694 | name, a number, which contains the bits to set in |
695 | @code{target_flags}, and a second string which is the description | |
561c1ae1 | 696 | displayed by @option{--help}. If the number is negative then the bits specified |
b8468bc7 NC |
697 | by the number are cleared instead of being set. If the description |
698 | string is present but empty, then no help information will be displayed | |
699 | for that option, but it will not count as an undocumented option. The | |
700 | actual option name is made by appending @samp{-m} to the specified name. | |
561c1ae1 | 701 | Non-empty description strings should be marked with @code{N_(@dots{})} for |
404ae494 PT |
702 | @command{xgettext}. Please do not mark empty strings because the empty |
703 | string is reserved by GNU gettext. @code{gettext("")} returns the header entry | |
704 | of the message catalog with meta information, not the empty string. | |
705 | ||
706 | In addition to the description for @option{--help}, | |
561c1ae1 JM |
707 | more detailed documentation for each option should be added to |
708 | @file{invoke.texi}. | |
feca2ed3 JW |
709 | |
710 | One of the subgroupings should have a null string. The number in | |
711 | this grouping is the default value for @code{target_flags}. Any | |
712 | target options act starting with that value. | |
713 | ||
630d3d5a | 714 | Here is an example which defines @option{-m68000} and @option{-m68020} |
feca2ed3 JW |
715 | with opposite meanings, and picks the latter as the default: |
716 | ||
717 | @smallexample | |
718 | #define TARGET_SWITCHES \ | |
9f6dc500 | 719 | @{ @{ "68020", TARGET_MASK_68020, "" @}, \ |
561c1ae1 JM |
720 | @{ "68000", -TARGET_MASK_68020, \ |
721 | N_("Compile for the 68000") @}, \ | |
9f6dc500 | 722 | @{ "", TARGET_MASK_68020, "" @}@} |
feca2ed3 JW |
723 | @end smallexample |
724 | ||
725 | @findex TARGET_OPTIONS | |
726 | @item TARGET_OPTIONS | |
727 | This macro is similar to @code{TARGET_SWITCHES} but defines names of command | |
728 | options that have values. Its definition is an initializer with a | |
729 | subgrouping for each command option. | |
730 | ||
731 | Each subgrouping contains a string constant, that defines the fixed part | |
404ae494 PT |
732 | of the option name, the address of a variable, and a description string. |
733 | Non-empty description strings should be marked with @code{N_(@dots{})} for | |
734 | @command{xgettext}. Please do not mark empty strings because the empty | |
735 | string is reserved by GNU gettext. @code{gettext("")} returns the header entry | |
736 | of the message catalog with meta information, not the empty string. | |
737 | ||
b8468bc7 NC |
738 | The variable, type @code{char *}, is set to the variable part of the |
739 | given option if the fixed part matches. The actual option name is made | |
561c1ae1 JM |
740 | by appending @samp{-m} to the specified name. Again, each option should |
741 | also be documented in @file{invoke.texi}. | |
feca2ed3 | 742 | |
630d3d5a JM |
743 | Here is an example which defines @option{-mshort-data-@var{number}}. If the |
744 | given option is @option{-mshort-data-512}, the variable @code{m88k_short_data} | |
feca2ed3 JW |
745 | will be set to the string @code{"512"}. |
746 | ||
747 | @smallexample | |
748 | extern char *m88k_short_data; | |
749 | #define TARGET_OPTIONS \ | |
561c1ae1 JM |
750 | @{ @{ "short-data-", &m88k_short_data, \ |
751 | N_("Specify the size of the short data section") @} @} | |
feca2ed3 JW |
752 | @end smallexample |
753 | ||
754 | @findex TARGET_VERSION | |
755 | @item TARGET_VERSION | |
756 | This macro is a C statement to print on @code{stderr} a string | |
757 | describing the particular machine description choice. Every machine | |
758 | description should define @code{TARGET_VERSION}. For example: | |
759 | ||
760 | @smallexample | |
761 | #ifdef MOTOROLA | |
762 | #define TARGET_VERSION \ | |
763 | fprintf (stderr, " (68k, Motorola syntax)"); | |
764 | #else | |
765 | #define TARGET_VERSION \ | |
766 | fprintf (stderr, " (68k, MIT syntax)"); | |
767 | #endif | |
768 | @end smallexample | |
769 | ||
770 | @findex OVERRIDE_OPTIONS | |
771 | @item OVERRIDE_OPTIONS | |
772 | Sometimes certain combinations of command options do not make sense on | |
773 | a particular target machine. You can define a macro | |
774 | @code{OVERRIDE_OPTIONS} to take account of this. This macro, if | |
775 | defined, is executed once just after all the command options have been | |
776 | parsed. | |
777 | ||
778 | Don't use this macro to turn on various extra optimizations for | |
630d3d5a | 779 | @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for. |
feca2ed3 JW |
780 | |
781 | @findex OPTIMIZATION_OPTIONS | |
c6aded7c | 782 | @item OPTIMIZATION_OPTIONS (@var{level}, @var{size}) |
feca2ed3 JW |
783 | Some machines may desire to change what optimizations are performed for |
784 | various optimization levels. This macro, if defined, is executed once | |
785 | just after the optimization level is determined and before the remainder | |
786 | of the command options have been parsed. Values set in this macro are | |
787 | used as the default values for the other command line options. | |
788 | ||
630d3d5a JM |
789 | @var{level} is the optimization level specified; 2 if @option{-O2} is |
790 | specified, 1 if @option{-O} is specified, and 0 if neither is specified. | |
feca2ed3 | 791 | |
df2a54e9 | 792 | @var{size} is nonzero if @option{-Os} is specified and zero otherwise. |
c6aded7c | 793 | |
feca2ed3 JW |
794 | You should not use this macro to change options that are not |
795 | machine-specific. These should uniformly selected by the same | |
796 | optimization level on all supported machines. Use this macro to enable | |
797 | machine-specific optimizations. | |
798 | ||
799 | @strong{Do not examine @code{write_symbols} in | |
800 | this macro!} The debugging options are not supposed to alter the | |
801 | generated code. | |
802 | ||
803 | @findex CAN_DEBUG_WITHOUT_FP | |
804 | @item CAN_DEBUG_WITHOUT_FP | |
805 | Define this macro if debugging can be performed even without a frame | |
a3a15b4d | 806 | pointer. If this macro is defined, GCC will turn on the |
630d3d5a | 807 | @option{-fomit-frame-pointer} option whenever @option{-O} is specified. |
feca2ed3 JW |
808 | @end table |
809 | ||
414c4dc4 NC |
810 | @node Per-Function Data |
811 | @section Defining data structures for per-function information. | |
812 | @cindex per-function data | |
813 | @cindex data structures | |
814 | ||
815 | If the target needs to store information on a per-function basis, GCC | |
816 | provides a macro and a couple of variables to allow this. Note, just | |
817 | using statics to store the information is a bad idea, since GCC supports | |
818 | nested functions, so you can be halfway through encoding one function | |
819 | when another one comes along. | |
820 | ||
821 | GCC defines a data structure called @code{struct function} which | |
822 | contains all of the data specific to an individual function. This | |
823 | structure contains a field called @code{machine} whose type is | |
824 | @code{struct machine_function *}, which can be used by targets to point | |
825 | to their own specific data. | |
826 | ||
827 | If a target needs per-function specific data it should define the type | |
e2500fed GK |
828 | @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}. |
829 | This macro should be used to initialize the function pointer | |
830 | @code{init_machine_status}. This pointer is explained below. | |
414c4dc4 NC |
831 | |
832 | One typical use of per-function, target specific data is to create an | |
833 | RTX to hold the register containing the function's return address. This | |
834 | RTX can then be used to implement the @code{__builtin_return_address} | |
835 | function, for level 0. | |
836 | ||
aee96fe9 | 837 | Note---earlier implementations of GCC used a single data area to hold |
414c4dc4 NC |
838 | all of the per-function information. Thus when processing of a nested |
839 | function began the old per-function data had to be pushed onto a | |
840 | stack, and when the processing was finished, it had to be popped off the | |
841 | stack. GCC used to provide function pointers called | |
02f52e19 | 842 | @code{save_machine_status} and @code{restore_machine_status} to handle |
414c4dc4 NC |
843 | the saving and restoring of the target specific information. Since the |
844 | single data area approach is no longer used, these pointers are no | |
845 | longer supported. | |
846 | ||
847 | The macro and function pointers are described below. | |
848 | ||
849 | @table @code | |
850 | @findex INIT_EXPANDERS | |
851 | @item INIT_EXPANDERS | |
c21cd8b1 | 852 | Macro called to initialize any target specific information. This macro |
414c4dc4 | 853 | is called once per function, before generation of any RTL has begun. |
c21cd8b1 | 854 | The intention of this macro is to allow the initialization of the |
414c4dc4 NC |
855 | function pointers below. |
856 | ||
857 | @findex init_machine_status | |
858 | @item init_machine_status | |
859 | This is a @code{void (*)(struct function *)} function pointer. If this | |
59d42021 | 860 | pointer is non-@code{NULL} it will be called once per function, before function |
414c4dc4 | 861 | compilation starts, in order to allow the target to perform any target |
c21cd8b1 JM |
862 | specific initialization of the @code{struct function} structure. It is |
863 | intended that this would be used to initialize the @code{machine} of | |
c771326b | 864 | that structure. |
414c4dc4 | 865 | |
e2500fed GK |
866 | @code{struct machine_function} structures are expected to be freed by GC. |
867 | Generally, any memory that they reference must be allocated by using | |
868 | @code{ggc_alloc}, including the structure itself. | |
414c4dc4 NC |
869 | |
870 | @end table | |
871 | ||
feca2ed3 JW |
872 | @node Storage Layout |
873 | @section Storage Layout | |
874 | @cindex storage layout | |
875 | ||
876 | Note that the definitions of the macros in this table which are sizes or | |
877 | alignments measured in bits do not need to be constant. They can be C | |
878 | expressions that refer to static variables, such as the @code{target_flags}. | |
879 | @xref{Run-time Target}. | |
880 | ||
881 | @table @code | |
882 | @findex BITS_BIG_ENDIAN | |
883 | @item BITS_BIG_ENDIAN | |
884 | Define this macro to have the value 1 if the most significant bit in a | |
885 | byte has the lowest number; otherwise define it to have the value zero. | |
886 | This means that bit-field instructions count from the most significant | |
887 | bit. If the machine has no bit-field instructions, then this must still | |
888 | be defined, but it doesn't matter which value it is defined to. This | |
889 | macro need not be a constant. | |
890 | ||
891 | This macro does not affect the way structure fields are packed into | |
892 | bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}. | |
893 | ||
894 | @findex BYTES_BIG_ENDIAN | |
895 | @item BYTES_BIG_ENDIAN | |
896 | Define this macro to have the value 1 if the most significant byte in a | |
897 | word has the lowest number. This macro need not be a constant. | |
898 | ||
899 | @findex WORDS_BIG_ENDIAN | |
900 | @item WORDS_BIG_ENDIAN | |
901 | Define this macro to have the value 1 if, in a multiword object, the | |
902 | most significant word has the lowest number. This applies to both | |
a3a15b4d | 903 | memory locations and registers; GCC fundamentally assumes that the |
feca2ed3 JW |
904 | order of words in memory is the same as the order in registers. This |
905 | macro need not be a constant. | |
906 | ||
907 | @findex LIBGCC2_WORDS_BIG_ENDIAN | |
908 | @item LIBGCC2_WORDS_BIG_ENDIAN | |
aee96fe9 JM |
909 | Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a |
910 | constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be | |
911 | used only when compiling @file{libgcc2.c}. Typically the value will be set | |
feca2ed3 JW |
912 | based on preprocessor defines. |
913 | ||
914 | @findex FLOAT_WORDS_BIG_ENDIAN | |
915 | @item FLOAT_WORDS_BIG_ENDIAN | |
916 | Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or | |
917 | @code{TFmode} floating point numbers are stored in memory with the word | |
918 | containing the sign bit at the lowest address; otherwise define it to | |
919 | have the value 0. This macro need not be a constant. | |
920 | ||
921 | You need not define this macro if the ordering is the same as for | |
922 | multi-word integers. | |
923 | ||
924 | @findex BITS_PER_UNIT | |
925 | @item BITS_PER_UNIT | |
926 | Define this macro to be the number of bits in an addressable storage | |
5c60f03d | 927 | unit (byte). If you do not define this macro the default is 8. |
feca2ed3 JW |
928 | |
929 | @findex BITS_PER_WORD | |
930 | @item BITS_PER_WORD | |
e81dd381 KG |
931 | Number of bits in a word. If you do not define this macro, the default |
932 | is @code{BITS_PER_UNIT * UNITS_PER_WORD}. | |
feca2ed3 JW |
933 | |
934 | @findex MAX_BITS_PER_WORD | |
935 | @item MAX_BITS_PER_WORD | |
936 | Maximum number of bits in a word. If this is undefined, the default is | |
937 | @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the | |
938 | largest value that @code{BITS_PER_WORD} can have at run-time. | |
939 | ||
940 | @findex UNITS_PER_WORD | |
941 | @item UNITS_PER_WORD | |
942 | Number of storage units in a word; normally 4. | |
943 | ||
944 | @findex MIN_UNITS_PER_WORD | |
945 | @item MIN_UNITS_PER_WORD | |
946 | Minimum number of units in a word. If this is undefined, the default is | |
947 | @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the | |
948 | smallest value that @code{UNITS_PER_WORD} can have at run-time. | |
949 | ||
950 | @findex POINTER_SIZE | |
951 | @item POINTER_SIZE | |
952 | Width of a pointer, in bits. You must specify a value no wider than the | |
953 | width of @code{Pmode}. If it is not equal to the width of @code{Pmode}, | |
2465bf76 KG |
954 | you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify |
955 | a value the default is @code{BITS_PER_WORD}. | |
feca2ed3 JW |
956 | |
957 | @findex POINTERS_EXTEND_UNSIGNED | |
958 | @item POINTERS_EXTEND_UNSIGNED | |
6dd12198 | 959 | A C expression whose value is greater than zero if pointers that need to be |
f5963e61 | 960 | extended from being @code{POINTER_SIZE} bits wide to @code{Pmode} are to |
6dd12198 SE |
961 | be zero-extended and zero if they are to be sign-extended. If the value |
962 | is less then zero then there must be an "ptr_extend" instruction that | |
963 | extends a pointer from @code{POINTER_SIZE} to @code{Pmode}. | |
feca2ed3 JW |
964 | |
965 | You need not define this macro if the @code{POINTER_SIZE} is equal | |
966 | to the width of @code{Pmode}. | |
967 | ||
968 | @findex PROMOTE_MODE | |
969 | @item PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type}) | |
970 | A macro to update @var{m} and @var{unsignedp} when an object whose type | |
971 | is @var{type} and which has the specified mode and signedness is to be | |
972 | stored in a register. This macro is only called when @var{type} is a | |
973 | scalar type. | |
974 | ||
975 | On most RISC machines, which only have operations that operate on a full | |
976 | register, define this macro to set @var{m} to @code{word_mode} if | |
977 | @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most | |
978 | cases, only integer modes should be widened because wider-precision | |
979 | floating-point operations are usually more expensive than their narrower | |
980 | counterparts. | |
981 | ||
982 | For most machines, the macro definition does not change @var{unsignedp}. | |
983 | However, some machines, have instructions that preferentially handle | |
984 | either signed or unsigned quantities of certain modes. For example, on | |
985 | the DEC Alpha, 32-bit loads from memory and 32-bit add instructions | |
986 | sign-extend the result to 64 bits. On such machines, set | |
987 | @var{unsignedp} according to which kind of extension is more efficient. | |
988 | ||
989 | Do not define this macro if it would never modify @var{m}. | |
990 | ||
991 | @findex PROMOTE_FUNCTION_ARGS | |
992 | @item PROMOTE_FUNCTION_ARGS | |
993 | Define this macro if the promotion described by @code{PROMOTE_MODE} | |
994 | should also be done for outgoing function arguments. | |
995 | ||
996 | @findex PROMOTE_FUNCTION_RETURN | |
997 | @item PROMOTE_FUNCTION_RETURN | |
998 | Define this macro if the promotion described by @code{PROMOTE_MODE} | |
999 | should also be done for the return value of functions. | |
1000 | ||
1001 | If this macro is defined, @code{FUNCTION_VALUE} must perform the same | |
1002 | promotions done by @code{PROMOTE_MODE}. | |
1003 | ||
1004 | @findex PROMOTE_FOR_CALL_ONLY | |
1005 | @item PROMOTE_FOR_CALL_ONLY | |
1006 | Define this macro if the promotion described by @code{PROMOTE_MODE} | |
1007 | should @emph{only} be performed for outgoing function arguments or | |
1008 | function return values, as specified by @code{PROMOTE_FUNCTION_ARGS} | |
1009 | and @code{PROMOTE_FUNCTION_RETURN}, respectively. | |
1010 | ||
1011 | @findex PARM_BOUNDARY | |
1012 | @item PARM_BOUNDARY | |
1013 | Normal alignment required for function parameters on the stack, in | |
1014 | bits. All stack parameters receive at least this much alignment | |
1015 | regardless of data type. On most machines, this is the same as the | |
1016 | size of an integer. | |
1017 | ||
1018 | @findex STACK_BOUNDARY | |
1019 | @item STACK_BOUNDARY | |
31cdd499 ZW |
1020 | Define this macro to the minimum alignment enforced by hardware for the |
1021 | stack pointer on this machine. The definition is a C expression for the | |
1022 | desired alignment (measured in bits). This value is used as a default | |
1023 | if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines, | |
1024 | this should be the same as @code{PARM_BOUNDARY}. | |
c795bca9 BS |
1025 | |
1026 | @findex PREFERRED_STACK_BOUNDARY | |
1027 | @item PREFERRED_STACK_BOUNDARY | |
31cdd499 ZW |
1028 | Define this macro if you wish to preserve a certain alignment for the |
1029 | stack pointer, greater than what the hardware enforces. The definition | |
1030 | is a C expression for the desired alignment (measured in bits). This | |
1031 | macro must evaluate to a value equal to or larger than | |
1032 | @code{STACK_BOUNDARY}. | |
feca2ed3 | 1033 | |
1d482056 RH |
1034 | @findex FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN |
1035 | @item FORCE_PREFERRED_STACK_BOUNDARY_IN_MAIN | |
1036 | A C expression that evaluates true if @code{PREFERRED_STACK_BOUNDARY} is | |
1037 | not guaranteed by the runtime and we should emit code to align the stack | |
1038 | at the beginning of @code{main}. | |
1039 | ||
c795bca9 | 1040 | @cindex @code{PUSH_ROUNDING}, interaction with @code{PREFERRED_STACK_BOUNDARY} |
feca2ed3 | 1041 | If @code{PUSH_ROUNDING} is not defined, the stack will always be aligned |
c795bca9 BS |
1042 | to the specified boundary. If @code{PUSH_ROUNDING} is defined and specifies |
1043 | a less strict alignment than @code{PREFERRED_STACK_BOUNDARY}, the stack may | |
1044 | be momentarily unaligned while pushing arguments. | |
feca2ed3 JW |
1045 | |
1046 | @findex FUNCTION_BOUNDARY | |
1047 | @item FUNCTION_BOUNDARY | |
1048 | Alignment required for a function entry point, in bits. | |
1049 | ||
1050 | @findex BIGGEST_ALIGNMENT | |
1051 | @item BIGGEST_ALIGNMENT | |
1052 | Biggest alignment that any data type can require on this machine, in bits. | |
1053 | ||
861bb6c1 JL |
1054 | @findex MINIMUM_ATOMIC_ALIGNMENT |
1055 | @item MINIMUM_ATOMIC_ALIGNMENT | |
1056 | If defined, the smallest alignment, in bits, that can be given to an | |
1057 | object that can be referenced in one operation, without disturbing any | |
1058 | nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger | |
1059 | on machines that don't have byte or half-word store operations. | |
1060 | ||
feca2ed3 JW |
1061 | @findex BIGGEST_FIELD_ALIGNMENT |
1062 | @item BIGGEST_FIELD_ALIGNMENT | |
11cf4d18 JJ |
1063 | Biggest alignment that any structure or union field can require on this |
1064 | machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for | |
1065 | structure and union fields only, unless the field alignment has been set | |
1066 | by the @code{__attribute__ ((aligned (@var{n})))} construct. | |
feca2ed3 JW |
1067 | |
1068 | @findex ADJUST_FIELD_ALIGN | |
1069 | @item ADJUST_FIELD_ALIGN (@var{field}, @var{computed}) | |
1070 | An expression for the alignment of a structure field @var{field} if the | |
ad9335eb JJ |
1071 | alignment computed in the usual way (including applying of |
1072 | @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the | |
1073 | alignment) is @var{computed}. It overrides alignment only if the | |
1074 | field alignment has not been set by the | |
1075 | @code{__attribute__ ((aligned (@var{n})))} construct. | |
feca2ed3 JW |
1076 | |
1077 | @findex MAX_OFILE_ALIGNMENT | |
1078 | @item MAX_OFILE_ALIGNMENT | |
1079 | Biggest alignment supported by the object file format of this machine. | |
1080 | Use this macro to limit the alignment which can be specified using the | |
1081 | @code{__attribute__ ((aligned (@var{n})))} construct. If not defined, | |
1082 | the default value is @code{BIGGEST_ALIGNMENT}. | |
1083 | ||
1084 | @findex DATA_ALIGNMENT | |
1085 | @item DATA_ALIGNMENT (@var{type}, @var{basic-align}) | |
a8d1550a | 1086 | If defined, a C expression to compute the alignment for a variable in |
8a198bd2 JW |
1087 | the static store. @var{type} is the data type, and @var{basic-align} is |
1088 | the alignment that the object would ordinarily have. The value of this | |
feca2ed3 JW |
1089 | macro is used instead of that alignment to align the object. |
1090 | ||
1091 | If this macro is not defined, then @var{basic-align} is used. | |
1092 | ||
1093 | @findex strcpy | |
1094 | One use of this macro is to increase alignment of medium-size data to | |
1095 | make it all fit in fewer cache lines. Another is to cause character | |
1096 | arrays to be word-aligned so that @code{strcpy} calls that copy | |
1097 | constants to character arrays can be done inline. | |
1098 | ||
1099 | @findex CONSTANT_ALIGNMENT | |
1100 | @item CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align}) | |
1101 | If defined, a C expression to compute the alignment given to a constant | |
1102 | that is being placed in memory. @var{constant} is the constant and | |
1103 | @var{basic-align} is the alignment that the object would ordinarily | |
1104 | have. The value of this macro is used instead of that alignment to | |
1105 | align the object. | |
1106 | ||
1107 | If this macro is not defined, then @var{basic-align} is used. | |
1108 | ||
1109 | The typical use of this macro is to increase alignment for string | |
1110 | constants to be word aligned so that @code{strcpy} calls that copy | |
1111 | constants can be done inline. | |
1112 | ||
d16790f2 JW |
1113 | @findex LOCAL_ALIGNMENT |
1114 | @item LOCAL_ALIGNMENT (@var{type}, @var{basic-align}) | |
a8d1550a | 1115 | If defined, a C expression to compute the alignment for a variable in |
d16790f2 JW |
1116 | the local store. @var{type} is the data type, and @var{basic-align} is |
1117 | the alignment that the object would ordinarily have. The value of this | |
1118 | macro is used instead of that alignment to align the object. | |
1119 | ||
1120 | If this macro is not defined, then @var{basic-align} is used. | |
1121 | ||
1122 | One use of this macro is to increase alignment of medium-size data to | |
1123 | make it all fit in fewer cache lines. | |
1124 | ||
feca2ed3 JW |
1125 | @findex EMPTY_FIELD_BOUNDARY |
1126 | @item EMPTY_FIELD_BOUNDARY | |
c771326b | 1127 | Alignment in bits to be given to a structure bit-field that follows an |
feca2ed3 JW |
1128 | empty field such as @code{int : 0;}. |
1129 | ||
1130 | Note that @code{PCC_BITFIELD_TYPE_MATTERS} also affects the alignment | |
1131 | that results from an empty field. | |
1132 | ||
1133 | @findex STRUCTURE_SIZE_BOUNDARY | |
1134 | @item STRUCTURE_SIZE_BOUNDARY | |
1135 | Number of bits which any structure or union's size must be a multiple of. | |
1136 | Each structure or union's size is rounded up to a multiple of this. | |
1137 | ||
1138 | If you do not define this macro, the default is the same as | |
1139 | @code{BITS_PER_UNIT}. | |
1140 | ||
1141 | @findex STRICT_ALIGNMENT | |
1142 | @item STRICT_ALIGNMENT | |
1143 | Define this macro to be the value 1 if instructions will fail to work | |
1144 | if given data not on the nominal alignment. If instructions will merely | |
1145 | go slower in that case, define this macro as 0. | |
1146 | ||
1147 | @findex PCC_BITFIELD_TYPE_MATTERS | |
1148 | @item PCC_BITFIELD_TYPE_MATTERS | |
1149 | Define this if you wish to imitate the way many other C compilers handle | |
c771326b | 1150 | alignment of bit-fields and the structures that contain them. |
feca2ed3 | 1151 | |
c771326b | 1152 | The behavior is that the type written for a bit-field (@code{int}, |
feca2ed3 JW |
1153 | @code{short}, or other integer type) imposes an alignment for the |
1154 | entire structure, as if the structure really did contain an ordinary | |
c771326b | 1155 | field of that type. In addition, the bit-field is placed within the |
feca2ed3 JW |
1156 | structure so that it would fit within such a field, not crossing a |
1157 | boundary for it. | |
1158 | ||
c771326b | 1159 | Thus, on most machines, a bit-field whose type is written as @code{int} |
feca2ed3 JW |
1160 | would not cross a four-byte boundary, and would force four-byte |
1161 | alignment for the whole structure. (The alignment used may not be four | |
1162 | bytes; it is controlled by the other alignment parameters.) | |
1163 | ||
1164 | If the macro is defined, its definition should be a C expression; | |
1165 | a nonzero value for the expression enables this behavior. | |
1166 | ||
1167 | Note that if this macro is not defined, or its value is zero, some | |
c771326b | 1168 | bit-fields may cross more than one alignment boundary. The compiler can |
feca2ed3 JW |
1169 | support such references if there are @samp{insv}, @samp{extv}, and |
1170 | @samp{extzv} insns that can directly reference memory. | |
1171 | ||
c771326b | 1172 | The other known way of making bit-fields work is to define |
feca2ed3 JW |
1173 | @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}. |
1174 | Then every structure can be accessed with fullwords. | |
1175 | ||
c771326b | 1176 | Unless the machine has bit-field instructions or you define |
feca2ed3 JW |
1177 | @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define |
1178 | @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value. | |
1179 | ||
a3a15b4d | 1180 | If your aim is to make GCC use the same conventions for laying out |
c771326b | 1181 | bit-fields as are used by another compiler, here is how to investigate |
feca2ed3 JW |
1182 | what the other compiler does. Compile and run this program: |
1183 | ||
1184 | @example | |
1185 | struct foo1 | |
1186 | @{ | |
1187 | char x; | |
1188 | char :0; | |
1189 | char y; | |
1190 | @}; | |
1191 | ||
1192 | struct foo2 | |
1193 | @{ | |
1194 | char x; | |
1195 | int :0; | |
1196 | char y; | |
1197 | @}; | |
1198 | ||
1199 | main () | |
1200 | @{ | |
1201 | printf ("Size of foo1 is %d\n", | |
1202 | sizeof (struct foo1)); | |
1203 | printf ("Size of foo2 is %d\n", | |
1204 | sizeof (struct foo2)); | |
1205 | exit (0); | |
1206 | @} | |
1207 | @end example | |
1208 | ||
1209 | If this prints 2 and 5, then the compiler's behavior is what you would | |
1210 | get from @code{PCC_BITFIELD_TYPE_MATTERS}. | |
1211 | ||
1212 | @findex BITFIELD_NBYTES_LIMITED | |
1213 | @item BITFIELD_NBYTES_LIMITED | |
f913c102 AO |
1214 | Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited |
1215 | to aligning a bit-field within the structure. | |
feca2ed3 | 1216 | |
31a02448 | 1217 | @findex MEMBER_TYPE_FORCES_BLK |
182e515e | 1218 | @item MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode}) |
31a02448 | 1219 | Return 1 if a structure or array containing @var{field} should be accessed using |
9f6dc500 HPN |
1220 | @code{BLKMODE}. |
1221 | ||
182e515e AH |
1222 | If @var{field} is the only field in the structure, @var{mode} is its |
1223 | mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the | |
1224 | case where structures of one field would require the structure's mode to | |
1225 | retain the field's mode. | |
1226 | ||
9f6dc500 HPN |
1227 | Normally, this is not needed. See the file @file{c4x.h} for an example |
1228 | of how to use this macro to prevent a structure having a floating point | |
1229 | field from being accessed in an integer mode. | |
1230 | ||
feca2ed3 | 1231 | @findex ROUND_TYPE_SIZE |
0003feb2 VM |
1232 | @item ROUND_TYPE_SIZE (@var{type}, @var{computed}, @var{specified}) |
1233 | Define this macro as an expression for the overall size of a type | |
1234 | (given by @var{type} as a tree node) when the size computed in the | |
1235 | usual way is @var{computed} and the alignment is @var{specified}. | |
feca2ed3 | 1236 | |
0003feb2 | 1237 | The default is to round @var{computed} up to a multiple of @var{specified}. |
feca2ed3 | 1238 | |
fed3cef0 RK |
1239 | @findex ROUND_TYPE_SIZE_UNIT |
1240 | @item ROUND_TYPE_SIZE_UNIT (@var{type}, @var{computed}, @var{specified}) | |
1241 | Similar to @code{ROUND_TYPE_SIZE}, but sizes and alignments are | |
1242 | specified in units (bytes). If you define @code{ROUND_TYPE_SIZE}, | |
1243 | you must also define this macro and they must be defined consistently | |
1244 | with each other. | |
1245 | ||
feca2ed3 | 1246 | @findex ROUND_TYPE_ALIGN |
0003feb2 VM |
1247 | @item ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified}) |
1248 | Define this macro as an expression for the alignment of a type (given | |
1249 | by @var{type} as a tree node) if the alignment computed in the usual | |
1250 | way is @var{computed} and the alignment explicitly specified was | |
feca2ed3 JW |
1251 | @var{specified}. |
1252 | ||
1253 | The default is to use @var{specified} if it is larger; otherwise, use | |
1254 | the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT} | |
1255 | ||
1256 | @findex MAX_FIXED_MODE_SIZE | |
1257 | @item MAX_FIXED_MODE_SIZE | |
1258 | An integer expression for the size in bits of the largest integer | |
1259 | machine mode that should actually be used. All integer machine modes of | |
1260 | this size or smaller can be used for structures and unions with the | |
1261 | appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE | |
1262 | (DImode)} is assumed. | |
1263 | ||
4061f623 BS |
1264 | @findex VECTOR_MODE_SUPPORTED_P |
1265 | @item VECTOR_MODE_SUPPORTED_P(@var{mode}) | |
1266 | Define this macro to be nonzero if the port is prepared to handle insns | |
1267 | involving vector mode @var{mode}. At the very least, it must have move | |
1268 | patterns for this mode. | |
1269 | ||
73c8090f DE |
1270 | @findex STACK_SAVEAREA_MODE |
1271 | @item STACK_SAVEAREA_MODE (@var{save_level}) | |
1272 | If defined, an expression of type @code{enum machine_mode} that | |
39403d82 DE |
1273 | specifies the mode of the save area operand of a |
1274 | @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}). | |
1275 | @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or | |
1276 | @code{SAVE_NONLOCAL} and selects which of the three named patterns is | |
1277 | having its mode specified. | |
73c8090f DE |
1278 | |
1279 | You need not define this macro if it always returns @code{Pmode}. You | |
1280 | would most commonly define this macro if the | |
1281 | @code{save_stack_@var{level}} patterns need to support both a 32- and a | |
1282 | 64-bit mode. | |
1283 | ||
39403d82 DE |
1284 | @findex STACK_SIZE_MODE |
1285 | @item STACK_SIZE_MODE | |
1286 | If defined, an expression of type @code{enum machine_mode} that | |
1287 | specifies the mode of the size increment operand of an | |
1288 | @code{allocate_stack} named pattern (@pxref{Standard Names}). | |
1289 | ||
1290 | You need not define this macro if it always returns @code{word_mode}. | |
1291 | You would most commonly define this macro if the @code{allocate_stack} | |
1292 | pattern needs to support both a 32- and a 64-bit mode. | |
1293 | ||
feca2ed3 JW |
1294 | @findex TARGET_FLOAT_FORMAT |
1295 | @item TARGET_FLOAT_FORMAT | |
1296 | A code distinguishing the floating point format of the target machine. | |
68eb4fb9 | 1297 | There are five defined values: |
feca2ed3 JW |
1298 | |
1299 | @table @code | |
1300 | @findex IEEE_FLOAT_FORMAT | |
1301 | @item IEEE_FLOAT_FORMAT | |
1302 | This code indicates IEEE floating point. It is the default; there is no | |
161d7b59 | 1303 | need to define this macro when the format is IEEE@. |
feca2ed3 JW |
1304 | |
1305 | @findex VAX_FLOAT_FORMAT | |
1306 | @item VAX_FLOAT_FORMAT | |
4226378a PK |
1307 | This code indicates the ``F float'' (for @code{float}) and ``D float'' |
1308 | or ``G float'' formats (for @code{double}) used on the VAX and PDP-11@. | |
feca2ed3 | 1309 | |
68eb4fb9 LB |
1310 | @findex IBM_FLOAT_FORMAT |
1311 | @item IBM_FLOAT_FORMAT | |
1312 | This code indicates the format used on the IBM System/370. | |
1313 | ||
1314 | @findex C4X_FLOAT_FORMAT | |
1315 | @item C4X_FLOAT_FORMAT | |
1316 | This code indicates the format used on the TMS320C3x/C4x. | |
1317 | ||
feca2ed3 JW |
1318 | @findex UNKNOWN_FLOAT_FORMAT |
1319 | @item UNKNOWN_FLOAT_FORMAT | |
1320 | This code indicates any other format. | |
1321 | @end table | |
1322 | ||
4226378a | 1323 | If any other |
807633e5 ZW |
1324 | formats are actually in use on supported machines, new codes should be |
1325 | defined for them. | |
feca2ed3 JW |
1326 | |
1327 | The ordering of the component words of floating point values stored in | |
807633e5 | 1328 | memory is controlled by @code{FLOAT_WORDS_BIG_ENDIAN}. |
e9a25f70 | 1329 | |
71925bc0 RS |
1330 | @findex MODE_HAS_NANS |
1331 | @item MODE_HAS_NANS (@var{mode}) | |
1332 | When defined, this macro should be true if @var{mode} has a NaN | |
1333 | representation. The compiler assumes that NaNs are not equal to | |
1334 | anything (including themselves) and that addition, subtraction, | |
1335 | multiplication and division all return NaNs when one operand is | |
1336 | NaN@. | |
1337 | ||
1338 | By default, this macro is true if @var{mode} is a floating-point | |
1339 | mode and the target floating-point format is IEEE@. | |
1340 | ||
1341 | @findex MODE_HAS_INFINITIES | |
1342 | @item MODE_HAS_INFINITIES (@var{mode}) | |
1343 | This macro should be true if @var{mode} can represent infinity. At | |
1344 | present, the compiler uses this macro to decide whether @samp{x - x} | |
1345 | is always defined. By default, the macro is true when @var{mode} | |
1346 | is a floating-point mode and the target format is IEEE@. | |
1347 | ||
1348 | @findex MODE_HAS_SIGNED_ZEROS | |
1349 | @item MODE_HAS_SIGNED_ZEROS (@var{mode}) | |
1350 | True if @var{mode} distinguishes between positive and negative zero. | |
1351 | The rules are expected to follow the IEEE standard: | |
1352 | ||
1353 | @itemize @bullet | |
1354 | @item | |
1355 | @samp{x + x} has the same sign as @samp{x}. | |
1356 | ||
1357 | @item | |
1358 | If the sum of two values with opposite sign is zero, the result is | |
1359 | positive for all rounding modes expect towards @minus{}infinity, for | |
1360 | which it is negative. | |
1361 | ||
1362 | @item | |
1363 | The sign of a product or quotient is negative when exactly one | |
1364 | of the operands is negative. | |
1365 | @end itemize | |
1366 | ||
1367 | The default definition is true if @var{mode} is a floating-point | |
1368 | mode and the target format is IEEE@. | |
1369 | ||
1370 | @findex MODE_HAS_SIGN_DEPENDENT_ROUNDING | |
1371 | @item MODE_HAS_SIGN_DEPENDENT_ROUNDING (@var{mode}) | |
1372 | If defined, this macro should be true for @var{mode} if it has at | |
1373 | least one rounding mode in which @samp{x} and @samp{-x} can be | |
1374 | rounded to numbers of different magnitude. Two such modes are | |
1375 | towards @minus{}infinity and towards +infinity. | |
1376 | ||
1377 | The default definition of this macro is true if @var{mode} is | |
1378 | a floating-point mode and the target format is IEEE@. | |
3fcaac1d RS |
1379 | |
1380 | @findex ROUND_TOWARDS_ZERO | |
1381 | @item ROUND_TOWARDS_ZERO | |
1382 | If defined, this macro should be true if the prevailing rounding | |
1383 | mode is towards zero. A true value has the following effects: | |
1384 | ||
1385 | @itemize @bullet | |
1386 | @item | |
1387 | @code{MODE_HAS_SIGN_DEPENDENT_ROUNDING} will be false for all modes. | |
1388 | ||
1389 | @item | |
1390 | @file{libgcc.a}'s floating-point emulator will round towards zero | |
1391 | rather than towards nearest. | |
1392 | ||
1393 | @item | |
1394 | The compiler's floating-point emulator will round towards zero after | |
1395 | doing arithmetic, and when converting from the internal float format to | |
1396 | the target format. | |
1397 | @end itemize | |
1398 | ||
1399 | The macro does not affect the parsing of string literals. When the | |
1400 | primary rounding mode is towards zero, library functions like | |
1401 | @code{strtod} might still round towards nearest, and the compiler's | |
1402 | parser should behave like the target's @code{strtod} where possible. | |
1403 | ||
1404 | Not defining this macro is equivalent to returning zero. | |
1405 | ||
1406 | @findex LARGEST_EXPONENT_IS_NORMAL | |
1407 | @item LARGEST_EXPONENT_IS_NORMAL (@var{size}) | |
4226378a | 1408 | This macro should return true if floats with @var{size} |
3fcaac1d RS |
1409 | bits do not have a NaN or infinity representation, but use the largest |
1410 | exponent for normal numbers instead. | |
1411 | ||
1412 | Defining this macro to true for @var{size} causes @code{MODE_HAS_NANS} | |
1413 | and @code{MODE_HAS_INFINITIES} to be false for @var{size}-bit modes. | |
1414 | It also affects the way @file{libgcc.a} and @file{real.c} emulate | |
1415 | floating-point arithmetic. | |
1416 | ||
1417 | The default definition of this macro returns false for all sizes. | |
feca2ed3 JW |
1418 | @end table |
1419 | ||
f913c102 AO |
1420 | @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type}) |
1421 | This target hook returns @code{true} if bit-fields in the given | |
1422 | @var{record_type} are to be laid out following the rules of Microsoft | |
1423 | Visual C/C++, namely: (i) a bit-field won't share the same storage | |
1424 | unit with the previous bit-field if their underlying types have | |
1425 | different sizes, and the bit-field will be aligned to the highest | |
1426 | alignment of the underlying types of itself and of the previous | |
1427 | bit-field; (ii) a zero-sized bit-field will affect the alignment of | |
1428 | the whole enclosing structure, even if it is unnamed; except that | |
1429 | (iii) a zero-sized bit-field will be disregarded unless it follows | |
1430 | another bit-field of non-zero size. If this hook returns @code{true}, | |
1431 | other macros that control bit-field layout are ignored. | |
e4850f36 DR |
1432 | |
1433 | When a bit-field is inserted into a packed record, the whole size | |
1434 | of the underlying type is used by one or more same-size adjacent | |
1435 | bit-fields (that is, if its long:3, 32 bits is used in the record, | |
1436 | and any additional adjacent long bit-fields are packed into the same | |
1437 | chunk of 32 bits. However, if the size changes, a new field of that | |
1438 | size is allocated). In an unpacked record, this is the same as using | |
1439 | alignment, but not equivalent when packing. | |
1440 | ||
1441 | If both MS bit-fields and @samp{__attribute__((packed))} are used, | |
1442 | the latter will take precedence. If @samp{__attribute__((packed))} is | |
1443 | used on a single field when MS bit-fields are in use, it will take | |
1444 | precedence for that field, but the alignment of the rest of the structure | |
1445 | may affect its placement. | |
f913c102 AO |
1446 | @end deftypefn |
1447 | ||
feca2ed3 JW |
1448 | @node Type Layout |
1449 | @section Layout of Source Language Data Types | |
1450 | ||
1451 | These macros define the sizes and other characteristics of the standard | |
1452 | basic data types used in programs being compiled. Unlike the macros in | |
1453 | the previous section, these apply to specific features of C and related | |
1454 | languages, rather than to fundamental aspects of storage layout. | |
1455 | ||
1456 | @table @code | |
1457 | @findex INT_TYPE_SIZE | |
1458 | @item INT_TYPE_SIZE | |
1459 | A C expression for the size in bits of the type @code{int} on the | |
1460 | target machine. If you don't define this, the default is one word. | |
1461 | ||
feca2ed3 JW |
1462 | @findex SHORT_TYPE_SIZE |
1463 | @item SHORT_TYPE_SIZE | |
1464 | A C expression for the size in bits of the type @code{short} on the | |
1465 | target machine. If you don't define this, the default is half a word. | |
1466 | (If this would be less than one storage unit, it is rounded up to one | |
1467 | unit.) | |
1468 | ||
1469 | @findex LONG_TYPE_SIZE | |
1470 | @item LONG_TYPE_SIZE | |
1471 | A C expression for the size in bits of the type @code{long} on the | |
1472 | target machine. If you don't define this, the default is one word. | |
1473 | ||
1615c261 RK |
1474 | @findex ADA_LONG_TYPE_SIZE |
1475 | @item ADA_LONG_TYPE_SIZE | |
1476 | On some machines, the size used for the Ada equivalent of the type | |
1477 | @code{long} by a native Ada compiler differs from that used by C. In | |
1478 | that situation, define this macro to be a C expression to be used for | |
1479 | the size of that type. If you don't define this, the default is the | |
1480 | value of @code{LONG_TYPE_SIZE}. | |
1481 | ||
feca2ed3 JW |
1482 | @findex MAX_LONG_TYPE_SIZE |
1483 | @item MAX_LONG_TYPE_SIZE | |
1484 | Maximum number for the size in bits of the type @code{long} on the | |
1485 | target machine. If this is undefined, the default is | |
1486 | @code{LONG_TYPE_SIZE}. Otherwise, it is the constant value that is the | |
1487 | largest value that @code{LONG_TYPE_SIZE} can have at run-time. This is | |
1488 | used in @code{cpp}. | |
1489 | ||
1490 | @findex LONG_LONG_TYPE_SIZE | |
1491 | @item LONG_LONG_TYPE_SIZE | |
1492 | A C expression for the size in bits of the type @code{long long} on the | |
1493 | target machine. If you don't define this, the default is two | |
047c1c92 | 1494 | words. If you want to support GNU Ada on your machine, the value of this |
feca2ed3 JW |
1495 | macro must be at least 64. |
1496 | ||
1497 | @findex CHAR_TYPE_SIZE | |
1498 | @item CHAR_TYPE_SIZE | |
1499 | A C expression for the size in bits of the type @code{char} on the | |
c294bd99 HPN |
1500 | target machine. If you don't define this, the default is |
1501 | @code{BITS_PER_UNIT}. | |
feca2ed3 | 1502 | |
68eb4fb9 LB |
1503 | @findex BOOL_TYPE_SIZE |
1504 | @item BOOL_TYPE_SIZE | |
3d1ad9e5 JM |
1505 | A C expression for the size in bits of the C++ type @code{bool} and |
1506 | C99 type @code{_Bool} on the target machine. If you don't define | |
1507 | this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}. | |
68eb4fb9 | 1508 | |
feca2ed3 JW |
1509 | @findex FLOAT_TYPE_SIZE |
1510 | @item FLOAT_TYPE_SIZE | |
1511 | A C expression for the size in bits of the type @code{float} on the | |
1512 | target machine. If you don't define this, the default is one word. | |
1513 | ||
1514 | @findex DOUBLE_TYPE_SIZE | |
1515 | @item DOUBLE_TYPE_SIZE | |
1516 | A C expression for the size in bits of the type @code{double} on the | |
1517 | target machine. If you don't define this, the default is two | |
1518 | words. | |
1519 | ||
1520 | @findex LONG_DOUBLE_TYPE_SIZE | |
1521 | @item LONG_DOUBLE_TYPE_SIZE | |
1522 | A C expression for the size in bits of the type @code{long double} on | |
1523 | the target machine. If you don't define this, the default is two | |
1524 | words. | |
1525 | ||
aaa2e8ef JW |
1526 | @findex MAX_LONG_DOUBLE_TYPE_SIZE |
1527 | Maximum number for the size in bits of the type @code{long double} on the | |
1528 | target machine. If this is undefined, the default is | |
1529 | @code{LONG_DOUBLE_TYPE_SIZE}. Otherwise, it is the constant value that is | |
1530 | the largest value that @code{LONG_DOUBLE_TYPE_SIZE} can have at run-time. | |
1531 | This is used in @code{cpp}. | |
1532 | ||
d57a4b98 RH |
1533 | @findex TARGET_FLT_EVAL_METHOD |
1534 | @item TARGET_FLT_EVAL_METHOD | |
1535 | A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h}, | |
1536 | assuming, if applicable, that the floating-point control word is in its | |
1537 | default state. If you do not define this macro the value of | |
1538 | @code{FLT_EVAL_METHOD} will be zero. | |
aaa2e8ef | 1539 | |
e9a25f70 JL |
1540 | @findex WIDEST_HARDWARE_FP_SIZE |
1541 | @item WIDEST_HARDWARE_FP_SIZE | |
1542 | A C expression for the size in bits of the widest floating-point format | |
1543 | supported by the hardware. If you define this macro, you must specify a | |
1544 | value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}. | |
1545 | If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE} | |
1546 | is the default. | |
1547 | ||
feca2ed3 JW |
1548 | @findex DEFAULT_SIGNED_CHAR |
1549 | @item DEFAULT_SIGNED_CHAR | |
1550 | An expression whose value is 1 or 0, according to whether the type | |
1551 | @code{char} should be signed or unsigned by default. The user can | |
630d3d5a JM |
1552 | always override this default with the options @option{-fsigned-char} |
1553 | and @option{-funsigned-char}. | |
feca2ed3 JW |
1554 | |
1555 | @findex DEFAULT_SHORT_ENUMS | |
1556 | @item DEFAULT_SHORT_ENUMS | |
1557 | A C expression to determine whether to give an @code{enum} type | |
1558 | only as many bytes as it takes to represent the range of possible values | |
1559 | of that type. A nonzero value means to do that; a zero value means all | |
1560 | @code{enum} types should be allocated like @code{int}. | |
1561 | ||
1562 | If you don't define the macro, the default is 0. | |
1563 | ||
1564 | @findex SIZE_TYPE | |
1565 | @item SIZE_TYPE | |
1566 | A C expression for a string describing the name of the data type to use | |
1567 | for size values. The typedef name @code{size_t} is defined using the | |
1568 | contents of the string. | |
1569 | ||
1570 | The string can contain more than one keyword. If so, separate them with | |
1571 | spaces, and write first any length keyword, then @code{unsigned} if | |
1572 | appropriate, and finally @code{int}. The string must exactly match one | |
1573 | of the data type names defined in the function | |
1574 | @code{init_decl_processing} in the file @file{c-decl.c}. You may not | |
1575 | omit @code{int} or change the order---that would cause the compiler to | |
1576 | crash on startup. | |
1577 | ||
1578 | If you don't define this macro, the default is @code{"long unsigned | |
1579 | int"}. | |
1580 | ||
1581 | @findex PTRDIFF_TYPE | |
1582 | @item PTRDIFF_TYPE | |
1583 | A C expression for a string describing the name of the data type to use | |
1584 | for the result of subtracting two pointers. The typedef name | |
1585 | @code{ptrdiff_t} is defined using the contents of the string. See | |
1586 | @code{SIZE_TYPE} above for more information. | |
1587 | ||
1588 | If you don't define this macro, the default is @code{"long int"}. | |
1589 | ||
1590 | @findex WCHAR_TYPE | |
1591 | @item WCHAR_TYPE | |
1592 | A C expression for a string describing the name of the data type to use | |
1593 | for wide characters. The typedef name @code{wchar_t} is defined using | |
1594 | the contents of the string. See @code{SIZE_TYPE} above for more | |
1595 | information. | |
1596 | ||
1597 | If you don't define this macro, the default is @code{"int"}. | |
1598 | ||
1599 | @findex WCHAR_TYPE_SIZE | |
1600 | @item WCHAR_TYPE_SIZE | |
1601 | A C expression for the size in bits of the data type for wide | |
1602 | characters. This is used in @code{cpp}, which cannot make use of | |
1603 | @code{WCHAR_TYPE}. | |
1604 | ||
1605 | @findex MAX_WCHAR_TYPE_SIZE | |
1606 | @item MAX_WCHAR_TYPE_SIZE | |
1607 | Maximum number for the size in bits of the data type for wide | |
1608 | characters. If this is undefined, the default is | |
1609 | @code{WCHAR_TYPE_SIZE}. Otherwise, it is the constant value that is the | |
1610 | largest value that @code{WCHAR_TYPE_SIZE} can have at run-time. This is | |
1611 | used in @code{cpp}. | |
1612 | ||
b2aec5c0 JH |
1613 | @findex GCOV_TYPE_SIZE |
1614 | @item GCOV_TYPE_SIZE | |
1615 | A C expression for the size in bits of the type used for gcov counters on the | |
1616 | target machine. If you don't define this, the default is one | |
1617 | @code{LONG_TYPE_SIZE} in case it is greater or equal to 64-bit and | |
1618 | @code{LONG_LONG_TYPE_SIZE} otherwise. You may want to re-define the type to | |
1619 | ensure atomicity for counters in multithreaded programs. | |
1620 | ||
1a67c7d3 JL |
1621 | @findex WINT_TYPE |
1622 | @item WINT_TYPE | |
1623 | A C expression for a string describing the name of the data type to | |
1624 | use for wide characters passed to @code{printf} and returned from | |
1625 | @code{getwc}. The typedef name @code{wint_t} is defined using the | |
1626 | contents of the string. See @code{SIZE_TYPE} above for more | |
1627 | information. | |
1628 | ||
1629 | If you don't define this macro, the default is @code{"unsigned int"}. | |
1630 | ||
b15ad712 JM |
1631 | @findex INTMAX_TYPE |
1632 | @item INTMAX_TYPE | |
1633 | A C expression for a string describing the name of the data type that | |
1634 | can represent any value of any standard or extended signed integer type. | |
1635 | The typedef name @code{intmax_t} is defined using the contents of the | |
1636 | string. See @code{SIZE_TYPE} above for more information. | |
1637 | ||
1638 | If you don't define this macro, the default is the first of | |
1639 | @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as | |
1640 | much precision as @code{long long int}. | |
1641 | ||
1642 | @findex UINTMAX_TYPE | |
1643 | @item UINTMAX_TYPE | |
1644 | A C expression for a string describing the name of the data type that | |
1645 | can represent any value of any standard or extended unsigned integer | |
1646 | type. The typedef name @code{uintmax_t} is defined using the contents | |
1647 | of the string. See @code{SIZE_TYPE} above for more information. | |
1648 | ||
1649 | If you don't define this macro, the default is the first of | |
1650 | @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long | |
1651 | unsigned int"} that has as much precision as @code{long long unsigned | |
1652 | int}. | |
1653 | ||
f3c55c97 AO |
1654 | @findex TARGET_PTRMEMFUNC_VBIT_LOCATION |
1655 | @item TARGET_PTRMEMFUNC_VBIT_LOCATION | |
1656 | The C++ compiler represents a pointer-to-member-function with a struct | |
1657 | that looks like: | |
1658 | ||
1659 | @example | |
1660 | struct @{ | |
1661 | union @{ | |
1662 | void (*fn)(); | |
1663 | ptrdiff_t vtable_index; | |
1664 | @}; | |
1665 | ptrdiff_t delta; | |
1666 | @}; | |
1667 | @end example | |
1668 | ||
1669 | @noindent | |
1670 | The C++ compiler must use one bit to indicate whether the function that | |
1671 | will be called through a pointer-to-member-function is virtual. | |
1672 | Normally, we assume that the low-order bit of a function pointer must | |
1673 | always be zero. Then, by ensuring that the vtable_index is odd, we can | |
1674 | distinguish which variant of the union is in use. But, on some | |
1675 | platforms function pointers can be odd, and so this doesn't work. In | |
1676 | that case, we use the low-order bit of the @code{delta} field, and shift | |
1677 | the remainder of the @code{delta} field to the left. | |
1678 | ||
1679 | GCC will automatically make the right selection about where to store | |
1680 | this bit using the @code{FUNCTION_BOUNDARY} setting for your platform. | |
1681 | However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY} | |
1682 | set such that functions always start at even addresses, but the lowest | |
1683 | bit of pointers to functions indicate whether the function at that | |
1684 | address is in ARM or Thumb mode. If this is the case of your | |
1685 | architecture, you should define this macro to | |
1686 | @code{ptrmemfunc_vbit_in_delta}. | |
1687 | ||
1688 | In general, you should not have to define this macro. On architectures | |
1689 | in which function addresses are always even, according to | |
1690 | @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to | |
1691 | @code{ptrmemfunc_vbit_in_pfn}. | |
67231816 RH |
1692 | |
1693 | @findex TARGET_VTABLE_USES_DESCRIPTORS | |
1694 | @item TARGET_VTABLE_USES_DESCRIPTORS | |
1695 | Normally, the C++ compiler uses function pointers in vtables. This | |
f282ffb3 | 1696 | macro allows the target to change to use ``function descriptors'' |
67231816 RH |
1697 | instead. Function descriptors are found on targets for whom a |
1698 | function pointer is actually a small data structure. Normally the | |
f282ffb3 | 1699 | data structure consists of the actual code address plus a data |
67231816 RH |
1700 | pointer to which the function's data is relative. |
1701 | ||
1702 | If vtables are used, the value of this macro should be the number | |
1703 | of words that the function descriptor occupies. | |
a6f5e048 RH |
1704 | |
1705 | @findex TARGET_VTABLE_ENTRY_ALIGN | |
1706 | @item TARGET_VTABLE_ENTRY_ALIGN | |
1707 | By default, the vtable entries are void pointers, the so the alignment | |
1708 | is the same as pointer alignment. The value of this macro specifies | |
1709 | the alignment of the vtable entry in bits. It should be defined only | |
1710 | when special alignment is necessary. */ | |
1711 | ||
1712 | @findex TARGET_VTABLE_DATA_ENTRY_DISTANCE | |
1713 | @item TARGET_VTABLE_DATA_ENTRY_DISTANCE | |
1714 | There are a few non-descriptor entries in the vtable at offsets below | |
1715 | zero. If these entries must be padded (say, to preserve the alignment | |
1716 | specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number | |
1717 | of words in each data entry. | |
b2b263e1 NB |
1718 | @end table |
1719 | ||
1720 | @node Escape Sequences | |
1721 | @section Target Character Escape Sequences | |
1722 | @cindex escape sequences | |
f3c55c97 | 1723 | |
b2b263e1 NB |
1724 | By default, GCC assumes that the C character escape sequences take on |
1725 | their ASCII values for the target. If this is not correct, you must | |
1726 | explicitly define all of the macros below. | |
1727 | ||
1728 | @table @code | |
feca2ed3 JW |
1729 | @findex TARGET_BELL |
1730 | @item TARGET_BELL | |
1731 | A C constant expression for the integer value for escape sequence | |
1732 | @samp{\a}. | |
1733 | ||
501990bb NB |
1734 | @findex TARGET_ESC |
1735 | @item TARGET_ESC | |
1736 | A C constant expression for the integer value of the target escape | |
1737 | character. As an extension, GCC evaluates the escape sequences | |
1738 | @samp{\e} and @samp{\E} to this. | |
1739 | ||
feca2ed3 JW |
1740 | @findex TARGET_TAB |
1741 | @findex TARGET_BS | |
1742 | @findex TARGET_NEWLINE | |
1743 | @item TARGET_BS | |
1744 | @itemx TARGET_TAB | |
1745 | @itemx TARGET_NEWLINE | |
1746 | C constant expressions for the integer values for escape sequences | |
1747 | @samp{\b}, @samp{\t} and @samp{\n}. | |
1748 | ||
1749 | @findex TARGET_VT | |
1750 | @findex TARGET_FF | |
1751 | @findex TARGET_CR | |
1752 | @item TARGET_VT | |
1753 | @itemx TARGET_FF | |
1754 | @itemx TARGET_CR | |
1755 | C constant expressions for the integer values for escape sequences | |
1756 | @samp{\v}, @samp{\f} and @samp{\r}. | |
1757 | @end table | |
1758 | ||
1759 | @node Registers | |
1760 | @section Register Usage | |
1761 | @cindex register usage | |
1762 | ||
1763 | This section explains how to describe what registers the target machine | |
1764 | has, and how (in general) they can be used. | |
1765 | ||
1766 | The description of which registers a specific instruction can use is | |
1767 | done with register classes; see @ref{Register Classes}. For information | |
1768 | on using registers to access a stack frame, see @ref{Frame Registers}. | |
1769 | For passing values in registers, see @ref{Register Arguments}. | |
1770 | For returning values in registers, see @ref{Scalar Return}. | |
1771 | ||
1772 | @menu | |
1773 | * Register Basics:: Number and kinds of registers. | |
1774 | * Allocation Order:: Order in which registers are allocated. | |
1775 | * Values in Registers:: What kinds of values each reg can hold. | |
1776 | * Leaf Functions:: Renumbering registers for leaf functions. | |
1777 | * Stack Registers:: Handling a register stack such as 80387. | |
feca2ed3 JW |
1778 | @end menu |
1779 | ||
1780 | @node Register Basics | |
1781 | @subsection Basic Characteristics of Registers | |
1782 | ||
1783 | @c prevent bad page break with this line | |
1784 | Registers have various characteristics. | |
1785 | ||
1786 | @table @code | |
1787 | @findex FIRST_PSEUDO_REGISTER | |
1788 | @item FIRST_PSEUDO_REGISTER | |
1789 | Number of hardware registers known to the compiler. They receive | |
1790 | numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first | |
1791 | pseudo register's number really is assigned the number | |
1792 | @code{FIRST_PSEUDO_REGISTER}. | |
1793 | ||
1794 | @item FIXED_REGISTERS | |
1795 | @findex FIXED_REGISTERS | |
1796 | @cindex fixed register | |
1797 | An initializer that says which registers are used for fixed purposes | |
1798 | all throughout the compiled code and are therefore not available for | |
1799 | general allocation. These would include the stack pointer, the frame | |
1800 | pointer (except on machines where that can be used as a general | |
1801 | register when no frame pointer is needed), the program counter on | |
1802 | machines where that is considered one of the addressable registers, | |
1803 | and any other numbered register with a standard use. | |
1804 | ||
1805 | This information is expressed as a sequence of numbers, separated by | |
1806 | commas and surrounded by braces. The @var{n}th number is 1 if | |
1807 | register @var{n} is fixed, 0 otherwise. | |
1808 | ||
1809 | The table initialized from this macro, and the table initialized by | |
1810 | the following one, may be overridden at run time either automatically, | |
1811 | by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by | |
630d3d5a JM |
1812 | the user with the command options @option{-ffixed-@var{reg}}, |
1813 | @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}. | |
feca2ed3 JW |
1814 | |
1815 | @findex CALL_USED_REGISTERS | |
1816 | @item CALL_USED_REGISTERS | |
1817 | @cindex call-used register | |
1818 | @cindex call-clobbered register | |
1819 | @cindex call-saved register | |
1820 | Like @code{FIXED_REGISTERS} but has 1 for each register that is | |
1821 | clobbered (in general) by function calls as well as for fixed | |
1822 | registers. This macro therefore identifies the registers that are not | |
1823 | available for general allocation of values that must live across | |
1824 | function calls. | |
1825 | ||
1826 | If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler | |
1827 | automatically saves it on function entry and restores it on function | |
1828 | exit, if the register is used within the function. | |
1829 | ||
fc1296b7 AM |
1830 | @findex CALL_REALLY_USED_REGISTERS |
1831 | @item CALL_REALLY_USED_REGISTERS | |
1832 | @cindex call-used register | |
1833 | @cindex call-clobbered register | |
1834 | @cindex call-saved register | |
f282ffb3 JM |
1835 | Like @code{CALL_USED_REGISTERS} except this macro doesn't require |
1836 | that the entire set of @code{FIXED_REGISTERS} be included. | |
fc1296b7 | 1837 | (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}). |
f282ffb3 | 1838 | This macro is optional. If not specified, it defaults to the value |
fc1296b7 AM |
1839 | of @code{CALL_USED_REGISTERS}. |
1840 | ||
1e326708 MH |
1841 | @findex HARD_REGNO_CALL_PART_CLOBBERED |
1842 | @item HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode}) | |
1843 | @cindex call-used register | |
1844 | @cindex call-clobbered register | |
1845 | @cindex call-saved register | |
df2a54e9 | 1846 | A C expression that is nonzero if it is not permissible to store a |
1e326708 MH |
1847 | value of mode @var{mode} in hard register number @var{regno} across a |
1848 | call without some part of it being clobbered. For most machines this | |
1849 | macro need not be defined. It is only required for machines that do not | |
1850 | preserve the entire contents of a register across a call. | |
1851 | ||
feca2ed3 JW |
1852 | @findex CONDITIONAL_REGISTER_USAGE |
1853 | @findex fixed_regs | |
1854 | @findex call_used_regs | |
1855 | @item CONDITIONAL_REGISTER_USAGE | |
055177dc NC |
1856 | Zero or more C statements that may conditionally modify five variables |
1857 | @code{fixed_regs}, @code{call_used_regs}, @code{global_regs}, | |
c237e94a ZW |
1858 | @code{reg_names}, and @code{reg_class_contents}, to take into account |
1859 | any dependence of these register sets on target flags. The first three | |
1860 | of these are of type @code{char []} (interpreted as Boolean vectors). | |
1861 | @code{global_regs} is a @code{const char *[]}, and | |
1862 | @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is | |
1863 | called, @code{fixed_regs}, @code{call_used_regs}, | |
1864 | @code{reg_class_contents}, and @code{reg_names} have been initialized | |
055177dc | 1865 | from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS}, |
c237e94a | 1866 | @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively. |
630d3d5a | 1867 | @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}}, |
c237e94a ZW |
1868 | @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}} |
1869 | command options have been applied. | |
feca2ed3 JW |
1870 | |
1871 | You need not define this macro if it has no work to do. | |
1872 | ||
1873 | @cindex disabling certain registers | |
1874 | @cindex controlling register usage | |
1875 | If the usage of an entire class of registers depends on the target | |
1876 | flags, you may indicate this to GCC by using this macro to modify | |
1877 | @code{fixed_regs} and @code{call_used_regs} to 1 for each of the | |
161d7b59 | 1878 | registers in the classes which should not be used by GCC@. Also define |
feca2ed3 JW |
1879 | the macro @code{REG_CLASS_FROM_LETTER} to return @code{NO_REGS} if it |
1880 | is called with a letter for a class that shouldn't be used. | |
1881 | ||
1882 | (However, if this class is not included in @code{GENERAL_REGS} and all | |
1883 | of the insn patterns whose constraints permit this class are | |
1884 | controlled by target switches, then GCC will automatically avoid using | |
1885 | these registers when the target switches are opposed to them.) | |
1886 | ||
1887 | @findex NON_SAVING_SETJMP | |
1888 | @item NON_SAVING_SETJMP | |
1889 | If this macro is defined and has a nonzero value, it means that | |
1890 | @code{setjmp} and related functions fail to save the registers, or that | |
1891 | @code{longjmp} fails to restore them. To compensate, the compiler | |
1892 | avoids putting variables in registers in functions that use | |
1893 | @code{setjmp}. | |
1894 | ||
1895 | @findex INCOMING_REGNO | |
1896 | @item INCOMING_REGNO (@var{out}) | |
1897 | Define this macro if the target machine has register windows. This C | |
1898 | expression returns the register number as seen by the called function | |
1899 | corresponding to the register number @var{out} as seen by the calling | |
1900 | function. Return @var{out} if register number @var{out} is not an | |
1901 | outbound register. | |
1902 | ||
1903 | @findex OUTGOING_REGNO | |
1904 | @item OUTGOING_REGNO (@var{in}) | |
1905 | Define this macro if the target machine has register windows. This C | |
1906 | expression returns the register number as seen by the calling function | |
1907 | corresponding to the register number @var{in} as seen by the called | |
1908 | function. Return @var{in} if register number @var{in} is not an inbound | |
1909 | register. | |
1910 | ||
fa80e43d JL |
1911 | @findex LOCAL_REGNO |
1912 | @item LOCAL_REGNO (@var{regno}) | |
1913 | Define this macro if the target machine has register windows. This C | |
1914 | expression returns true if the register is call-saved but is in the | |
1915 | register window. Unlike most call-saved registers, such registers | |
1916 | need not be explicitly restored on function exit or during non-local | |
1917 | gotos. | |
1918 | ||
feca2ed3 JW |
1919 | @ignore |
1920 | @findex PC_REGNUM | |
1921 | @item PC_REGNUM | |
1922 | If the program counter has a register number, define this as that | |
1923 | register number. Otherwise, do not define it. | |
1924 | @end ignore | |
1925 | @end table | |
1926 | ||
1927 | @node Allocation Order | |
1928 | @subsection Order of Allocation of Registers | |
1929 | @cindex order of register allocation | |
1930 | @cindex register allocation order | |
1931 | ||
1932 | @c prevent bad page break with this line | |
1933 | Registers are allocated in order. | |
1934 | ||
1935 | @table @code | |
1936 | @findex REG_ALLOC_ORDER | |
1937 | @item REG_ALLOC_ORDER | |
1938 | If defined, an initializer for a vector of integers, containing the | |
a3a15b4d | 1939 | numbers of hard registers in the order in which GCC should prefer |
feca2ed3 JW |
1940 | to use them (from most preferred to least). |
1941 | ||
1942 | If this macro is not defined, registers are used lowest numbered first | |
1943 | (all else being equal). | |
1944 | ||
1945 | One use of this macro is on machines where the highest numbered | |
1946 | registers must always be saved and the save-multiple-registers | |
1947 | instruction supports only sequences of consecutive registers. On such | |
1948 | machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists | |
956d6950 | 1949 | the highest numbered allocable register first. |
feca2ed3 JW |
1950 | |
1951 | @findex ORDER_REGS_FOR_LOCAL_ALLOC | |
1952 | @item ORDER_REGS_FOR_LOCAL_ALLOC | |
1953 | A C statement (sans semicolon) to choose the order in which to allocate | |
1954 | hard registers for pseudo-registers local to a basic block. | |
1955 | ||
1956 | Store the desired register order in the array @code{reg_alloc_order}. | |
1957 | Element 0 should be the register to allocate first; element 1, the next | |
1958 | register; and so on. | |
1959 | ||
1960 | The macro body should not assume anything about the contents of | |
1961 | @code{reg_alloc_order} before execution of the macro. | |
1962 | ||
1963 | On most machines, it is not necessary to define this macro. | |
1964 | @end table | |
1965 | ||
1966 | @node Values in Registers | |
1967 | @subsection How Values Fit in Registers | |
1968 | ||
1969 | This section discusses the macros that describe which kinds of values | |
1970 | (specifically, which machine modes) each register can hold, and how many | |
1971 | consecutive registers are needed for a given mode. | |
1972 | ||
1973 | @table @code | |
1974 | @findex HARD_REGNO_NREGS | |
1975 | @item HARD_REGNO_NREGS (@var{regno}, @var{mode}) | |
1976 | A C expression for the number of consecutive hard registers, starting | |
1977 | at register number @var{regno}, required to hold a value of mode | |
1978 | @var{mode}. | |
1979 | ||
1980 | On a machine where all registers are exactly one word, a suitable | |
1981 | definition of this macro is | |
1982 | ||
1983 | @smallexample | |
1984 | #define HARD_REGNO_NREGS(REGNO, MODE) \ | |
1985 | ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \ | |
32bd3974 | 1986 | / UNITS_PER_WORD) |
feca2ed3 JW |
1987 | @end smallexample |
1988 | ||
1989 | @findex HARD_REGNO_MODE_OK | |
1990 | @item HARD_REGNO_MODE_OK (@var{regno}, @var{mode}) | |
1991 | A C expression that is nonzero if it is permissible to store a value | |
1992 | of mode @var{mode} in hard register number @var{regno} (or in several | |
1993 | registers starting with that one). For a machine where all registers | |
1994 | are equivalent, a suitable definition is | |
1995 | ||
1996 | @smallexample | |
1997 | #define HARD_REGNO_MODE_OK(REGNO, MODE) 1 | |
1998 | @end smallexample | |
1999 | ||
e9a25f70 JL |
2000 | You need not include code to check for the numbers of fixed registers, |
2001 | because the allocation mechanism considers them to be always occupied. | |
feca2ed3 JW |
2002 | |
2003 | @cindex register pairs | |
2004 | On some machines, double-precision values must be kept in even/odd | |
e9a25f70 JL |
2005 | register pairs. You can implement that by defining this macro to reject |
2006 | odd register numbers for such modes. | |
feca2ed3 JW |
2007 | |
2008 | The minimum requirement for a mode to be OK in a register is that the | |
2009 | @samp{mov@var{mode}} instruction pattern support moves between the | |
e9a25f70 JL |
2010 | register and other hard register in the same class and that moving a |
2011 | value into the register and back out not alter it. | |
feca2ed3 | 2012 | |
e9a25f70 JL |
2013 | Since the same instruction used to move @code{word_mode} will work for |
2014 | all narrower integer modes, it is not necessary on any machine for | |
feca2ed3 JW |
2015 | @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided |
2016 | you define patterns @samp{movhi}, etc., to take advantage of this. This | |
2017 | is useful because of the interaction between @code{HARD_REGNO_MODE_OK} | |
2018 | and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes | |
2019 | to be tieable. | |
2020 | ||
2021 | Many machines have special registers for floating point arithmetic. | |
2022 | Often people assume that floating point machine modes are allowed only | |
2023 | in floating point registers. This is not true. Any registers that | |
2024 | can hold integers can safely @emph{hold} a floating point machine | |
2025 | mode, whether or not floating arithmetic can be done on it in those | |
2026 | registers. Integer move instructions can be used to move the values. | |
2027 | ||
2028 | On some machines, though, the converse is true: fixed-point machine | |
2029 | modes may not go in floating registers. This is true if the floating | |
2030 | registers normalize any value stored in them, because storing a | |
2031 | non-floating value there would garble it. In this case, | |
2032 | @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in | |
2033 | floating registers. But if the floating registers do not automatically | |
2034 | normalize, if you can store any bit pattern in one and retrieve it | |
2035 | unchanged without a trap, then any machine mode may go in a floating | |
2036 | register, so you can define this macro to say so. | |
2037 | ||
2038 | The primary significance of special floating registers is rather that | |
2039 | they are the registers acceptable in floating point arithmetic | |
2040 | instructions. However, this is of no concern to | |
2041 | @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper | |
2042 | constraints for those instructions. | |
2043 | ||
2044 | On some machines, the floating registers are especially slow to access, | |
2045 | so that it is better to store a value in a stack frame than in such a | |
2046 | register if floating point arithmetic is not being done. As long as the | |
2047 | floating registers are not in class @code{GENERAL_REGS}, they will not | |
2048 | be used unless some pattern's constraint asks for one. | |
2049 | ||
2050 | @findex MODES_TIEABLE_P | |
2051 | @item MODES_TIEABLE_P (@var{mode1}, @var{mode2}) | |
e9a25f70 | 2052 | A C expression that is nonzero if a value of mode |
956d6950 | 2053 | @var{mode1} is accessible in mode @var{mode2} without copying. |
feca2ed3 JW |
2054 | |
2055 | If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and | |
e9a25f70 JL |
2056 | @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for |
2057 | any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})} | |
2058 | should be nonzero. If they differ for any @var{r}, you should define | |
2059 | this macro to return zero unless some other mechanism ensures the | |
956d6950 | 2060 | accessibility of the value in a narrower mode. |
e9a25f70 JL |
2061 | |
2062 | You should define this macro to return nonzero in as many cases as | |
a3a15b4d | 2063 | possible since doing so will allow GCC to perform better register |
e9a25f70 | 2064 | allocation. |
7506f491 DE |
2065 | |
2066 | @findex AVOID_CCMODE_COPIES | |
2067 | @item AVOID_CCMODE_COPIES | |
2068 | Define this macro if the compiler should avoid copies to/from @code{CCmode} | |
a89608cb | 2069 | registers. You should only define this macro if support for copying to/from |
7506f491 | 2070 | @code{CCmode} is incomplete. |
feca2ed3 JW |
2071 | @end table |
2072 | ||
2073 | @node Leaf Functions | |
2074 | @subsection Handling Leaf Functions | |
2075 | ||
2076 | @cindex leaf functions | |
2077 | @cindex functions, leaf | |
2078 | On some machines, a leaf function (i.e., one which makes no calls) can run | |
2079 | more efficiently if it does not make its own register window. Often this | |
2080 | means it is required to receive its arguments in the registers where they | |
2081 | are passed by the caller, instead of the registers where they would | |
2082 | normally arrive. | |
2083 | ||
2084 | The special treatment for leaf functions generally applies only when | |
2085 | other conditions are met; for example, often they may use only those | |
2086 | registers for its own variables and temporaries. We use the term ``leaf | |
2087 | function'' to mean a function that is suitable for this special | |
2088 | handling, so that functions with no calls are not necessarily ``leaf | |
2089 | functions''. | |
2090 | ||
a3a15b4d | 2091 | GCC assigns register numbers before it knows whether the function is |
feca2ed3 JW |
2092 | suitable for leaf function treatment. So it needs to renumber the |
2093 | registers in order to output a leaf function. The following macros | |
2094 | accomplish this. | |
2095 | ||
2096 | @table @code | |
2097 | @findex LEAF_REGISTERS | |
2098 | @item LEAF_REGISTERS | |
7d167afd | 2099 | Name of a char vector, indexed by hard register number, which |
feca2ed3 JW |
2100 | contains 1 for a register that is allowable in a candidate for leaf |
2101 | function treatment. | |
2102 | ||
2103 | If leaf function treatment involves renumbering the registers, then the | |
2104 | registers marked here should be the ones before renumbering---those that | |
a3a15b4d | 2105 | GCC would ordinarily allocate. The registers which will actually be |
feca2ed3 JW |
2106 | used in the assembler code, after renumbering, should not be marked with 1 |
2107 | in this vector. | |
2108 | ||
2109 | Define this macro only if the target machine offers a way to optimize | |
2110 | the treatment of leaf functions. | |
2111 | ||
2112 | @findex LEAF_REG_REMAP | |
2113 | @item LEAF_REG_REMAP (@var{regno}) | |
2114 | A C expression whose value is the register number to which @var{regno} | |
2115 | should be renumbered, when a function is treated as a leaf function. | |
2116 | ||
2117 | If @var{regno} is a register number which should not appear in a leaf | |
630d3d5a | 2118 | function before renumbering, then the expression should yield @minus{}1, which |
feca2ed3 JW |
2119 | will cause the compiler to abort. |
2120 | ||
2121 | Define this macro only if the target machine offers a way to optimize the | |
2122 | treatment of leaf functions, and registers need to be renumbered to do | |
2123 | this. | |
2124 | @end table | |
2125 | ||
54ff41b7 JW |
2126 | @findex current_function_is_leaf |
2127 | @findex current_function_uses_only_leaf_regs | |
c237e94a ZW |
2128 | @code{TARGET_ASM_FUNCTION_PROLOGUE} and |
2129 | @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions | |
2130 | specially. They can test the C variable @code{current_function_is_leaf} | |
2131 | which is nonzero for leaf functions. @code{current_function_is_leaf} is | |
2132 | set prior to local register allocation and is valid for the remaining | |
08c148a8 NB |
2133 | compiler passes. They can also test the C variable |
2134 | @code{current_function_uses_only_leaf_regs} which is nonzero for leaf | |
2135 | functions which only use leaf registers. | |
54ff41b7 JW |
2136 | @code{current_function_uses_only_leaf_regs} is valid after reload and is |
2137 | only useful if @code{LEAF_REGISTERS} is defined. | |
feca2ed3 JW |
2138 | @c changed this to fix overfull. ALSO: why the "it" at the beginning |
2139 | @c of the next paragraph?! --mew 2feb93 | |
2140 | ||
2141 | @node Stack Registers | |
2142 | @subsection Registers That Form a Stack | |
2143 | ||
2144 | There are special features to handle computers where some of the | |
2145 | ``registers'' form a stack, as in the 80387 coprocessor for the 80386. | |
2146 | Stack registers are normally written by pushing onto the stack, and are | |
2147 | numbered relative to the top of the stack. | |
2148 | ||
a3a15b4d | 2149 | Currently, GCC can only handle one group of stack-like registers, and |
feca2ed3 JW |
2150 | they must be consecutively numbered. |
2151 | ||
2152 | @table @code | |
2153 | @findex STACK_REGS | |
2154 | @item STACK_REGS | |
2155 | Define this if the machine has any stack-like registers. | |
2156 | ||
2157 | @findex FIRST_STACK_REG | |
2158 | @item FIRST_STACK_REG | |
2159 | The number of the first stack-like register. This one is the top | |
2160 | of the stack. | |
2161 | ||
2162 | @findex LAST_STACK_REG | |
2163 | @item LAST_STACK_REG | |
2164 | The number of the last stack-like register. This one is the bottom of | |
2165 | the stack. | |
2166 | @end table | |
2167 | ||
feca2ed3 JW |
2168 | @node Register Classes |
2169 | @section Register Classes | |
2170 | @cindex register class definitions | |
2171 | @cindex class definitions, register | |
2172 | ||
2173 | On many machines, the numbered registers are not all equivalent. | |
2174 | For example, certain registers may not be allowed for indexed addressing; | |
2175 | certain registers may not be allowed in some instructions. These machine | |
2176 | restrictions are described to the compiler using @dfn{register classes}. | |
2177 | ||
2178 | You define a number of register classes, giving each one a name and saying | |
2179 | which of the registers belong to it. Then you can specify register classes | |
2180 | that are allowed as operands to particular instruction patterns. | |
2181 | ||
2182 | @findex ALL_REGS | |
2183 | @findex NO_REGS | |
2184 | In general, each register will belong to several classes. In fact, one | |
2185 | class must be named @code{ALL_REGS} and contain all the registers. Another | |
2186 | class must be named @code{NO_REGS} and contain no registers. Often the | |
2187 | union of two classes will be another class; however, this is not required. | |
2188 | ||
2189 | @findex GENERAL_REGS | |
2190 | One of the classes must be named @code{GENERAL_REGS}. There is nothing | |
2191 | terribly special about the name, but the operand constraint letters | |
2192 | @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is | |
2193 | the same as @code{ALL_REGS}, just define it as a macro which expands | |
2194 | to @code{ALL_REGS}. | |
2195 | ||
2196 | Order the classes so that if class @var{x} is contained in class @var{y} | |
2197 | then @var{x} has a lower class number than @var{y}. | |
2198 | ||
2199 | The way classes other than @code{GENERAL_REGS} are specified in operand | |
2200 | constraints is through machine-dependent operand constraint letters. | |
2201 | You can define such letters to correspond to various classes, then use | |
2202 | them in operand constraints. | |
2203 | ||
2204 | You should define a class for the union of two classes whenever some | |
2205 | instruction allows both classes. For example, if an instruction allows | |
2206 | either a floating point (coprocessor) register or a general register for a | |
2207 | certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS} | |
2208 | which includes both of them. Otherwise you will get suboptimal code. | |
2209 | ||
2210 | You must also specify certain redundant information about the register | |
2211 | classes: for each class, which classes contain it and which ones are | |
2212 | contained in it; for each pair of classes, the largest class contained | |
2213 | in their union. | |
2214 | ||
2215 | When a value occupying several consecutive registers is expected in a | |
2216 | certain class, all the registers used must belong to that class. | |
2217 | Therefore, register classes cannot be used to enforce a requirement for | |
2218 | a register pair to start with an even-numbered register. The way to | |
2219 | specify this requirement is with @code{HARD_REGNO_MODE_OK}. | |
2220 | ||
2221 | Register classes used for input-operands of bitwise-and or shift | |
2222 | instructions have a special requirement: each such class must have, for | |
2223 | each fixed-point machine mode, a subclass whose registers can transfer that | |
2224 | mode to or from memory. For example, on some machines, the operations for | |
2225 | single-byte values (@code{QImode}) are limited to certain registers. When | |
2226 | this is so, each register class that is used in a bitwise-and or shift | |
2227 | instruction must have a subclass consisting of registers from which | |
2228 | single-byte values can be loaded or stored. This is so that | |
2229 | @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return. | |
2230 | ||
2231 | @table @code | |
2232 | @findex enum reg_class | |
2233 | @item enum reg_class | |
2234 | An enumeral type that must be defined with all the register class names | |
2235 | as enumeral values. @code{NO_REGS} must be first. @code{ALL_REGS} | |
2236 | must be the last register class, followed by one more enumeral value, | |
2237 | @code{LIM_REG_CLASSES}, which is not a register class but rather | |
2238 | tells how many classes there are. | |
2239 | ||
2240 | Each register class has a number, which is the value of casting | |
2241 | the class name to type @code{int}. The number serves as an index | |
2242 | in many of the tables described below. | |
2243 | ||
2244 | @findex N_REG_CLASSES | |
2245 | @item N_REG_CLASSES | |
2246 | The number of distinct register classes, defined as follows: | |
2247 | ||
2248 | @example | |
2249 | #define N_REG_CLASSES (int) LIM_REG_CLASSES | |
2250 | @end example | |
2251 | ||
2252 | @findex REG_CLASS_NAMES | |
2253 | @item REG_CLASS_NAMES | |
2254 | An initializer containing the names of the register classes as C string | |
2255 | constants. These names are used in writing some of the debugging dumps. | |
2256 | ||
2257 | @findex REG_CLASS_CONTENTS | |
2258 | @item REG_CLASS_CONTENTS | |
2259 | An initializer containing the contents of the register classes, as integers | |
2260 | which are bit masks. The @var{n}th integer specifies the contents of class | |
2261 | @var{n}. The way the integer @var{mask} is interpreted is that | |
2262 | register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1. | |
2263 | ||
2264 | When the machine has more than 32 registers, an integer does not suffice. | |
2265 | Then the integers are replaced by sub-initializers, braced groupings containing | |
2266 | several integers. Each sub-initializer must be suitable as an initializer | |
2267 | for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}. | |
7c272079 MP |
2268 | In this situation, the first integer in each sub-initializer corresponds to |
2269 | registers 0 through 31, the second integer to registers 32 through 63, and | |
2270 | so on. | |
feca2ed3 JW |
2271 | |
2272 | @findex REGNO_REG_CLASS | |
2273 | @item REGNO_REG_CLASS (@var{regno}) | |
2274 | A C expression whose value is a register class containing hard register | |
2275 | @var{regno}. In general there is more than one such class; choose a class | |
2276 | which is @dfn{minimal}, meaning that no smaller class also contains the | |
2277 | register. | |
2278 | ||
2279 | @findex BASE_REG_CLASS | |
2280 | @item BASE_REG_CLASS | |
2281 | A macro whose definition is the name of the class to which a valid | |
2282 | base register must belong. A base register is one used in an address | |
2283 | which is the register value plus a displacement. | |
2284 | ||
3dcc68a4 NC |
2285 | @findex MODE_BASE_REG_CLASS |
2286 | @item MODE_BASE_REG_CLASS (@var{mode}) | |
2287 | This is a variation of the @code{BASE_REG_CLASS} macro which allows | |
2288 | the selection of a base register in a mode depenedent manner. If | |
2289 | @var{mode} is VOIDmode then it should return the same value as | |
2290 | @code{BASE_REG_CLASS}. | |
2291 | ||
feca2ed3 JW |
2292 | @findex INDEX_REG_CLASS |
2293 | @item INDEX_REG_CLASS | |
2294 | A macro whose definition is the name of the class to which a valid | |
2295 | index register must belong. An index register is one used in an | |
2296 | address where its value is either multiplied by a scale factor or | |
2297 | added to another register (as well as added to a displacement). | |
2298 | ||
2299 | @findex REG_CLASS_FROM_LETTER | |
2300 | @item REG_CLASS_FROM_LETTER (@var{char}) | |
2301 | A C expression which defines the machine-dependent operand constraint | |
2302 | letters for register classes. If @var{char} is such a letter, the | |
2303 | value should be the register class corresponding to it. Otherwise, | |
2304 | the value should be @code{NO_REGS}. The register letter @samp{r}, | |
2305 | corresponding to class @code{GENERAL_REGS}, will not be passed | |
2306 | to this macro; you do not need to handle it. | |
2307 | ||
2308 | @findex REGNO_OK_FOR_BASE_P | |
2309 | @item REGNO_OK_FOR_BASE_P (@var{num}) | |
2310 | A C expression which is nonzero if register number @var{num} is | |
2311 | suitable for use as a base register in operand addresses. It may be | |
2312 | either a suitable hard register or a pseudo register that has been | |
2313 | allocated such a hard register. | |
2314 | ||
861bb6c1 JL |
2315 | @findex REGNO_MODE_OK_FOR_BASE_P |
2316 | @item REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode}) | |
2317 | A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that | |
2318 | that expression may examine the mode of the memory reference in | |
2319 | @var{mode}. You should define this macro if the mode of the memory | |
2320 | reference affects whether a register may be used as a base register. If | |
2321 | you define this macro, the compiler will use it instead of | |
2322 | @code{REGNO_OK_FOR_BASE_P}. | |
2323 | ||
feca2ed3 JW |
2324 | @findex REGNO_OK_FOR_INDEX_P |
2325 | @item REGNO_OK_FOR_INDEX_P (@var{num}) | |
2326 | A C expression which is nonzero if register number @var{num} is | |
2327 | suitable for use as an index register in operand addresses. It may be | |
2328 | either a suitable hard register or a pseudo register that has been | |
2329 | allocated such a hard register. | |
2330 | ||
2331 | The difference between an index register and a base register is that | |
2332 | the index register may be scaled. If an address involves the sum of | |
2333 | two registers, neither one of them scaled, then either one may be | |
2334 | labeled the ``base'' and the other the ``index''; but whichever | |
2335 | labeling is used must fit the machine's constraints of which registers | |
2336 | may serve in each capacity. The compiler will try both labelings, | |
2337 | looking for one that is valid, and will reload one or both registers | |
2338 | only if neither labeling works. | |
2339 | ||
2340 | @findex PREFERRED_RELOAD_CLASS | |
2341 | @item PREFERRED_RELOAD_CLASS (@var{x}, @var{class}) | |
2342 | A C expression that places additional restrictions on the register class | |
2343 | to use when it is necessary to copy value @var{x} into a register in class | |
2344 | @var{class}. The value is a register class; perhaps @var{class}, or perhaps | |
2345 | another, smaller class. On many machines, the following definition is | |
2346 | safe: | |
2347 | ||
2348 | @example | |
2349 | #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS | |
2350 | @end example | |
2351 | ||
2352 | Sometimes returning a more restrictive class makes better code. For | |
2353 | example, on the 68000, when @var{x} is an integer constant that is in range | |
2354 | for a @samp{moveq} instruction, the value of this macro is always | |
2355 | @code{DATA_REGS} as long as @var{class} includes the data registers. | |
2356 | Requiring a data register guarantees that a @samp{moveq} will be used. | |
2357 | ||
2358 | If @var{x} is a @code{const_double}, by returning @code{NO_REGS} | |
2359 | you can force @var{x} into a memory constant. This is useful on | |
2360 | certain machines where immediate floating values cannot be loaded into | |
2361 | certain kinds of registers. | |
2362 | ||
2363 | @findex PREFERRED_OUTPUT_RELOAD_CLASS | |
2364 | @item PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class}) | |
2365 | Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of | |
2366 | input reloads. If you don't define this macro, the default is to use | |
2367 | @var{class}, unchanged. | |
2368 | ||
2369 | @findex LIMIT_RELOAD_CLASS | |
2370 | @item LIMIT_RELOAD_CLASS (@var{mode}, @var{class}) | |
2371 | A C expression that places additional restrictions on the register class | |
2372 | to use when it is necessary to be able to hold a value of mode | |
2373 | @var{mode} in a reload register for which class @var{class} would | |
2374 | ordinarily be used. | |
2375 | ||
2376 | Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when | |
2377 | there are certain modes that simply can't go in certain reload classes. | |
2378 | ||
2379 | The value is a register class; perhaps @var{class}, or perhaps another, | |
2380 | smaller class. | |
2381 | ||
2382 | Don't define this macro unless the target machine has limitations which | |
2383 | require the macro to do something nontrivial. | |
2384 | ||
2385 | @findex SECONDARY_RELOAD_CLASS | |
2386 | @findex SECONDARY_INPUT_RELOAD_CLASS | |
2387 | @findex SECONDARY_OUTPUT_RELOAD_CLASS | |
2388 | @item SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) | |
2389 | @itemx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) | |
2390 | @itemx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x}) | |
2391 | Many machines have some registers that cannot be copied directly to or | |
2392 | from memory or even from other types of registers. An example is the | |
2393 | @samp{MQ} register, which on most machines, can only be copied to or | |
2394 | from general registers, but not memory. Some machines allow copying all | |
2395 | registers to and from memory, but require a scratch register for stores | |
2396 | to some memory locations (e.g., those with symbolic address on the RT, | |
981f6289 | 2397 | and those with certain symbolic address on the SPARC when compiling |
161d7b59 | 2398 | PIC)@. In some cases, both an intermediate and a scratch register are |
feca2ed3 JW |
2399 | required. |
2400 | ||
2401 | You should define these macros to indicate to the reload phase that it may | |
2402 | need to allocate at least one register for a reload in addition to the | |
2403 | register to contain the data. Specifically, if copying @var{x} to a | |
2404 | register @var{class} in @var{mode} requires an intermediate register, | |
2405 | you should define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the | |
2406 | largest register class all of whose registers can be used as | |
2407 | intermediate registers or scratch registers. | |
2408 | ||
2409 | If copying a register @var{class} in @var{mode} to @var{x} requires an | |
2410 | intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS} | |
2411 | should be defined to return the largest register class required. If the | |
2412 | requirements for input and output reloads are the same, the macro | |
2413 | @code{SECONDARY_RELOAD_CLASS} should be used instead of defining both | |
2414 | macros identically. | |
2415 | ||
2416 | The values returned by these macros are often @code{GENERAL_REGS}. | |
2417 | Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x} | |
2418 | can be directly copied to or from a register of @var{class} in | |
2419 | @var{mode} without requiring a scratch register. Do not define this | |
2420 | macro if it would always return @code{NO_REGS}. | |
2421 | ||
2422 | If a scratch register is required (either with or without an | |
2423 | intermediate register), you should define patterns for | |
2424 | @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required | |
2425 | (@pxref{Standard Names}. These patterns, which will normally be | |
2426 | implemented with a @code{define_expand}, should be similar to the | |
2427 | @samp{mov@var{m}} patterns, except that operand 2 is the scratch | |
2428 | register. | |
2429 | ||
2430 | Define constraints for the reload register and scratch register that | |
2431 | contain a single register class. If the original reload register (whose | |
2432 | class is @var{class}) can meet the constraint given in the pattern, the | |
2433 | value returned by these macros is used for the class of the scratch | |
2434 | register. Otherwise, two additional reload registers are required. | |
2435 | Their classes are obtained from the constraints in the insn pattern. | |
2436 | ||
2437 | @var{x} might be a pseudo-register or a @code{subreg} of a | |
2438 | pseudo-register, which could either be in a hard register or in memory. | |
630d3d5a | 2439 | Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is |
feca2ed3 JW |
2440 | in memory and the hard register number if it is in a register. |
2441 | ||
2442 | These macros should not be used in the case where a particular class of | |
2443 | registers can only be copied to memory and not to another class of | |
2444 | registers. In that case, secondary reload registers are not needed and | |
2445 | would not be helpful. Instead, a stack location must be used to perform | |
a8154559 | 2446 | the copy and the @code{mov@var{m}} pattern should use memory as an |
feca2ed3 JW |
2447 | intermediate storage. This case often occurs between floating-point and |
2448 | general registers. | |
2449 | ||
2450 | @findex SECONDARY_MEMORY_NEEDED | |
2451 | @item SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m}) | |
2452 | Certain machines have the property that some registers cannot be copied | |
2453 | to some other registers without using memory. Define this macro on | |
df2a54e9 | 2454 | those machines to be a C expression that is nonzero if objects of mode |
feca2ed3 JW |
2455 | @var{m} in registers of @var{class1} can only be copied to registers of |
2456 | class @var{class2} by storing a register of @var{class1} into memory | |
2457 | and loading that memory location into a register of @var{class2}. | |
2458 | ||
2459 | Do not define this macro if its value would always be zero. | |
2460 | ||
2461 | @findex SECONDARY_MEMORY_NEEDED_RTX | |
2462 | @item SECONDARY_MEMORY_NEEDED_RTX (@var{mode}) | |
2463 | Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler | |
2464 | allocates a stack slot for a memory location needed for register copies. | |
2465 | If this macro is defined, the compiler instead uses the memory location | |
2466 | defined by this macro. | |
2467 | ||
2468 | Do not define this macro if you do not define | |
2469 | @code{SECONDARY_MEMORY_NEEDED}. | |
2470 | ||
2471 | @findex SECONDARY_MEMORY_NEEDED_MODE | |
2472 | @item SECONDARY_MEMORY_NEEDED_MODE (@var{mode}) | |
2473 | When the compiler needs a secondary memory location to copy between two | |
2474 | registers of mode @var{mode}, it normally allocates sufficient memory to | |
2475 | hold a quantity of @code{BITS_PER_WORD} bits and performs the store and | |
2476 | load operations in a mode that many bits wide and whose class is the | |
2477 | same as that of @var{mode}. | |
2478 | ||
2479 | This is right thing to do on most machines because it ensures that all | |
2480 | bits of the register are copied and prevents accesses to the registers | |
2481 | in a narrower mode, which some machines prohibit for floating-point | |
2482 | registers. | |
2483 | ||
2484 | However, this default behavior is not correct on some machines, such as | |
2485 | the DEC Alpha, that store short integers in floating-point registers | |
2486 | differently than in integer registers. On those machines, the default | |
2487 | widening will not work correctly and you must define this macro to | |
2488 | suppress that widening in some cases. See the file @file{alpha.h} for | |
2489 | details. | |
2490 | ||
2491 | Do not define this macro if you do not define | |
2492 | @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that | |
2493 | is @code{BITS_PER_WORD} bits wide is correct for your machine. | |
2494 | ||
2495 | @findex SMALL_REGISTER_CLASSES | |
2496 | @item SMALL_REGISTER_CLASSES | |
faa9eb19 BS |
2497 | On some machines, it is risky to let hard registers live across arbitrary |
2498 | insns. Typically, these machines have instructions that require values | |
2499 | to be in specific registers (like an accumulator), and reload will fail | |
2500 | if the required hard register is used for another purpose across such an | |
2501 | insn. | |
feca2ed3 | 2502 | |
df2a54e9 JM |
2503 | Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero |
2504 | value on these machines. When this macro has a nonzero value, the | |
faa9eb19 | 2505 | compiler will try to minimize the lifetime of hard registers. |
feca2ed3 | 2506 | |
df2a54e9 | 2507 | It is always safe to define this macro with a nonzero value, but if you |
861bb6c1 JL |
2508 | unnecessarily define it, you will reduce the amount of optimizations |
2509 | that can be performed in some cases. If you do not define this macro | |
df2a54e9 | 2510 | with a nonzero value when it is required, the compiler will run out of |
861bb6c1 JL |
2511 | spill registers and print a fatal error message. For most machines, you |
2512 | should not define this macro at all. | |
feca2ed3 JW |
2513 | |
2514 | @findex CLASS_LIKELY_SPILLED_P | |
2515 | @item CLASS_LIKELY_SPILLED_P (@var{class}) | |
2516 | A C expression whose value is nonzero if pseudos that have been assigned | |
2517 | to registers of class @var{class} would likely be spilled because | |
2518 | registers of @var{class} are needed for spill registers. | |
2519 | ||
2520 | The default value of this macro returns 1 if @var{class} has exactly one | |
2521 | register and zero otherwise. On most machines, this default should be | |
40687a9e | 2522 | used. Only define this macro to some other expression if pseudos |
feca2ed3 JW |
2523 | allocated by @file{local-alloc.c} end up in memory because their hard |
2524 | registers were needed for spill registers. If this macro returns nonzero | |
2525 | for those classes, those pseudos will only be allocated by | |
2526 | @file{global.c}, which knows how to reallocate the pseudo to another | |
2527 | register. If there would not be another register available for | |
2528 | reallocation, you should not change the definition of this macro since | |
2529 | the only effect of such a definition would be to slow down register | |
2530 | allocation. | |
2531 | ||
2532 | @findex CLASS_MAX_NREGS | |
2533 | @item CLASS_MAX_NREGS (@var{class}, @var{mode}) | |
2534 | A C expression for the maximum number of consecutive registers | |
2535 | of class @var{class} needed to hold a value of mode @var{mode}. | |
2536 | ||
2537 | This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact, | |
2538 | the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})} | |
2539 | should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno}, | |
2540 | @var{mode})} for all @var{regno} values in the class @var{class}. | |
2541 | ||
2542 | This macro helps control the handling of multiple-word values | |
2543 | in the reload pass. | |
2544 | ||
02188693 RH |
2545 | @item CLASS_CANNOT_CHANGE_MODE |
2546 | If defined, a C expression for a class that contains registers for | |
2547 | which the compiler may not change modes arbitrarily. | |
2548 | ||
2549 | @item CLASS_CANNOT_CHANGE_MODE_P(@var{from}, @var{to}) | |
2550 | A C expression that is true if, for a register in | |
13ba36b4 | 2551 | @code{CLASS_CANNOT_CHANGE_MODE}, the requested mode punning is invalid. |
feca2ed3 JW |
2552 | |
2553 | For the example, loading 32-bit integer or floating-point objects into | |
57694e40 | 2554 | floating-point registers on the Alpha extends them to 64 bits. |
feca2ed3 | 2555 | Therefore loading a 64-bit object and then storing it as a 32-bit object |
57694e40 | 2556 | does not store the low-order 32 bits, as would be the case for a normal |
02188693 RH |
2557 | register. Therefore, @file{alpha.h} defines @code{CLASS_CANNOT_CHANGE_MODE} |
2558 | as @code{FLOAT_REGS} and @code{CLASS_CANNOT_CHANGE_MODE_P} restricts | |
2559 | mode changes to same-size modes. | |
2560 | ||
2561 | Compare this to IA-64, which extends floating-point values to 82-bits, | |
2562 | and stores 64-bit integers in a different format than 64-bit doubles. | |
2563 | Therefore @code{CLASS_CANNOT_CHANGE_MODE_P} is always true. | |
feca2ed3 JW |
2564 | @end table |
2565 | ||
2566 | Three other special macros describe which operands fit which constraint | |
2567 | letters. | |
2568 | ||
2569 | @table @code | |
2570 | @findex CONST_OK_FOR_LETTER_P | |
2571 | @item CONST_OK_FOR_LETTER_P (@var{value}, @var{c}) | |
e119b68c MM |
2572 | A C expression that defines the machine-dependent operand constraint |
2573 | letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify | |
2574 | particular ranges of integer values. If @var{c} is one of those | |
2575 | letters, the expression should check that @var{value}, an integer, is in | |
2576 | the appropriate range and return 1 if so, 0 otherwise. If @var{c} is | |
2577 | not one of those letters, the value should be 0 regardless of | |
2578 | @var{value}. | |
feca2ed3 JW |
2579 | |
2580 | @findex CONST_DOUBLE_OK_FOR_LETTER_P | |
2581 | @item CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c}) | |
2582 | A C expression that defines the machine-dependent operand constraint | |
e119b68c MM |
2583 | letters that specify particular ranges of @code{const_double} values |
2584 | (@samp{G} or @samp{H}). | |
feca2ed3 JW |
2585 | |
2586 | If @var{c} is one of those letters, the expression should check that | |
2587 | @var{value}, an RTX of code @code{const_double}, is in the appropriate | |
2588 | range and return 1 if so, 0 otherwise. If @var{c} is not one of those | |
2589 | letters, the value should be 0 regardless of @var{value}. | |
2590 | ||
2591 | @code{const_double} is used for all floating-point constants and for | |
2592 | @code{DImode} fixed-point constants. A given letter can accept either | |
2593 | or both kinds of values. It can use @code{GET_MODE} to distinguish | |
2594 | between these kinds. | |
2595 | ||
2596 | @findex EXTRA_CONSTRAINT | |
2597 | @item EXTRA_CONSTRAINT (@var{value}, @var{c}) | |
2598 | A C expression that defines the optional machine-dependent constraint | |
c2cba7a9 RH |
2599 | letters that can be used to segregate specific types of operands, usually |
2600 | memory references, for the target machine. Any letter that is not | |
2601 | elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} | |
2602 | may be used. Normally this macro will not be defined. | |
2603 | ||
2604 | If it is required for a particular target machine, it should return 1 | |
2605 | if @var{value} corresponds to the operand type represented by the | |
2606 | constraint letter @var{c}. If @var{c} is not defined as an extra | |
e119b68c | 2607 | constraint, the value returned should be 0 regardless of @var{value}. |
feca2ed3 | 2608 | |
c2cba7a9 RH |
2609 | For example, on the ROMP, load instructions cannot have their output |
2610 | in r0 if the memory reference contains a symbolic address. Constraint | |
2611 | letter @samp{Q} is defined as representing a memory address that does | |
feca2ed3 JW |
2612 | @emph{not} contain a symbolic address. An alternative is specified with |
2613 | a @samp{Q} constraint on the input and @samp{r} on the output. The next | |
2614 | alternative specifies @samp{m} on the input and a register class that | |
2615 | does not include r0 on the output. | |
ccfc6cc8 UW |
2616 | |
2617 | @findex EXTRA_MEMORY_CONSTRAINT | |
2618 | @item EXTRA_MEMORY_CONSTRAINT (@var{c}) | |
2619 | A C expression that defines the optional machine-dependent constraint | |
2620 | letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should | |
2621 | be treated like memory constraints by the reload pass. | |
2622 | ||
2623 | It should return 1 if the operand type represented by the constraint | |
2624 | letter @var{c} comprises a subset of all memory references including | |
2625 | all those whose address is simply a base register. This allows the reload | |
2626 | pass to reload an operand, if it does not directly correspond to the operand | |
2627 | type of @var{c}, by copying its address into a base register. | |
2628 | ||
2629 | For example, on the S/390, some instructions do not accept arbitrary | |
2630 | memory references, but only those that do not make use of an index | |
2631 | register. The constraint letter @samp{Q} is defined via | |
2632 | @code{EXTRA_CONSTRAINT} as representing a memory address of this type. | |
2633 | If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT}, | |
2634 | a @samp{Q} constraint can handle any memory operand, because the | |
2635 | reload pass knows it can be reloaded by copying the memory address | |
2636 | into a base register if required. This is analogous to the way | |
2637 | a @samp{o} constraint can handle any memory operand. | |
2638 | ||
2639 | @findex EXTRA_ADDRESS_CONSTRAINT | |
2640 | @item EXTRA_ADDRESS_CONSTRAINT (@var{c}) | |
2641 | A C expression that defines the optional machine-dependent constraint | |
2642 | letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should | |
2643 | be treated like address constraints by the reload pass. | |
2644 | ||
2645 | It should return 1 if the operand type represented by the constraint | |
2646 | letter @var{c} comprises a subset of all memory addresses including | |
2647 | all those that consist of just a base register. This allows the reload | |
2648 | pass to reload an operand, if it does not directly correspond to the operand | |
2649 | type of @var{c}, by copying it into a base register. | |
2650 | ||
2651 | Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only | |
2652 | be used with the @code{address_operand} predicate. It is treated | |
2653 | analogously to the @samp{p} constraint. | |
feca2ed3 JW |
2654 | @end table |
2655 | ||
2656 | @node Stack and Calling | |
2657 | @section Stack Layout and Calling Conventions | |
2658 | @cindex calling conventions | |
2659 | ||
2660 | @c prevent bad page break with this line | |
2661 | This describes the stack layout and calling conventions. | |
2662 | ||
2663 | @menu | |
2664 | * Frame Layout:: | |
7c16328b | 2665 | * Exception Handling:: |
861bb6c1 | 2666 | * Stack Checking:: |
feca2ed3 JW |
2667 | * Frame Registers:: |
2668 | * Elimination:: | |
2669 | * Stack Arguments:: | |
2670 | * Register Arguments:: | |
2671 | * Scalar Return:: | |
2672 | * Aggregate Return:: | |
2673 | * Caller Saves:: | |
2674 | * Function Entry:: | |
2675 | * Profiling:: | |
91d231cb | 2676 | * Tail Calls:: |
feca2ed3 JW |
2677 | @end menu |
2678 | ||
2679 | @node Frame Layout | |
2680 | @subsection Basic Stack Layout | |
2681 | @cindex stack frame layout | |
2682 | @cindex frame layout | |
2683 | ||
2684 | @c prevent bad page break with this line | |
2685 | Here is the basic stack layout. | |
2686 | ||
2687 | @table @code | |
2688 | @findex STACK_GROWS_DOWNWARD | |
2689 | @item STACK_GROWS_DOWNWARD | |
2690 | Define this macro if pushing a word onto the stack moves the stack | |
2691 | pointer to a smaller address. | |
2692 | ||
2693 | When we say, ``define this macro if @dots{},'' it means that the | |
2694 | compiler checks this macro only with @code{#ifdef} so the precise | |
2695 | definition used does not matter. | |
2696 | ||
918a6124 GK |
2697 | @findex STACK_PUSH_CODE |
2698 | @item STACK_PUSH_CODE | |
2699 | ||
2700 | This macro defines the operation used when something is pushed | |
2701 | on the stack. In RTL, a push operation will be | |
04a5176a | 2702 | @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})} |
918a6124 GK |
2703 | |
2704 | The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC}, | |
2705 | and @code{POST_INC}. Which of these is correct depends on | |
2706 | the stack direction and on whether the stack pointer points | |
2707 | to the last item on the stack or whether it points to the | |
2708 | space for the next item on the stack. | |
2709 | ||
2710 | The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is | |
2711 | defined, which is almost always right, and @code{PRE_INC} otherwise, | |
2712 | which is often wrong. | |
2713 | ||
feca2ed3 JW |
2714 | @findex FRAME_GROWS_DOWNWARD |
2715 | @item FRAME_GROWS_DOWNWARD | |
2716 | Define this macro if the addresses of local variable slots are at negative | |
2717 | offsets from the frame pointer. | |
2718 | ||
2719 | @findex ARGS_GROW_DOWNWARD | |
2720 | @item ARGS_GROW_DOWNWARD | |
2721 | Define this macro if successive arguments to a function occupy decreasing | |
2722 | addresses on the stack. | |
2723 | ||
2724 | @findex STARTING_FRAME_OFFSET | |
2725 | @item STARTING_FRAME_OFFSET | |
2726 | Offset from the frame pointer to the first local variable slot to be allocated. | |
2727 | ||
2728 | If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by | |
2729 | subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}. | |
2730 | Otherwise, it is found by adding the length of the first slot to the | |
2731 | value @code{STARTING_FRAME_OFFSET}. | |
2732 | @c i'm not sure if the above is still correct.. had to change it to get | |
2733 | @c rid of an overfull. --mew 2feb93 | |
2734 | ||
2735 | @findex STACK_POINTER_OFFSET | |
2736 | @item STACK_POINTER_OFFSET | |
2737 | Offset from the stack pointer register to the first location at which | |
2738 | outgoing arguments are placed. If not specified, the default value of | |
2739 | zero is used. This is the proper value for most machines. | |
2740 | ||
2741 | If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above | |
2742 | the first location at which outgoing arguments are placed. | |
2743 | ||
2744 | @findex FIRST_PARM_OFFSET | |
2745 | @item FIRST_PARM_OFFSET (@var{fundecl}) | |
2746 | Offset from the argument pointer register to the first argument's | |
2747 | address. On some machines it may depend on the data type of the | |
2748 | function. | |
2749 | ||
2750 | If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above | |
2751 | the first argument's address. | |
2752 | ||
2753 | @findex STACK_DYNAMIC_OFFSET | |
2754 | @item STACK_DYNAMIC_OFFSET (@var{fundecl}) | |
2755 | Offset from the stack pointer register to an item dynamically allocated | |
2756 | on the stack, e.g., by @code{alloca}. | |
2757 | ||
2758 | The default value for this macro is @code{STACK_POINTER_OFFSET} plus the | |
2759 | length of the outgoing arguments. The default is correct for most | |
2760 | machines. See @file{function.c} for details. | |
2761 | ||
2762 | @findex DYNAMIC_CHAIN_ADDRESS | |
2763 | @item DYNAMIC_CHAIN_ADDRESS (@var{frameaddr}) | |
2764 | A C expression whose value is RTL representing the address in a stack | |
2765 | frame where the pointer to the caller's frame is stored. Assume that | |
2766 | @var{frameaddr} is an RTL expression for the address of the stack frame | |
2767 | itself. | |
2768 | ||
2769 | If you don't define this macro, the default is to return the value | |
2770 | of @var{frameaddr}---that is, the stack frame address is also the | |
2771 | address of the stack word that points to the previous frame. | |
2772 | ||
2773 | @findex SETUP_FRAME_ADDRESSES | |
0bc02db4 | 2774 | @item SETUP_FRAME_ADDRESSES |
feca2ed3 JW |
2775 | If defined, a C expression that produces the machine-specific code to |
2776 | setup the stack so that arbitrary frames can be accessed. For example, | |
981f6289 | 2777 | on the SPARC, we must flush all of the register windows to the stack |
0bc02db4 MS |
2778 | before we can access arbitrary stack frames. You will seldom need to |
2779 | define this macro. | |
2780 | ||
2781 | @findex BUILTIN_SETJMP_FRAME_VALUE | |
2782 | @item BUILTIN_SETJMP_FRAME_VALUE | |
2783 | If defined, a C expression that contains an rtx that is used to store | |
2784 | the address of the current frame into the built in @code{setjmp} buffer. | |
2785 | The default value, @code{virtual_stack_vars_rtx}, is correct for most | |
2786 | machines. One reason you may need to define this macro is if | |
2787 | @code{hard_frame_pointer_rtx} is the appropriate value on your machine. | |
feca2ed3 JW |
2788 | |
2789 | @findex RETURN_ADDR_RTX | |
2790 | @item RETURN_ADDR_RTX (@var{count}, @var{frameaddr}) | |
2791 | A C expression whose value is RTL representing the value of the return | |
861bb6c1 JL |
2792 | address for the frame @var{count} steps up from the current frame, after |
2793 | the prologue. @var{frameaddr} is the frame pointer of the @var{count} | |
2794 | frame, or the frame pointer of the @var{count} @minus{} 1 frame if | |
feca2ed3 JW |
2795 | @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined. |
2796 | ||
e9a25f70 JL |
2797 | The value of the expression must always be the correct address when |
2798 | @var{count} is zero, but may be @code{NULL_RTX} if there is not way to | |
2799 | determine the return address of other frames. | |
2800 | ||
feca2ed3 JW |
2801 | @findex RETURN_ADDR_IN_PREVIOUS_FRAME |
2802 | @item RETURN_ADDR_IN_PREVIOUS_FRAME | |
2803 | Define this if the return address of a particular stack frame is accessed | |
2804 | from the frame pointer of the previous stack frame. | |
861bb6c1 JL |
2805 | |
2806 | @findex INCOMING_RETURN_ADDR_RTX | |
2807 | @item INCOMING_RETURN_ADDR_RTX | |
2808 | A C expression whose value is RTL representing the location of the | |
2809 | incoming return address at the beginning of any function, before the | |
2810 | prologue. This RTL is either a @code{REG}, indicating that the return | |
2811 | value is saved in @samp{REG}, or a @code{MEM} representing a location in | |
2812 | the stack. | |
2813 | ||
2814 | You only need to define this macro if you want to support call frame | |
2815 | debugging information like that provided by DWARF 2. | |
2816 | ||
2c849145 | 2817 | If this RTL is a @code{REG}, you should also define |
aee96fe9 | 2818 | @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}. |
2c849145 | 2819 | |
861bb6c1 JL |
2820 | @findex INCOMING_FRAME_SP_OFFSET |
2821 | @item INCOMING_FRAME_SP_OFFSET | |
2822 | A C expression whose value is an integer giving the offset, in bytes, | |
2823 | from the value of the stack pointer register to the top of the stack | |
2824 | frame at the beginning of any function, before the prologue. The top of | |
2825 | the frame is defined to be the value of the stack pointer in the | |
2826 | previous frame, just before the call instruction. | |
2827 | ||
71038426 RH |
2828 | You only need to define this macro if you want to support call frame |
2829 | debugging information like that provided by DWARF 2. | |
2830 | ||
2831 | @findex ARG_POINTER_CFA_OFFSET | |
2c849145 | 2832 | @item ARG_POINTER_CFA_OFFSET (@var{fundecl}) |
71038426 RH |
2833 | A C expression whose value is an integer giving the offset, in bytes, |
2834 | from the argument pointer to the canonical frame address (cfa). The | |
02f52e19 | 2835 | final value should coincide with that calculated by |
71038426 RH |
2836 | @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable |
2837 | during virtual register instantiation. | |
2838 | ||
2c849145 JM |
2839 | The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)}, |
2840 | which is correct for most machines; in general, the arguments are found | |
208e52d9 JM |
2841 | immediately before the stack frame. Note that this is not the case on |
2842 | some targets that save registers into the caller's frame, such as SPARC | |
2843 | and rs6000, and so such targets need to define this macro. | |
2c849145 | 2844 | |
208e52d9 | 2845 | You only need to define this macro if the default is incorrect, and you |
2c849145 JM |
2846 | want to support call frame debugging information like that provided by |
2847 | DWARF 2. | |
512b62fb | 2848 | |
7c16328b RH |
2849 | @findex SMALL_STACK |
2850 | @item SMALL_STACK | |
2851 | Define this macro if the stack size for the target is very small. This | |
2852 | has the effect of disabling gcc's built-in @samp{alloca}, though | |
2853 | @samp{__builtin_alloca} is not affected. | |
2854 | @end table | |
2855 | ||
2856 | @node Exception Handling | |
2857 | @subsection Exception Handling Support | |
2858 | @cindex exception handling | |
2859 | ||
2860 | @table @code | |
52a11cbf RH |
2861 | @findex EH_RETURN_DATA_REGNO |
2862 | @item EH_RETURN_DATA_REGNO (@var{N}) | |
2863 | A C expression whose value is the @var{N}th register number used for | |
2864 | data by exception handlers, or @code{INVALID_REGNUM} if fewer than | |
2865 | @var{N} registers are usable. | |
2866 | ||
2867 | The exception handling library routines communicate with the exception | |
2868 | handlers via a set of agreed upon registers. Ideally these registers | |
2869 | should be call-clobbered; it is possible to use call-saved registers, | |
2870 | but may negatively impact code size. The target must support at least | |
2871 | 2 data registers, but should define 4 if there are enough free registers. | |
2872 | ||
2873 | You must define this macro if you want to support call frame exception | |
2874 | handling like that provided by DWARF 2. | |
2875 | ||
2876 | @findex EH_RETURN_STACKADJ_RTX | |
2877 | @item EH_RETURN_STACKADJ_RTX | |
2878 | A C expression whose value is RTL representing a location in which | |
2879 | to store a stack adjustment to be applied before function return. | |
2880 | This is used to unwind the stack to an exception handler's call frame. | |
2881 | It will be assigned zero on code paths that return normally. | |
2882 | ||
02f52e19 | 2883 | Typically this is a call-clobbered hard register that is otherwise |
52a11cbf RH |
2884 | untouched by the epilogue, but could also be a stack slot. |
2885 | ||
2886 | You must define this macro if you want to support call frame exception | |
2887 | handling like that provided by DWARF 2. | |
2888 | ||
2889 | @findex EH_RETURN_HANDLER_RTX | |
2890 | @item EH_RETURN_HANDLER_RTX | |
2891 | A C expression whose value is RTL representing a location in which | |
02f52e19 | 2892 | to store the address of an exception handler to which we should |
52a11cbf RH |
2893 | return. It will not be assigned on code paths that return normally. |
2894 | ||
2895 | Typically this is the location in the call frame at which the normal | |
02f52e19 AJ |
2896 | return address is stored. For targets that return by popping an |
2897 | address off the stack, this might be a memory address just below | |
52a11cbf RH |
2898 | the @emph{target} call frame rather than inside the current call |
2899 | frame. @code{EH_RETURN_STACKADJ_RTX} will have already been assigned, | |
2900 | so it may be used to calculate the location of the target call frame. | |
2901 | ||
2902 | Some targets have more complex requirements than storing to an | |
2903 | address calculable during initial code generation. In that case | |
2904 | the @code{eh_return} instruction pattern should be used instead. | |
2905 | ||
2906 | If you want to support call frame exception handling, you must | |
2907 | define either this macro or the @code{eh_return} instruction pattern. | |
2908 | ||
2a1ee410 | 2909 | @findex ASM_PREFERRED_EH_DATA_FORMAT |
aee96fe9 | 2910 | @item ASM_PREFERRED_EH_DATA_FORMAT(@var{code}, @var{global}) |
2a1ee410 RH |
2911 | This macro chooses the encoding of pointers embedded in the exception |
2912 | handling sections. If at all possible, this should be defined such | |
2913 | that the exception handling section will not require dynamic relocations, | |
2914 | and so may be read-only. | |
2915 | ||
aee96fe9 JM |
2916 | @var{code} is 0 for data, 1 for code labels, 2 for function pointers. |
2917 | @var{global} is true if the symbol may be affected by dynamic relocations. | |
2a1ee410 RH |
2918 | The macro should return a combination of the @code{DW_EH_PE_*} defines |
2919 | as found in @file{dwarf2.h}. | |
2920 | ||
ebb48a4d | 2921 | If this macro is not defined, pointers will not be encoded but |
2a1ee410 RH |
2922 | represented directly. |
2923 | ||
2924 | @findex ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX | |
aee96fe9 | 2925 | @item ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX(@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done}) |
2a1ee410 RH |
2926 | This macro allows the target to emit whatever special magic is required |
2927 | to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}. | |
2928 | Generic code takes care of pc-relative and indirect encodings; this must | |
2929 | be defined if the target uses text-relative or data-relative encodings. | |
2930 | ||
aee96fe9 JM |
2931 | This is a C statement that branches to @var{done} if the format was |
2932 | handled. @var{encoding} is the format chosen, @var{size} is the number | |
2933 | of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF} | |
2a1ee410 RH |
2934 | to be emitted. |
2935 | ||
7c16328b RH |
2936 | @findex MD_FALLBACK_FRAME_STATE_FOR |
2937 | @item MD_FALLBACK_FRAME_STATE_FOR(@var{context}, @var{fs}, @var{success}) | |
2938 | This macro allows the target to add cpu and operating system specific | |
2939 | code to the call-frame unwinder for use when there is no unwind data | |
2940 | available. The most common reason to implement this macro is to unwind | |
2941 | through signal frames. | |
2942 | ||
2943 | This macro is called from @code{uw_frame_state_for} in @file{unwind-dw2.c} | |
2944 | and @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context}; | |
2945 | @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra} | |
2946 | for the address of the code being executed and @code{context->cfa} for | |
2947 | the stack pointer value. If the frame can be decoded, the register save | |
2948 | addresses should be updated in @var{fs} and the macro should branch to | |
f282ffb3 | 2949 | @var{success}. If the frame cannot be decoded, the macro should do |
7c16328b | 2950 | nothing. |
861bb6c1 JL |
2951 | @end table |
2952 | ||
2953 | @node Stack Checking | |
2954 | @subsection Specifying How Stack Checking is Done | |
2955 | ||
a3a15b4d | 2956 | GCC will check that stack references are within the boundaries of |
630d3d5a | 2957 | the stack, if the @option{-fstack-check} is specified, in one of three ways: |
861bb6c1 JL |
2958 | |
2959 | @enumerate | |
2960 | @item | |
a3a15b4d | 2961 | If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC |
861bb6c1 JL |
2962 | will assume that you have arranged for stack checking to be done at |
2963 | appropriate places in the configuration files, e.g., in | |
08c148a8 NB |
2964 | @code{TARGET_ASM_FUNCTION_PROLOGUE}. GCC will do not other special |
2965 | processing. | |
861bb6c1 JL |
2966 | |
2967 | @item | |
2968 | If @code{STACK_CHECK_BUILTIN} is zero and you defined a named pattern | |
a3a15b4d | 2969 | called @code{check_stack} in your @file{md} file, GCC will call that |
861bb6c1 JL |
2970 | pattern with one argument which is the address to compare the stack |
2971 | value against. You must arrange for this pattern to report an error if | |
2972 | the stack pointer is out of range. | |
2973 | ||
2974 | @item | |
a3a15b4d | 2975 | If neither of the above are true, GCC will generate code to periodically |
861bb6c1 JL |
2976 | ``probe'' the stack pointer using the values of the macros defined below. |
2977 | @end enumerate | |
2978 | ||
a3a15b4d | 2979 | Normally, you will use the default values of these macros, so GCC |
861bb6c1 JL |
2980 | will use the third approach. |
2981 | ||
2982 | @table @code | |
2983 | @findex STACK_CHECK_BUILTIN | |
2984 | @item STACK_CHECK_BUILTIN | |
2985 | A nonzero value if stack checking is done by the configuration files in a | |
02f52e19 AJ |
2986 | machine-dependent manner. You should define this macro if stack checking |
2987 | is require by the ABI of your machine or if you would like to have to stack | |
a3a15b4d | 2988 | checking in some more efficient way than GCC's portable approach. |
861bb6c1 JL |
2989 | The default value of this macro is zero. |
2990 | ||
2991 | @findex STACK_CHECK_PROBE_INTERVAL | |
2992 | @item STACK_CHECK_PROBE_INTERVAL | |
a3a15b4d | 2993 | An integer representing the interval at which GCC must generate stack |
861bb6c1 JL |
2994 | probe instructions. You will normally define this macro to be no larger |
2995 | than the size of the ``guard pages'' at the end of a stack area. The | |
2996 | default value of 4096 is suitable for most systems. | |
2997 | ||
2998 | @findex STACK_CHECK_PROBE_LOAD | |
2999 | @item STACK_CHECK_PROBE_LOAD | |
02f52e19 | 3000 | A integer which is nonzero if GCC should perform the stack probe |
a3a15b4d | 3001 | as a load instruction and zero if GCC should use a store instruction. |
861bb6c1 JL |
3002 | The default is zero, which is the most efficient choice on most systems. |
3003 | ||
3004 | @findex STACK_CHECK_PROTECT | |
3005 | @item STACK_CHECK_PROTECT | |
3006 | The number of bytes of stack needed to recover from a stack overflow, | |
3007 | for languages where such a recovery is supported. The default value of | |
3008 | 75 words should be adequate for most machines. | |
3009 | ||
3010 | @findex STACK_CHECK_MAX_FRAME_SIZE | |
3011 | @item STACK_CHECK_MAX_FRAME_SIZE | |
a3a15b4d | 3012 | The maximum size of a stack frame, in bytes. GCC will generate probe |
861bb6c1 JL |
3013 | instructions in non-leaf functions to ensure at least this many bytes of |
3014 | stack are available. If a stack frame is larger than this size, stack | |
a3a15b4d JL |
3015 | checking will not be reliable and GCC will issue a warning. The |
3016 | default is chosen so that GCC only generates one instruction on most | |
861bb6c1 JL |
3017 | systems. You should normally not change the default value of this macro. |
3018 | ||
3019 | @findex STACK_CHECK_FIXED_FRAME_SIZE | |
3020 | @item STACK_CHECK_FIXED_FRAME_SIZE | |
a3a15b4d | 3021 | GCC uses this value to generate the above warning message. It |
861bb6c1 JL |
3022 | represents the amount of fixed frame used by a function, not including |
3023 | space for any callee-saved registers, temporaries and user variables. | |
3024 | You need only specify an upper bound for this amount and will normally | |
3025 | use the default of four words. | |
3026 | ||
3027 | @findex STACK_CHECK_MAX_VAR_SIZE | |
3028 | @item STACK_CHECK_MAX_VAR_SIZE | |
a3a15b4d | 3029 | The maximum size, in bytes, of an object that GCC will place in the |
861bb6c1 | 3030 | fixed area of the stack frame when the user specifies |
630d3d5a | 3031 | @option{-fstack-check}. |
a3a15b4d | 3032 | GCC computed the default from the values of the above macros and you will |
861bb6c1 | 3033 | normally not need to override that default. |
feca2ed3 JW |
3034 | @end table |
3035 | ||
3036 | @need 2000 | |
3037 | @node Frame Registers | |
3038 | @subsection Registers That Address the Stack Frame | |
3039 | ||
3040 | @c prevent bad page break with this line | |
3041 | This discusses registers that address the stack frame. | |
3042 | ||
3043 | @table @code | |
3044 | @findex STACK_POINTER_REGNUM | |
3045 | @item STACK_POINTER_REGNUM | |
3046 | The register number of the stack pointer register, which must also be a | |
3047 | fixed register according to @code{FIXED_REGISTERS}. On most machines, | |
3048 | the hardware determines which register this is. | |
3049 | ||
3050 | @findex FRAME_POINTER_REGNUM | |
3051 | @item FRAME_POINTER_REGNUM | |
3052 | The register number of the frame pointer register, which is used to | |
3053 | access automatic variables in the stack frame. On some machines, the | |
3054 | hardware determines which register this is. On other machines, you can | |
3055 | choose any register you wish for this purpose. | |
3056 | ||
3057 | @findex HARD_FRAME_POINTER_REGNUM | |
3058 | @item HARD_FRAME_POINTER_REGNUM | |
3059 | On some machines the offset between the frame pointer and starting | |
3060 | offset of the automatic variables is not known until after register | |
3061 | allocation has been done (for example, because the saved registers are | |
3062 | between these two locations). On those machines, define | |
3063 | @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to | |
3064 | be used internally until the offset is known, and define | |
556e0f21 | 3065 | @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number |
feca2ed3 JW |
3066 | used for the frame pointer. |
3067 | ||
3068 | You should define this macro only in the very rare circumstances when it | |
3069 | is not possible to calculate the offset between the frame pointer and | |
3070 | the automatic variables until after register allocation has been | |
3071 | completed. When this macro is defined, you must also indicate in your | |
3072 | definition of @code{ELIMINABLE_REGS} how to eliminate | |
3073 | @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM} | |
3074 | or @code{STACK_POINTER_REGNUM}. | |
3075 | ||
3076 | Do not define this macro if it would be the same as | |
3077 | @code{FRAME_POINTER_REGNUM}. | |
3078 | ||
3079 | @findex ARG_POINTER_REGNUM | |
3080 | @item ARG_POINTER_REGNUM | |
3081 | The register number of the arg pointer register, which is used to access | |
3082 | the function's argument list. On some machines, this is the same as the | |
3083 | frame pointer register. On some machines, the hardware determines which | |
3084 | register this is. On other machines, you can choose any register you | |
3085 | wish for this purpose. If this is not the same register as the frame | |
3086 | pointer register, then you must mark it as a fixed register according to | |
3087 | @code{FIXED_REGISTERS}, or arrange to be able to eliminate it | |
3088 | (@pxref{Elimination}). | |
3089 | ||
3090 | @findex RETURN_ADDRESS_POINTER_REGNUM | |
3091 | @item RETURN_ADDRESS_POINTER_REGNUM | |
3092 | The register number of the return address pointer register, which is used to | |
3093 | access the current function's return address from the stack. On some | |
3094 | machines, the return address is not at a fixed offset from the frame | |
3095 | pointer or stack pointer or argument pointer. This register can be defined | |
3096 | to point to the return address on the stack, and then be converted by | |
3097 | @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer. | |
3098 | ||
3099 | Do not define this macro unless there is no other way to get the return | |
3100 | address from the stack. | |
3101 | ||
3102 | @findex STATIC_CHAIN_REGNUM | |
3103 | @findex STATIC_CHAIN_INCOMING_REGNUM | |
3104 | @item STATIC_CHAIN_REGNUM | |
3105 | @itemx STATIC_CHAIN_INCOMING_REGNUM | |
3106 | Register numbers used for passing a function's static chain pointer. If | |
3107 | register windows are used, the register number as seen by the called | |
3108 | function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register | |
3109 | number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If | |
3110 | these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need | |
bd819a4a | 3111 | not be defined. |
feca2ed3 JW |
3112 | |
3113 | The static chain register need not be a fixed register. | |
3114 | ||
3115 | If the static chain is passed in memory, these macros should not be | |
3116 | defined; instead, the next two macros should be defined. | |
3117 | ||
3118 | @findex STATIC_CHAIN | |
3119 | @findex STATIC_CHAIN_INCOMING | |
3120 | @item STATIC_CHAIN | |
3121 | @itemx STATIC_CHAIN_INCOMING | |
3122 | If the static chain is passed in memory, these macros provide rtx giving | |
3123 | @code{mem} expressions that denote where they are stored. | |
3124 | @code{STATIC_CHAIN} and @code{STATIC_CHAIN_INCOMING} give the locations | |
3125 | as seen by the calling and called functions, respectively. Often the former | |
3126 | will be at an offset from the stack pointer and the latter at an offset from | |
bd819a4a | 3127 | the frame pointer. |
feca2ed3 JW |
3128 | |
3129 | @findex stack_pointer_rtx | |
3130 | @findex frame_pointer_rtx | |
3131 | @findex arg_pointer_rtx | |
3132 | The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and | |
3133 | @code{arg_pointer_rtx} will have been initialized prior to the use of these | |
3134 | macros and should be used to refer to those items. | |
3135 | ||
3136 | If the static chain is passed in a register, the two previous macros should | |
3137 | be defined instead. | |
919543ab AH |
3138 | |
3139 | @findex DWARF_FRAME_REGISTERS | |
3140 | @item DWARF_FRAME_REGISTERS | |
3141 | This macro specifies the maximum number of hard registers that can be | |
3142 | saved in a call frame. This is used to size data structures used in | |
3143 | DWARF2 exception handling. | |
3144 | ||
3145 | Prior to GCC 3.0, this macro was needed in order to establish a stable | |
3146 | exception handling ABI in the face of adding new hard registers for ISA | |
3147 | extensions. In GCC 3.0 and later, the EH ABI is insulated from changes | |
3148 | in the number of hard registers. Nevertheless, this macro can still be | |
3149 | used to reduce the runtime memory requirements of the exception handling | |
3150 | routines, which can be substantial if the ISA contains a lot of | |
3151 | registers that are not call-saved. | |
3152 | ||
3153 | If this macro is not defined, it defaults to | |
3154 | @code{FIRST_PSEUDO_REGISTER}. | |
3155 | ||
3156 | @findex PRE_GCC3_DWARF_FRAME_REGISTERS | |
3157 | @item PRE_GCC3_DWARF_FRAME_REGISTERS | |
3158 | ||
3159 | This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided | |
3160 | for backward compatibility in pre GCC 3.0 compiled code. | |
3161 | ||
3162 | If this macro is not defined, it defaults to | |
3163 | @code{DWARF_FRAME_REGISTERS}. | |
3164 | ||
feca2ed3 JW |
3165 | @end table |
3166 | ||
3167 | @node Elimination | |
3168 | @subsection Eliminating Frame Pointer and Arg Pointer | |
3169 | ||
3170 | @c prevent bad page break with this line | |
3171 | This is about eliminating the frame pointer and arg pointer. | |
3172 | ||
3173 | @table @code | |
3174 | @findex FRAME_POINTER_REQUIRED | |
3175 | @item FRAME_POINTER_REQUIRED | |
3176 | A C expression which is nonzero if a function must have and use a frame | |
3177 | pointer. This expression is evaluated in the reload pass. If its value is | |
3178 | nonzero the function will have a frame pointer. | |
3179 | ||
3180 | The expression can in principle examine the current function and decide | |
3181 | according to the facts, but on most machines the constant 0 or the | |
3182 | constant 1 suffices. Use 0 when the machine allows code to be generated | |
3183 | with no frame pointer, and doing so saves some time or space. Use 1 | |
3184 | when there is no possible advantage to avoiding a frame pointer. | |
3185 | ||
3186 | In certain cases, the compiler does not know how to produce valid code | |
3187 | without a frame pointer. The compiler recognizes those cases and | |
3188 | automatically gives the function a frame pointer regardless of what | |
3189 | @code{FRAME_POINTER_REQUIRED} says. You don't need to worry about | |
bd819a4a | 3190 | them. |
feca2ed3 JW |
3191 | |
3192 | In a function that does not require a frame pointer, the frame pointer | |
3193 | register can be allocated for ordinary usage, unless you mark it as a | |
3194 | fixed register. See @code{FIXED_REGISTERS} for more information. | |
3195 | ||
3196 | @findex INITIAL_FRAME_POINTER_OFFSET | |
3197 | @findex get_frame_size | |
3198 | @item INITIAL_FRAME_POINTER_OFFSET (@var{depth-var}) | |
3199 | A C statement to store in the variable @var{depth-var} the difference | |
3200 | between the frame pointer and the stack pointer values immediately after | |
3201 | the function prologue. The value would be computed from information | |
3202 | such as the result of @code{get_frame_size ()} and the tables of | |
3203 | registers @code{regs_ever_live} and @code{call_used_regs}. | |
3204 | ||
3205 | If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and | |
3206 | need not be defined. Otherwise, it must be defined even if | |
3207 | @code{FRAME_POINTER_REQUIRED} is defined to always be true; in that | |
3208 | case, you may set @var{depth-var} to anything. | |
3209 | ||
3210 | @findex ELIMINABLE_REGS | |
3211 | @item ELIMINABLE_REGS | |
3212 | If defined, this macro specifies a table of register pairs used to | |
3213 | eliminate unneeded registers that point into the stack frame. If it is not | |
3214 | defined, the only elimination attempted by the compiler is to replace | |
3215 | references to the frame pointer with references to the stack pointer. | |
3216 | ||
3217 | The definition of this macro is a list of structure initializations, each | |
3218 | of which specifies an original and replacement register. | |
3219 | ||
3220 | On some machines, the position of the argument pointer is not known until | |
3221 | the compilation is completed. In such a case, a separate hard register | |
3222 | must be used for the argument pointer. This register can be eliminated by | |
3223 | replacing it with either the frame pointer or the argument pointer, | |
3224 | depending on whether or not the frame pointer has been eliminated. | |
3225 | ||
3226 | In this case, you might specify: | |
3227 | @example | |
3228 | #define ELIMINABLE_REGS \ | |
3229 | @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \ | |
3230 | @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \ | |
3231 | @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@} | |
3232 | @end example | |
3233 | ||
3234 | Note that the elimination of the argument pointer with the stack pointer is | |
3235 | specified first since that is the preferred elimination. | |
3236 | ||
3237 | @findex CAN_ELIMINATE | |
3238 | @item CAN_ELIMINATE (@var{from-reg}, @var{to-reg}) | |
df2a54e9 | 3239 | A C expression that returns nonzero if the compiler is allowed to try |
feca2ed3 JW |
3240 | to replace register number @var{from-reg} with register number |
3241 | @var{to-reg}. This macro need only be defined if @code{ELIMINABLE_REGS} | |
3242 | is defined, and will usually be the constant 1, since most of the cases | |
3243 | preventing register elimination are things that the compiler already | |
3244 | knows about. | |
3245 | ||
3246 | @findex INITIAL_ELIMINATION_OFFSET | |
3247 | @item INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var}) | |
3248 | This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It | |
3249 | specifies the initial difference between the specified pair of | |
3250 | registers. This macro must be defined if @code{ELIMINABLE_REGS} is | |
3251 | defined. | |
feca2ed3 JW |
3252 | @end table |
3253 | ||
3254 | @node Stack Arguments | |
3255 | @subsection Passing Function Arguments on the Stack | |
3256 | @cindex arguments on stack | |
3257 | @cindex stack arguments | |
3258 | ||
3259 | The macros in this section control how arguments are passed | |
3260 | on the stack. See the following section for other macros that | |
3261 | control passing certain arguments in registers. | |
3262 | ||
3263 | @table @code | |
3264 | @findex PROMOTE_PROTOTYPES | |
3265 | @item PROMOTE_PROTOTYPES | |
7d473569 JJ |
3266 | A C expression whose value is nonzero if an argument declared in |
3267 | a prototype as an integral type smaller than @code{int} should | |
3268 | actually be passed as an @code{int}. In addition to avoiding | |
3269 | errors in certain cases of mismatch, it also makes for better | |
3270 | code on certain machines. If the macro is not defined in target | |
3271 | header files, it defaults to 0. | |
feca2ed3 | 3272 | |
f73ad30e JH |
3273 | @findex PUSH_ARGS |
3274 | @item PUSH_ARGS | |
767094dd | 3275 | A C expression. If nonzero, push insns will be used to pass |
f73ad30e JH |
3276 | outgoing arguments. |
3277 | If the target machine does not have a push instruction, set it to zero. | |
3278 | That directs GCC to use an alternate strategy: to | |
3279 | allocate the entire argument block and then store the arguments into | |
aee96fe9 | 3280 | it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too. |
f73ad30e JH |
3281 | On some machines, the definition |
3282 | ||
feca2ed3 JW |
3283 | @findex PUSH_ROUNDING |
3284 | @item PUSH_ROUNDING (@var{npushed}) | |
3285 | A C expression that is the number of bytes actually pushed onto the | |
3286 | stack when an instruction attempts to push @var{npushed} bytes. | |
feca2ed3 JW |
3287 | |
3288 | On some machines, the definition | |
3289 | ||
3290 | @example | |
3291 | #define PUSH_ROUNDING(BYTES) (BYTES) | |
3292 | @end example | |
3293 | ||
3294 | @noindent | |
3295 | will suffice. But on other machines, instructions that appear | |
3296 | to push one byte actually push two bytes in an attempt to maintain | |
3297 | alignment. Then the definition should be | |
3298 | ||
3299 | @example | |
3300 | #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1) | |
3301 | @end example | |
3302 | ||
3303 | @findex ACCUMULATE_OUTGOING_ARGS | |
3304 | @findex current_function_outgoing_args_size | |
3305 | @item ACCUMULATE_OUTGOING_ARGS | |
767094dd | 3306 | A C expression. If nonzero, the maximum amount of space required for outgoing arguments |
feca2ed3 JW |
3307 | will be computed and placed into the variable |
3308 | @code{current_function_outgoing_args_size}. No space will be pushed | |
3309 | onto the stack for each call; instead, the function prologue should | |
3310 | increase the stack frame size by this amount. | |
3311 | ||
f73ad30e | 3312 | Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS} |
feca2ed3 JW |
3313 | is not proper. |
3314 | ||
3315 | @findex REG_PARM_STACK_SPACE | |
3316 | @item REG_PARM_STACK_SPACE (@var{fndecl}) | |
3317 | Define this macro if functions should assume that stack space has been | |
3318 | allocated for arguments even when their values are passed in | |
3319 | registers. | |
3320 | ||
3321 | The value of this macro is the size, in bytes, of the area reserved for | |
ab87f8c8 | 3322 | arguments passed in registers for the function represented by @var{fndecl}, |
a3a15b4d | 3323 | which can be zero if GCC is calling a library function. |
feca2ed3 JW |
3324 | |
3325 | This space can be allocated by the caller, or be a part of the | |
3326 | machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says | |
3327 | which. | |
3328 | @c above is overfull. not sure what to do. --mew 5feb93 did | |
3329 | @c something, not sure if it looks good. --mew 10feb93 | |
3330 | ||
3331 | @findex MAYBE_REG_PARM_STACK_SPACE | |
3332 | @findex FINAL_REG_PARM_STACK_SPACE | |
3333 | @item MAYBE_REG_PARM_STACK_SPACE | |
3334 | @itemx FINAL_REG_PARM_STACK_SPACE (@var{const_size}, @var{var_size}) | |
3335 | Define these macros in addition to the one above if functions might | |
3336 | allocate stack space for arguments even when their values are passed | |
3337 | in registers. These should be used when the stack space allocated | |
3338 | for arguments in registers is not a simple constant independent of the | |
3339 | function declaration. | |
3340 | ||
3341 | The value of the first macro is the size, in bytes, of the area that | |
3342 | we should initially assume would be reserved for arguments passed in registers. | |
3343 | ||
3344 | The value of the second macro is the actual size, in bytes, of the area | |
3345 | that will be reserved for arguments passed in registers. This takes two | |
3346 | arguments: an integer representing the number of bytes of fixed sized | |
3347 | arguments on the stack, and a tree representing the number of bytes of | |
3348 | variable sized arguments on the stack. | |
3349 | ||
3350 | When these macros are defined, @code{REG_PARM_STACK_SPACE} will only be | |
3351 | called for libcall functions, the current function, or for a function | |
3352 | being called when it is known that such stack space must be allocated. | |
3353 | In each case this value can be easily computed. | |
3354 | ||
3355 | When deciding whether a called function needs such stack space, and how | |
a3a15b4d | 3356 | much space to reserve, GCC uses these two macros instead of |
feca2ed3 JW |
3357 | @code{REG_PARM_STACK_SPACE}. |
3358 | ||
3359 | @findex OUTGOING_REG_PARM_STACK_SPACE | |
3360 | @item OUTGOING_REG_PARM_STACK_SPACE | |
3361 | Define this if it is the responsibility of the caller to allocate the area | |
3362 | reserved for arguments passed in registers. | |
3363 | ||
3364 | If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls | |
3365 | whether the space for these arguments counts in the value of | |
3366 | @code{current_function_outgoing_args_size}. | |
3367 | ||
3368 | @findex STACK_PARMS_IN_REG_PARM_AREA | |
3369 | @item STACK_PARMS_IN_REG_PARM_AREA | |
3370 | Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the | |
3371 | stack parameters don't skip the area specified by it. | |
3372 | @c i changed this, makes more sens and it should have taken care of the | |
3373 | @c overfull.. not as specific, tho. --mew 5feb93 | |
3374 | ||
3375 | Normally, when a parameter is not passed in registers, it is placed on the | |
3376 | stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro | |
3377 | suppresses this behavior and causes the parameter to be passed on the | |
3378 | stack in its natural location. | |
3379 | ||
3380 | @findex RETURN_POPS_ARGS | |
3381 | @item RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size}) | |
3382 | A C expression that should indicate the number of bytes of its own | |
3383 | arguments that a function pops on returning, or 0 if the | |
3384 | function pops no arguments and the caller must therefore pop them all | |
3385 | after the function returns. | |
3386 | ||
3387 | @var{fundecl} is a C variable whose value is a tree node that describes | |
3388 | the function in question. Normally it is a node of type | |
3389 | @code{FUNCTION_DECL} that describes the declaration of the function. | |
91d231cb | 3390 | From this you can obtain the @code{DECL_ATTRIBUTES} of the function. |
feca2ed3 JW |
3391 | |
3392 | @var{funtype} is a C variable whose value is a tree node that | |
3393 | describes the function in question. Normally it is a node of type | |
3394 | @code{FUNCTION_TYPE} that describes the data type of the function. | |
3395 | From this it is possible to obtain the data types of the value and | |
3396 | arguments (if known). | |
3397 | ||
861bb6c1 | 3398 | When a call to a library function is being considered, @var{fundecl} |
feca2ed3 JW |
3399 | will contain an identifier node for the library function. Thus, if |
3400 | you need to distinguish among various library functions, you can do so | |
3401 | by their names. Note that ``library function'' in this context means | |
3402 | a function used to perform arithmetic, whose name is known specially | |
3403 | in the compiler and was not mentioned in the C code being compiled. | |
3404 | ||
3405 | @var{stack-size} is the number of bytes of arguments passed on the | |
3406 | stack. If a variable number of bytes is passed, it is zero, and | |
3407 | argument popping will always be the responsibility of the calling function. | |
3408 | ||
8aeea6e6 | 3409 | On the VAX, all functions always pop their arguments, so the definition |
feca2ed3 JW |
3410 | of this macro is @var{stack-size}. On the 68000, using the standard |
3411 | calling convention, no functions pop their arguments, so the value of | |
3412 | the macro is always 0 in this case. But an alternative calling | |
3413 | convention is available in which functions that take a fixed number of | |
3414 | arguments pop them but other functions (such as @code{printf}) pop | |
3415 | nothing (the caller pops all). When this convention is in use, | |
3416 | @var{funtype} is examined to determine whether a function takes a fixed | |
3417 | number of arguments. | |
fa5322fa AO |
3418 | |
3419 | @findex CALL_POPS_ARGS | |
3420 | @item CALL_POPS_ARGS (@var{cum}) | |
3421 | A C expression that should indicate the number of bytes a call sequence | |
3422 | pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS} | |
3423 | when compiling a function call. | |
3424 | ||
3425 | @var{cum} is the variable in which all arguments to the called function | |
3426 | have been accumulated. | |
3427 | ||
3428 | On certain architectures, such as the SH5, a call trampoline is used | |
3429 | that pops certain registers off the stack, depending on the arguments | |
3430 | that have been passed to the function. Since this is a property of the | |
3431 | call site, not of the called function, @code{RETURN_POPS_ARGS} is not | |
3432 | appropriate. | |
3433 | ||
feca2ed3 JW |
3434 | @end table |
3435 | ||
3436 | @node Register Arguments | |
3437 | @subsection Passing Arguments in Registers | |
3438 | @cindex arguments in registers | |
3439 | @cindex registers arguments | |
3440 | ||
3441 | This section describes the macros which let you control how various | |
3442 | types of arguments are passed in registers or how they are arranged in | |
3443 | the stack. | |
3444 | ||
3445 | @table @code | |
3446 | @findex FUNCTION_ARG | |
3447 | @item FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
3448 | A C expression that controls whether a function argument is passed | |
3449 | in a register, and which register. | |
3450 | ||
3451 | The arguments are @var{cum}, which summarizes all the previous | |
3452 | arguments; @var{mode}, the machine mode of the argument; @var{type}, | |
3453 | the data type of the argument as a tree node or 0 if that is not known | |
3454 | (which happens for C support library functions); and @var{named}, | |
3455 | which is 1 for an ordinary argument and 0 for nameless arguments that | |
3456 | correspond to @samp{@dots{}} in the called function's prototype. | |
3719d27b JO |
3457 | @var{type} can be an incomplete type if a syntax error has previously |
3458 | occurred. | |
feca2ed3 JW |
3459 | |
3460 | The value of the expression is usually either a @code{reg} RTX for the | |
3461 | hard register in which to pass the argument, or zero to pass the | |
3462 | argument on the stack. | |
3463 | ||
8aeea6e6 | 3464 | For machines like the VAX and 68000, where normally all arguments are |
feca2ed3 JW |
3465 | pushed, zero suffices as a definition. |
3466 | ||
161d7b59 | 3467 | The value of the expression can also be a @code{parallel} RTX@. This is |
feca2ed3 JW |
3468 | used when an argument is passed in multiple locations. The mode of the |
3469 | of the @code{parallel} should be the mode of the entire argument. The | |
3470 | @code{parallel} holds any number of @code{expr_list} pairs; each one | |
f797c10b NC |
3471 | describes where part of the argument is passed. In each |
3472 | @code{expr_list} the first operand must be a @code{reg} RTX for the hard | |
3473 | register in which to pass this part of the argument, and the mode of the | |
3474 | register RTX indicates how large this part of the argument is. The | |
3475 | second operand of the @code{expr_list} is a @code{const_int} which gives | |
3476 | the offset in bytes into the entire argument of where this part starts. | |
02f52e19 | 3477 | As a special exception the first @code{expr_list} in the @code{parallel} |
c980b85b NC |
3478 | RTX may have a first operand of zero. This indicates that the entire |
3479 | argument is also stored on the stack. | |
feca2ed3 | 3480 | |
1cc5e432 GK |
3481 | The last time this macro is called, it is called with @code{MODE == |
3482 | VOIDmode}, and its result is passed to the @code{call} or @code{call_value} | |
3483 | pattern as operands 2 and 3 respectively. | |
3484 | ||
feca2ed3 | 3485 | @cindex @file{stdarg.h} and register arguments |
5490d604 | 3486 | The usual way to make the ISO library @file{stdarg.h} work on a machine |
feca2ed3 JW |
3487 | where some arguments are usually passed in registers, is to cause |
3488 | nameless arguments to be passed on the stack instead. This is done | |
3489 | by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0. | |
3490 | ||
3491 | @cindex @code{MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG} | |
3492 | @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG} | |
3493 | You may use the macro @code{MUST_PASS_IN_STACK (@var{mode}, @var{type})} | |
3494 | in the definition of this macro to determine if this argument is of a | |
3495 | type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE} | |
df2a54e9 | 3496 | is not defined and @code{FUNCTION_ARG} returns nonzero for such an |
feca2ed3 JW |
3497 | argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is |
3498 | defined, the argument will be computed in the stack and then loaded into | |
3499 | a register. | |
3500 | ||
d9a4ee00 JL |
3501 | @findex MUST_PASS_IN_STACK |
3502 | @item MUST_PASS_IN_STACK (@var{mode}, @var{type}) | |
3503 | Define as a C expression that evaluates to nonzero if we do not know how | |
3504 | to pass TYPE solely in registers. The file @file{expr.h} defines a | |
3505 | definition that is usually appropriate, refer to @file{expr.h} for additional | |
3506 | documentation. | |
3507 | ||
feca2ed3 JW |
3508 | @findex FUNCTION_INCOMING_ARG |
3509 | @item FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
3510 | Define this macro if the target machine has ``register windows'', so | |
3511 | that the register in which a function sees an arguments is not | |
3512 | necessarily the same as the one in which the caller passed the | |
3513 | argument. | |
3514 | ||
3515 | For such machines, @code{FUNCTION_ARG} computes the register in which | |
3516 | the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should | |
3517 | be defined in a similar fashion to tell the function being called | |
3518 | where the arguments will arrive. | |
3519 | ||
3520 | If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG} | |
bd819a4a | 3521 | serves both purposes. |
feca2ed3 JW |
3522 | |
3523 | @findex FUNCTION_ARG_PARTIAL_NREGS | |
3524 | @item FUNCTION_ARG_PARTIAL_NREGS (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
3525 | A C expression for the number of words, at the beginning of an | |
6b72173a | 3526 | argument, that must be put in registers. The value must be zero for |
feca2ed3 JW |
3527 | arguments that are passed entirely in registers or that are entirely |
3528 | pushed on the stack. | |
3529 | ||
3530 | On some machines, certain arguments must be passed partially in | |
3531 | registers and partially in memory. On these machines, typically the | |
3532 | first @var{n} words of arguments are passed in registers, and the rest | |
3533 | on the stack. If a multi-word argument (a @code{double} or a | |
3534 | structure) crosses that boundary, its first few words must be passed | |
3535 | in registers and the rest must be pushed. This macro tells the | |
3536 | compiler when this occurs, and how many of the words should go in | |
3537 | registers. | |
3538 | ||
3539 | @code{FUNCTION_ARG} for these arguments should return the first | |
3540 | register to be used by the caller for this argument; likewise | |
3541 | @code{FUNCTION_INCOMING_ARG}, for the called function. | |
3542 | ||
3543 | @findex FUNCTION_ARG_PASS_BY_REFERENCE | |
3544 | @item FUNCTION_ARG_PASS_BY_REFERENCE (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
3545 | A C expression that indicates when an argument must be passed by reference. | |
3546 | If nonzero for an argument, a copy of that argument is made in memory and a | |
3547 | pointer to the argument is passed instead of the argument itself. | |
3548 | The pointer is passed in whatever way is appropriate for passing a pointer | |
3549 | to that type. | |
3550 | ||
3551 | On machines where @code{REG_PARM_STACK_SPACE} is not defined, a suitable | |
3552 | definition of this macro might be | |
3553 | @smallexample | |
3554 | #define FUNCTION_ARG_PASS_BY_REFERENCE\ | |
3555 | (CUM, MODE, TYPE, NAMED) \ | |
3556 | MUST_PASS_IN_STACK (MODE, TYPE) | |
3557 | @end smallexample | |
3558 | @c this is *still* too long. --mew 5feb93 | |
3559 | ||
3560 | @findex FUNCTION_ARG_CALLEE_COPIES | |
3561 | @item FUNCTION_ARG_CALLEE_COPIES (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
3562 | If defined, a C expression that indicates when it is the called function's | |
3563 | responsibility to make a copy of arguments passed by invisible reference. | |
3564 | Normally, the caller makes a copy and passes the address of the copy to the | |
aee96fe9 | 3565 | routine being called. When @code{FUNCTION_ARG_CALLEE_COPIES} is defined and is |
feca2ed3 JW |
3566 | nonzero, the caller does not make a copy. Instead, it passes a pointer to the |
3567 | ``live'' value. The called function must not modify this value. If it can be | |
3568 | determined that the value won't be modified, it need not make a copy; | |
3569 | otherwise a copy must be made. | |
3570 | ||
0d7839da SE |
3571 | @findex FUNCTION_ARG_REG_LITTLE_ENDIAN |
3572 | @item FUNCTION_ARG_REG_LITTLE_ENDIAN | |
3573 | If defined TRUE on a big-endian system then structure arguments passed | |
3574 | (and returned) in registers are passed in a little-endian manner instead of | |
3575 | the big-endian manner. On the HP-UX IA64 and PA64 platforms structures are | |
3576 | aligned differently then integral values and setting this value to true will | |
3577 | allow for the special handling of structure arguments and return values. | |
3578 | ||
feca2ed3 JW |
3579 | @findex CUMULATIVE_ARGS |
3580 | @item CUMULATIVE_ARGS | |
3581 | A C type for declaring a variable that is used as the first argument of | |
3582 | @code{FUNCTION_ARG} and other related values. For some target machines, | |
3583 | the type @code{int} suffices and can hold the number of bytes of | |
3584 | argument so far. | |
3585 | ||
3586 | There is no need to record in @code{CUMULATIVE_ARGS} anything about the | |
3587 | arguments that have been passed on the stack. The compiler has other | |
3588 | variables to keep track of that. For target machines on which all | |
3589 | arguments are passed on the stack, there is no need to store anything in | |
3590 | @code{CUMULATIVE_ARGS}; however, the data structure must exist and | |
3591 | should not be empty, so use @code{int}. | |
3592 | ||
3593 | @findex INIT_CUMULATIVE_ARGS | |
3594 | @item INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{indirect}) | |
3595 | A C statement (sans semicolon) for initializing the variable @var{cum} | |
3596 | for the state at the beginning of the argument list. The variable has | |
3597 | type @code{CUMULATIVE_ARGS}. The value of @var{fntype} is the tree node | |
3598 | for the data type of the function which will receive the args, or 0 | |
3599 | if the args are to a compiler support library function. The value of | |
3600 | @var{indirect} is nonzero when processing an indirect call, for example | |
3601 | a call through a function pointer. The value of @var{indirect} is zero | |
3602 | for a call to an explicitly named function, a library function call, or when | |
3603 | @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function | |
3604 | being compiled. | |
3605 | ||
3606 | When processing a call to a compiler support library function, | |
3607 | @var{libname} identifies which one. It is a @code{symbol_ref} rtx which | |
3608 | contains the name of the function, as a string. @var{libname} is 0 when | |
3609 | an ordinary C function call is being processed. Thus, each time this | |
3610 | macro is called, either @var{libname} or @var{fntype} is nonzero, but | |
3611 | never both of them at once. | |
3612 | ||
97fc4caf AO |
3613 | @findex INIT_CUMULATIVE_LIBCALL_ARGS |
3614 | @item INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname}) | |
3615 | Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls, | |
3616 | it gets a @code{MODE} argument instead of @var{fntype}, that would be | |
3617 | @code{NULL}. @var{indirect} would always be zero, too. If this macro | |
3618 | is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname, | |
3619 | 0)} is used instead. | |
3620 | ||
feca2ed3 JW |
3621 | @findex INIT_CUMULATIVE_INCOMING_ARGS |
3622 | @item INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname}) | |
3623 | Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of | |
3624 | finding the arguments for the function being compiled. If this macro is | |
3625 | undefined, @code{INIT_CUMULATIVE_ARGS} is used instead. | |
3626 | ||
3627 | The value passed for @var{libname} is always 0, since library routines | |
161d7b59 | 3628 | with special calling conventions are never compiled with GCC@. The |
feca2ed3 JW |
3629 | argument @var{libname} exists for symmetry with |
3630 | @code{INIT_CUMULATIVE_ARGS}. | |
3631 | @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe. | |
3632 | @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93 | |
3633 | ||
3634 | @findex FUNCTION_ARG_ADVANCE | |
3635 | @item FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named}) | |
3636 | A C statement (sans semicolon) to update the summarizer variable | |
3637 | @var{cum} to advance past an argument in the argument list. The | |
3638 | values @var{mode}, @var{type} and @var{named} describe that argument. | |
3639 | Once this is done, the variable @var{cum} is suitable for analyzing | |
bd819a4a | 3640 | the @emph{following} argument with @code{FUNCTION_ARG}, etc. |
feca2ed3 JW |
3641 | |
3642 | This macro need not do anything if the argument in question was passed | |
3643 | on the stack. The compiler knows how to track the amount of stack space | |
3644 | used for arguments without any special help. | |
3645 | ||
3646 | @findex FUNCTION_ARG_PADDING | |
3647 | @item FUNCTION_ARG_PADDING (@var{mode}, @var{type}) | |
3648 | If defined, a C expression which determines whether, and in which direction, | |
3649 | to pad out an argument with extra space. The value should be of type | |
3650 | @code{enum direction}: either @code{upward} to pad above the argument, | |
3651 | @code{downward} to pad below, or @code{none} to inhibit padding. | |
3652 | ||
3653 | The @emph{amount} of padding is always just enough to reach the next | |
3654 | multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control | |
3655 | it. | |
3656 | ||
3657 | This macro has a default definition which is right for most systems. | |
3658 | For little-endian machines, the default is to pad upward. For | |
3659 | big-endian machines, the default is to pad downward for an argument of | |
3660 | constant size shorter than an @code{int}, and upward otherwise. | |
3661 | ||
5e4f6244 CP |
3662 | @findex PAD_VARARGS_DOWN |
3663 | @item PAD_VARARGS_DOWN | |
02f52e19 AJ |
3664 | If defined, a C expression which determines whether the default |
3665 | implementation of va_arg will attempt to pad down before reading the | |
5e4f6244 CP |
3666 | next argument, if that argument is smaller than its aligned space as |
3667 | controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such | |
3668 | arguments are padded down if @code{BYTES_BIG_ENDIAN} is true. | |
3669 | ||
feca2ed3 JW |
3670 | @findex FUNCTION_ARG_BOUNDARY |
3671 | @item FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type}) | |
3672 | If defined, a C expression that gives the alignment boundary, in bits, | |
3673 | of an argument with the specified mode and type. If it is not defined, | |
3674 | @code{PARM_BOUNDARY} is used for all arguments. | |
3675 | ||
3676 | @findex FUNCTION_ARG_REGNO_P | |
3677 | @item FUNCTION_ARG_REGNO_P (@var{regno}) | |
3678 | A C expression that is nonzero if @var{regno} is the number of a hard | |
3679 | register in which function arguments are sometimes passed. This does | |
3680 | @emph{not} include implicit arguments such as the static chain and | |
3681 | the structure-value address. On many machines, no registers can be | |
3682 | used for this purpose since all function arguments are pushed on the | |
3683 | stack. | |
bb1b857a GK |
3684 | |
3685 | @findex LOAD_ARGS_REVERSED | |
3686 | @item LOAD_ARGS_REVERSED | |
3687 | If defined, the order in which arguments are loaded into their | |
02f52e19 | 3688 | respective argument registers is reversed so that the last |
4e5f1329 | 3689 | argument is loaded first. This macro only affects arguments |
bb1b857a GK |
3690 | passed in registers. |
3691 | ||
feca2ed3 JW |
3692 | @end table |
3693 | ||
3694 | @node Scalar Return | |
3695 | @subsection How Scalar Function Values Are Returned | |
3696 | @cindex return values in registers | |
3697 | @cindex values, returned by functions | |
3698 | @cindex scalars, returned as values | |
3699 | ||
3700 | This section discusses the macros that control returning scalars as | |
3701 | values---values that can fit in registers. | |
3702 | ||
3703 | @table @code | |
feca2ed3 JW |
3704 | @findex FUNCTION_VALUE |
3705 | @item FUNCTION_VALUE (@var{valtype}, @var{func}) | |
3706 | A C expression to create an RTX representing the place where a | |
3707 | function returns a value of data type @var{valtype}. @var{valtype} is | |
3708 | a tree node representing a data type. Write @code{TYPE_MODE | |
3709 | (@var{valtype})} to get the machine mode used to represent that type. | |
3710 | On many machines, only the mode is relevant. (Actually, on most | |
3711 | machines, scalar values are returned in the same place regardless of | |
bd819a4a | 3712 | mode). |
feca2ed3 JW |
3713 | |
3714 | The value of the expression is usually a @code{reg} RTX for the hard | |
3715 | register where the return value is stored. The value can also be a | |
3716 | @code{parallel} RTX, if the return value is in multiple places. See | |
3717 | @code{FUNCTION_ARG} for an explanation of the @code{parallel} form. | |
3718 | ||
3719 | If @code{PROMOTE_FUNCTION_RETURN} is defined, you must apply the same | |
3720 | promotion rules specified in @code{PROMOTE_MODE} if @var{valtype} is a | |
3721 | scalar type. | |
3722 | ||
3723 | If the precise function being called is known, @var{func} is a tree | |
3724 | node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null | |
3725 | pointer. This makes it possible to use a different value-returning | |
3726 | convention for specific functions when all their calls are | |
bd819a4a | 3727 | known. |
feca2ed3 JW |
3728 | |
3729 | @code{FUNCTION_VALUE} is not used for return vales with aggregate data | |
3730 | types, because these are returned in another way. See | |
3731 | @code{STRUCT_VALUE_REGNUM} and related macros, below. | |
3732 | ||
3733 | @findex FUNCTION_OUTGOING_VALUE | |
3734 | @item FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func}) | |
3735 | Define this macro if the target machine has ``register windows'' | |
3736 | so that the register in which a function returns its value is not | |
3737 | the same as the one in which the caller sees the value. | |
3738 | ||
3739 | For such machines, @code{FUNCTION_VALUE} computes the register in which | |
3740 | the caller will see the value. @code{FUNCTION_OUTGOING_VALUE} should be | |
3741 | defined in a similar fashion to tell the function where to put the | |
bd819a4a | 3742 | value. |
feca2ed3 JW |
3743 | |
3744 | If @code{FUNCTION_OUTGOING_VALUE} is not defined, | |
bd819a4a | 3745 | @code{FUNCTION_VALUE} serves both purposes. |
feca2ed3 JW |
3746 | |
3747 | @code{FUNCTION_OUTGOING_VALUE} is not used for return vales with | |
3748 | aggregate data types, because these are returned in another way. See | |
3749 | @code{STRUCT_VALUE_REGNUM} and related macros, below. | |
3750 | ||
3751 | @findex LIBCALL_VALUE | |
3752 | @item LIBCALL_VALUE (@var{mode}) | |
3753 | A C expression to create an RTX representing the place where a library | |
3754 | function returns a value of mode @var{mode}. If the precise function | |
3755 | being called is known, @var{func} is a tree node | |
3756 | (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null | |
3757 | pointer. This makes it possible to use a different value-returning | |
3758 | convention for specific functions when all their calls are | |
bd819a4a | 3759 | known. |
feca2ed3 JW |
3760 | |
3761 | Note that ``library function'' in this context means a compiler | |
3762 | support routine, used to perform arithmetic, whose name is known | |
3763 | specially by the compiler and was not mentioned in the C code being | |
3764 | compiled. | |
3765 | ||
3766 | The definition of @code{LIBRARY_VALUE} need not be concerned aggregate | |
3767 | data types, because none of the library functions returns such types. | |
3768 | ||
3769 | @findex FUNCTION_VALUE_REGNO_P | |
3770 | @item FUNCTION_VALUE_REGNO_P (@var{regno}) | |
3771 | A C expression that is nonzero if @var{regno} is the number of a hard | |
3772 | register in which the values of called function may come back. | |
3773 | ||
3774 | A register whose use for returning values is limited to serving as the | |
3775 | second of a pair (for a value of type @code{double}, say) need not be | |
3776 | recognized by this macro. So for most machines, this definition | |
3777 | suffices: | |
3778 | ||
3779 | @example | |
3780 | #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0) | |
3781 | @end example | |
3782 | ||
3783 | If the machine has register windows, so that the caller and the called | |
3784 | function use different registers for the return value, this macro | |
3785 | should recognize only the caller's register numbers. | |
3786 | ||
3787 | @findex APPLY_RESULT_SIZE | |
3788 | @item APPLY_RESULT_SIZE | |
3789 | Define this macro if @samp{untyped_call} and @samp{untyped_return} | |
3790 | need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for | |
3791 | saving and restoring an arbitrary return value. | |
3792 | @end table | |
3793 | ||
3794 | @node Aggregate Return | |
3795 | @subsection How Large Values Are Returned | |
3796 | @cindex aggregates as return values | |
3797 | @cindex large return values | |
3798 | @cindex returning aggregate values | |
3799 | @cindex structure value address | |
3800 | ||
3801 | When a function value's mode is @code{BLKmode} (and in some other | |
3802 | cases), the value is not returned according to @code{FUNCTION_VALUE} | |
3803 | (@pxref{Scalar Return}). Instead, the caller passes the address of a | |
3804 | block of memory in which the value should be stored. This address | |
3805 | is called the @dfn{structure value address}. | |
3806 | ||
3807 | This section describes how to control returning structure values in | |
3808 | memory. | |
3809 | ||
3810 | @table @code | |
3811 | @findex RETURN_IN_MEMORY | |
3812 | @item RETURN_IN_MEMORY (@var{type}) | |
3813 | A C expression which can inhibit the returning of certain function | |
3814 | values in registers, based on the type of value. A nonzero value says | |
3815 | to return the function value in memory, just as large structures are | |
3816 | always returned. Here @var{type} will be a C expression of type | |
3817 | @code{tree}, representing the data type of the value. | |
3818 | ||
3819 | Note that values of mode @code{BLKmode} must be explicitly handled | |
630d3d5a | 3820 | by this macro. Also, the option @option{-fpcc-struct-return} |
feca2ed3 JW |
3821 | takes effect regardless of this macro. On most systems, it is |
3822 | possible to leave the macro undefined; this causes a default | |
3823 | definition to be used, whose value is the constant 1 for @code{BLKmode} | |
3824 | values, and 0 otherwise. | |
3825 | ||
3826 | Do not use this macro to indicate that structures and unions should always | |
3827 | be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN} | |
3828 | to indicate this. | |
3829 | ||
3830 | @findex DEFAULT_PCC_STRUCT_RETURN | |
3831 | @item DEFAULT_PCC_STRUCT_RETURN | |
3832 | Define this macro to be 1 if all structure and union return values must be | |
3833 | in memory. Since this results in slower code, this should be defined | |
161d7b59 | 3834 | only if needed for compatibility with other compilers or with an ABI@. |
feca2ed3 JW |
3835 | If you define this macro to be 0, then the conventions used for structure |
3836 | and union return values are decided by the @code{RETURN_IN_MEMORY} macro. | |
3837 | ||
3838 | If not defined, this defaults to the value 1. | |
3839 | ||
3840 | @findex STRUCT_VALUE_REGNUM | |
3841 | @item STRUCT_VALUE_REGNUM | |
3842 | If the structure value address is passed in a register, then | |
3843 | @code{STRUCT_VALUE_REGNUM} should be the number of that register. | |
3844 | ||
3845 | @findex STRUCT_VALUE | |
3846 | @item STRUCT_VALUE | |
3847 | If the structure value address is not passed in a register, define | |
3848 | @code{STRUCT_VALUE} as an expression returning an RTX for the place | |
3849 | where the address is passed. If it returns 0, the address is passed as | |
3850 | an ``invisible'' first argument. | |
3851 | ||
3852 | @findex STRUCT_VALUE_INCOMING_REGNUM | |
3853 | @item STRUCT_VALUE_INCOMING_REGNUM | |
3854 | On some architectures the place where the structure value address | |
3855 | is found by the called function is not the same place that the | |
3856 | caller put it. This can be due to register windows, or it could | |
3857 | be because the function prologue moves it to a different place. | |
3858 | ||
3859 | If the incoming location of the structure value address is in a | |
3860 | register, define this macro as the register number. | |
3861 | ||
3862 | @findex STRUCT_VALUE_INCOMING | |
3863 | @item STRUCT_VALUE_INCOMING | |
3864 | If the incoming location is not a register, then you should define | |
3865 | @code{STRUCT_VALUE_INCOMING} as an expression for an RTX for where the | |
3866 | called function should find the value. If it should find the value on | |
3867 | the stack, define this to create a @code{mem} which refers to the frame | |
3868 | pointer. A definition of 0 means that the address is passed as an | |
3869 | ``invisible'' first argument. | |
3870 | ||
3871 | @findex PCC_STATIC_STRUCT_RETURN | |
3872 | @item PCC_STATIC_STRUCT_RETURN | |
3873 | Define this macro if the usual system convention on the target machine | |
3874 | for returning structures and unions is for the called function to return | |
3875 | the address of a static variable containing the value. | |
3876 | ||
3877 | Do not define this if the usual system convention is for the caller to | |
3878 | pass an address to the subroutine. | |
3879 | ||
630d3d5a JM |
3880 | This macro has effect in @option{-fpcc-struct-return} mode, but it does |
3881 | nothing when you use @option{-freg-struct-return} mode. | |
feca2ed3 JW |
3882 | @end table |
3883 | ||
3884 | @node Caller Saves | |
3885 | @subsection Caller-Saves Register Allocation | |
3886 | ||
a3a15b4d | 3887 | If you enable it, GCC can save registers around function calls. This |
feca2ed3 JW |
3888 | makes it possible to use call-clobbered registers to hold variables that |
3889 | must live across calls. | |
3890 | ||
3891 | @table @code | |
3892 | @findex DEFAULT_CALLER_SAVES | |
3893 | @item DEFAULT_CALLER_SAVES | |
3894 | Define this macro if function calls on the target machine do not preserve | |
3895 | any registers; in other words, if @code{CALL_USED_REGISTERS} has 1 | |
630d3d5a | 3896 | for all registers. When defined, this macro enables @option{-fcaller-saves} |
81610a0d | 3897 | by default for all optimization levels. It has no effect for optimization |
630d3d5a | 3898 | levels 2 and higher, where @option{-fcaller-saves} is the default. |
feca2ed3 JW |
3899 | |
3900 | @findex CALLER_SAVE_PROFITABLE | |
3901 | @item CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls}) | |
3902 | A C expression to determine whether it is worthwhile to consider placing | |
3903 | a pseudo-register in a call-clobbered hard register and saving and | |
3904 | restoring it around each function call. The expression should be 1 when | |
3905 | this is worth doing, and 0 otherwise. | |
3906 | ||
3907 | If you don't define this macro, a default is used which is good on most | |
3908 | machines: @code{4 * @var{calls} < @var{refs}}. | |
8d5c8167 JL |
3909 | |
3910 | @findex HARD_REGNO_CALLER_SAVE_MODE | |
3911 | @item HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs}) | |
3912 | A C expression specifying which mode is required for saving @var{nregs} | |
3913 | of a pseudo-register in call-clobbered hard register @var{regno}. If | |
3914 | @var{regno} is unsuitable for caller save, @code{VOIDmode} should be | |
3915 | returned. For most machines this macro need not be defined since GCC | |
3916 | will select the smallest suitable mode. | |
feca2ed3 JW |
3917 | @end table |
3918 | ||
3919 | @node Function Entry | |
3920 | @subsection Function Entry and Exit | |
3921 | @cindex function entry and exit | |
3922 | @cindex prologue | |
3923 | @cindex epilogue | |
3924 | ||
3925 | This section describes the macros that output function entry | |
3926 | (@dfn{prologue}) and exit (@dfn{epilogue}) code. | |
3927 | ||
08c148a8 NB |
3928 | @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size}) |
3929 | If defined, a function that outputs the assembler code for entry to a | |
feca2ed3 JW |
3930 | function. The prologue is responsible for setting up the stack frame, |
3931 | initializing the frame pointer register, saving registers that must be | |
3932 | saved, and allocating @var{size} additional bytes of storage for the | |
3933 | local variables. @var{size} is an integer. @var{file} is a stdio | |
3934 | stream to which the assembler code should be output. | |
3935 | ||
3936 | The label for the beginning of the function need not be output by this | |
3937 | macro. That has already been done when the macro is run. | |
3938 | ||
3939 | @findex regs_ever_live | |
3940 | To determine which registers to save, the macro can refer to the array | |
3941 | @code{regs_ever_live}: element @var{r} is nonzero if hard register | |
3942 | @var{r} is used anywhere within the function. This implies the function | |
3943 | prologue should save register @var{r}, provided it is not one of the | |
08c148a8 | 3944 | call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use |
feca2ed3 JW |
3945 | @code{regs_ever_live}.) |
3946 | ||
3947 | On machines that have ``register windows'', the function entry code does | |
3948 | not save on the stack the registers that are in the windows, even if | |
3949 | they are supposed to be preserved by function calls; instead it takes | |
3950 | appropriate steps to ``push'' the register stack, if any non-call-used | |
3951 | registers are used in the function. | |
3952 | ||
3953 | @findex frame_pointer_needed | |
3954 | On machines where functions may or may not have frame-pointers, the | |
3955 | function entry code must vary accordingly; it must set up the frame | |
3956 | pointer if one is wanted, and not otherwise. To determine whether a | |
3957 | frame pointer is in wanted, the macro can refer to the variable | |
3958 | @code{frame_pointer_needed}. The variable's value will be 1 at run | |
3959 | time in a function that needs a frame pointer. @xref{Elimination}. | |
3960 | ||
3961 | The function entry code is responsible for allocating any stack space | |
3962 | required for the function. This stack space consists of the regions | |
3963 | listed below. In most cases, these regions are allocated in the | |
3964 | order listed, with the last listed region closest to the top of the | |
3965 | stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and | |
3966 | the highest address if it is not defined). You can use a different order | |
3967 | for a machine if doing so is more convenient or required for | |
3968 | compatibility reasons. Except in cases where required by standard | |
3969 | or by a debugger, there is no reason why the stack layout used by GCC | |
3970 | need agree with that used by other compilers for a machine. | |
08c148a8 NB |
3971 | @end deftypefn |
3972 | ||
17b53c33 NB |
3973 | @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file}) |
3974 | If defined, a function that outputs assembler code at the end of a | |
3975 | prologue. This should be used when the function prologue is being | |
3976 | emitted as RTL, and you have some extra assembler that needs to be | |
3977 | emitted. @xref{prologue instruction pattern}. | |
3978 | @end deftypefn | |
3979 | ||
3980 | @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file}) | |
3981 | If defined, a function that outputs assembler code at the start of an | |
3982 | epilogue. This should be used when the function epilogue is being | |
3983 | emitted as RTL, and you have some extra assembler that needs to be | |
3984 | emitted. @xref{epilogue instruction pattern}. | |
3985 | @end deftypefn | |
3986 | ||
08c148a8 NB |
3987 | @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size}) |
3988 | If defined, a function that outputs the assembler code for exit from a | |
3989 | function. The epilogue is responsible for restoring the saved | |
3990 | registers and stack pointer to their values when the function was | |
3991 | called, and returning control to the caller. This macro takes the | |
3992 | same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the | |
3993 | registers to restore are determined from @code{regs_ever_live} and | |
3994 | @code{CALL_USED_REGISTERS} in the same way. | |
3995 | ||
3996 | On some machines, there is a single instruction that does all the work | |
3997 | of returning from the function. On these machines, give that | |
3998 | instruction the name @samp{return} and do not define the macro | |
3999 | @code{TARGET_ASM_FUNCTION_EPILOGUE} at all. | |
4000 | ||
4001 | Do not define a pattern named @samp{return} if you want the | |
4002 | @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target | |
4003 | switches to control whether return instructions or epilogues are used, | |
4004 | define a @samp{return} pattern with a validity condition that tests the | |
4005 | target switches appropriately. If the @samp{return} pattern's validity | |
4006 | condition is false, epilogues will be used. | |
4007 | ||
4008 | On machines where functions may or may not have frame-pointers, the | |
4009 | function exit code must vary accordingly. Sometimes the code for these | |
4010 | two cases is completely different. To determine whether a frame pointer | |
4011 | is wanted, the macro can refer to the variable | |
4012 | @code{frame_pointer_needed}. The variable's value will be 1 when compiling | |
4013 | a function that needs a frame pointer. | |
4014 | ||
4015 | Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and | |
4016 | @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially. | |
4017 | The C variable @code{current_function_is_leaf} is nonzero for such a | |
4018 | function. @xref{Leaf Functions}. | |
4019 | ||
4020 | On some machines, some functions pop their arguments on exit while | |
4021 | others leave that for the caller to do. For example, the 68020 when | |
4022 | given @option{-mrtd} pops arguments in functions that take a fixed | |
4023 | number of arguments. | |
4024 | ||
4025 | @findex current_function_pops_args | |
4026 | Your definition of the macro @code{RETURN_POPS_ARGS} decides which | |
4027 | functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE} | |
4028 | needs to know what was decided. The variable that is called | |
4029 | @code{current_function_pops_args} is the number of bytes of its | |
4030 | arguments that a function should pop. @xref{Scalar Return}. | |
4031 | @c what is the "its arguments" in the above sentence referring to, pray | |
4032 | @c tell? --mew 5feb93 | |
4033 | @end deftypefn | |
4034 | ||
4035 | @table @code | |
feca2ed3 JW |
4036 | |
4037 | @itemize @bullet | |
4038 | @item | |
4039 | @findex current_function_pretend_args_size | |
4040 | A region of @code{current_function_pretend_args_size} bytes of | |
4041 | uninitialized space just underneath the first argument arriving on the | |
4042 | stack. (This may not be at the very start of the allocated stack region | |
4043 | if the calling sequence has pushed anything else since pushing the stack | |
4044 | arguments. But usually, on such machines, nothing else has been pushed | |
4045 | yet, because the function prologue itself does all the pushing.) This | |
4046 | region is used on machines where an argument may be passed partly in | |
4047 | registers and partly in memory, and, in some cases to support the | |
6c535c69 | 4048 | features in @code{<stdarg.h>}. |
feca2ed3 JW |
4049 | |
4050 | @item | |
4051 | An area of memory used to save certain registers used by the function. | |
4052 | The size of this area, which may also include space for such things as | |
4053 | the return address and pointers to previous stack frames, is | |
4054 | machine-specific and usually depends on which registers have been used | |
4055 | in the function. Machines with register windows often do not require | |
4056 | a save area. | |
4057 | ||
4058 | @item | |
4059 | A region of at least @var{size} bytes, possibly rounded up to an allocation | |
4060 | boundary, to contain the local variables of the function. On some machines, | |
4061 | this region and the save area may occur in the opposite order, with the | |
4062 | save area closer to the top of the stack. | |
4063 | ||
4064 | @item | |
4065 | @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames | |
4066 | Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of | |
4067 | @code{current_function_outgoing_args_size} bytes to be used for outgoing | |
4068 | argument lists of the function. @xref{Stack Arguments}. | |
4069 | @end itemize | |
4070 | ||
08c148a8 NB |
4071 | Normally, it is necessary for the macros |
4072 | @code{TARGET_ASM_FUNCTION_PROLOGUE} and | |
4073 | @code{TARGET_ASM_FUNCTION_EPILOGUE} to treat leaf functions specially. | |
4074 | The C variable @code{current_function_is_leaf} is nonzero for such a | |
4075 | function. | |
feca2ed3 JW |
4076 | |
4077 | @findex EXIT_IGNORE_STACK | |
4078 | @item EXIT_IGNORE_STACK | |
4079 | Define this macro as a C expression that is nonzero if the return | |
4080 | instruction or the function epilogue ignores the value of the stack | |
4081 | pointer; in other words, if it is safe to delete an instruction to | |
4082 | adjust the stack pointer before a return from the function. | |
4083 | ||
4084 | Note that this macro's value is relevant only for functions for which | |
4085 | frame pointers are maintained. It is never safe to delete a final | |
4086 | stack adjustment in a function that has no frame pointer, and the | |
4087 | compiler knows this regardless of @code{EXIT_IGNORE_STACK}. | |
4088 | ||
4089 | @findex EPILOGUE_USES | |
4090 | @item EPILOGUE_USES (@var{regno}) | |
8760eaae | 4091 | Define this macro as a C expression that is nonzero for registers that are |
feca2ed3 JW |
4092 | used by the epilogue or the @samp{return} pattern. The stack and frame |
4093 | pointer registers are already be assumed to be used as needed. | |
4094 | ||
15b5aef3 RH |
4095 | @findex EH_USES |
4096 | @item EH_USES (@var{regno}) | |
4097 | Define this macro as a C expression that is nonzero for registers that are | |
4098 | used by the exception handling mechanism, and so should be considered live | |
4099 | on entry to an exception edge. | |
4100 | ||
feca2ed3 JW |
4101 | @findex DELAY_SLOTS_FOR_EPILOGUE |
4102 | @item DELAY_SLOTS_FOR_EPILOGUE | |
4103 | Define this macro if the function epilogue contains delay slots to which | |
4104 | instructions from the rest of the function can be ``moved''. The | |
4105 | definition should be a C expression whose value is an integer | |
4106 | representing the number of delay slots there. | |
4107 | ||
4108 | @findex ELIGIBLE_FOR_EPILOGUE_DELAY | |
4109 | @item ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n}) | |
4110 | A C expression that returns 1 if @var{insn} can be placed in delay | |
4111 | slot number @var{n} of the epilogue. | |
4112 | ||
4113 | The argument @var{n} is an integer which identifies the delay slot now | |
4114 | being considered (since different slots may have different rules of | |
4115 | eligibility). It is never negative and is always less than the number | |
4116 | of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns). | |
4117 | If you reject a particular insn for a given delay slot, in principle, it | |
4118 | may be reconsidered for a subsequent delay slot. Also, other insns may | |
4119 | (at least in principle) be considered for the so far unfilled delay | |
4120 | slot. | |
4121 | ||
4122 | @findex current_function_epilogue_delay_list | |
4123 | @findex final_scan_insn | |
4124 | The insns accepted to fill the epilogue delay slots are put in an RTL | |
4125 | list made with @code{insn_list} objects, stored in the variable | |
4126 | @code{current_function_epilogue_delay_list}. The insn for the first | |
4127 | delay slot comes first in the list. Your definition of the macro | |
08c148a8 NB |
4128 | @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by |
4129 | outputting the insns in this list, usually by calling | |
4130 | @code{final_scan_insn}. | |
feca2ed3 JW |
4131 | |
4132 | You need not define this macro if you did not define | |
4133 | @code{DELAY_SLOTS_FOR_EPILOGUE}. | |
4134 | ||
4135 | @findex ASM_OUTPUT_MI_THUNK | |
4136 | @item ASM_OUTPUT_MI_THUNK (@var{file}, @var{thunk_fndecl}, @var{delta}, @var{function}) | |
4137 | A C compound statement that outputs the assembler code for a thunk | |
4138 | function, used to implement C++ virtual function calls with multiple | |
4139 | inheritance. The thunk acts as a wrapper around a virtual function, | |
4140 | adjusting the implicit object parameter before handing control off to | |
4141 | the real function. | |
4142 | ||
4143 | First, emit code to add the integer @var{delta} to the location that | |
4144 | contains the incoming first argument. Assume that this argument | |
4145 | contains a pointer, and is the one used to pass the @code{this} pointer | |
4146 | in C++. This is the incoming argument @emph{before} the function prologue, | |
e979f9e8 | 4147 | e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of |
feca2ed3 JW |
4148 | all other incoming arguments. |
4149 | ||
4150 | After the addition, emit code to jump to @var{function}, which is a | |
4151 | @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does | |
4152 | not touch the return address. Hence returning from @var{FUNCTION} will | |
4153 | return to whoever called the current @samp{thunk}. | |
4154 | ||
4155 | The effect must be as if @var{function} had been called directly with | |
4156 | the adjusted first argument. This macro is responsible for emitting all | |
08c148a8 NB |
4157 | of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE} |
4158 | and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked. | |
feca2ed3 JW |
4159 | |
4160 | The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function} | |
4161 | have already been extracted from it.) It might possibly be useful on | |
4162 | some targets, but probably not. | |
4163 | ||
861bb6c1 | 4164 | If you do not define this macro, the target-independent code in the C++ |
c771326b | 4165 | front end will generate a less efficient heavyweight thunk that calls |
861bb6c1 JL |
4166 | @var{function} instead of jumping to it. The generic approach does |
4167 | not support varargs. | |
feca2ed3 JW |
4168 | @end table |
4169 | ||
4170 | @node Profiling | |
4171 | @subsection Generating Code for Profiling | |
4172 | @cindex profiling, code generation | |
4173 | ||
4174 | These macros will help you generate code for profiling. | |
4175 | ||
4176 | @table @code | |
4177 | @findex FUNCTION_PROFILER | |
4178 | @item FUNCTION_PROFILER (@var{file}, @var{labelno}) | |
4179 | A C statement or compound statement to output to @var{file} some | |
4180 | assembler code to call the profiling subroutine @code{mcount}. | |
feca2ed3 JW |
4181 | |
4182 | @findex mcount | |
980e2067 | 4183 | The details of how @code{mcount} expects to be called are determined by |
161d7b59 | 4184 | your operating system environment, not by GCC@. To figure them out, |
980e2067 JL |
4185 | compile a small program for profiling using the system's installed C |
4186 | compiler and look at the assembler code that results. | |
4187 | ||
4188 | Older implementations of @code{mcount} expect the address of a counter | |
4189 | variable to be loaded into some register. The name of this variable is | |
4190 | @samp{LP} followed by the number @var{labelno}, so you would generate | |
4191 | the name using @samp{LP%d} in a @code{fprintf}. | |
4192 | ||
411707f4 CC |
4193 | @findex PROFILE_HOOK |
4194 | @item PROFILE_HOOK | |
4195 | A C statement or compound statement to output to @var{file} some assembly | |
4196 | code to call the profiling subroutine @code{mcount} even the target does | |
4197 | not support profiling. | |
4198 | ||
980e2067 JL |
4199 | @findex NO_PROFILE_COUNTERS |
4200 | @item NO_PROFILE_COUNTERS | |
4201 | Define this macro if the @code{mcount} subroutine on your system does | |
4202 | not need a counter variable allocated for each function. This is true | |
4203 | for almost all modern implementations. If you define this macro, you | |
4204 | must not use the @var{labelno} argument to @code{FUNCTION_PROFILER}. | |
feca2ed3 JW |
4205 | |
4206 | @findex PROFILE_BEFORE_PROLOGUE | |
4207 | @item PROFILE_BEFORE_PROLOGUE | |
4208 | Define this macro if the code for function profiling should come before | |
4209 | the function prologue. Normally, the profiling code comes after. | |
feca2ed3 JW |
4210 | @end table |
4211 | ||
91d231cb JM |
4212 | @node Tail Calls |
4213 | @subsection Permitting tail calls | |
4214 | @cindex tail calls | |
b36f4ed3 | 4215 | |
4cb1433c RH |
4216 | @table @code |
4217 | @findex FUNCTION_OK_FOR_SIBCALL | |
4218 | @item FUNCTION_OK_FOR_SIBCALL (@var{decl}) | |
4219 | A C expression that evaluates to true if it is ok to perform a sibling | |
c237e94a | 4220 | call to @var{decl} from the current function. |
4cb1433c RH |
4221 | |
4222 | It is not uncommon for limitations of calling conventions to prevent | |
4223 | tail calls to functions outside the current unit of translation, or | |
4224 | during PIC compilation. Use this macro to enforce these restrictions, | |
02f52e19 | 4225 | as the @code{sibcall} md pattern can not fail, or fall over to a |
4cb1433c RH |
4226 | ``normal'' call. |
4227 | @end table | |
4228 | ||
feca2ed3 JW |
4229 | @node Varargs |
4230 | @section Implementing the Varargs Macros | |
4231 | @cindex varargs implementation | |
4232 | ||
aee96fe9 JM |
4233 | GCC comes with an implementation of @code{<varargs.h>} and |
4234 | @code{<stdarg.h>} that work without change on machines that pass arguments | |
feca2ed3 JW |
4235 | on the stack. Other machines require their own implementations of |
4236 | varargs, and the two machine independent header files must have | |
4237 | conditionals to include it. | |
4238 | ||
aee96fe9 | 4239 | ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in |
feca2ed3 JW |
4240 | the calling convention for @code{va_start}. The traditional |
4241 | implementation takes just one argument, which is the variable in which | |
5490d604 | 4242 | to store the argument pointer. The ISO implementation of |
feca2ed3 JW |
4243 | @code{va_start} takes an additional second argument. The user is |
4244 | supposed to write the last named argument of the function here. | |
4245 | ||
4246 | However, @code{va_start} should not use this argument. The way to find | |
4247 | the end of the named arguments is with the built-in functions described | |
4248 | below. | |
4249 | ||
4250 | @table @code | |
4251 | @findex __builtin_saveregs | |
4252 | @item __builtin_saveregs () | |
4253 | Use this built-in function to save the argument registers in memory so | |
5490d604 | 4254 | that the varargs mechanism can access them. Both ISO and traditional |
feca2ed3 JW |
4255 | versions of @code{va_start} must use @code{__builtin_saveregs}, unless |
4256 | you use @code{SETUP_INCOMING_VARARGS} (see below) instead. | |
4257 | ||
4258 | On some machines, @code{__builtin_saveregs} is open-coded under the | |
4259 | control of the macro @code{EXPAND_BUILTIN_SAVEREGS}. On other machines, | |
4260 | it calls a routine written in assembler language, found in | |
4261 | @file{libgcc2.c}. | |
4262 | ||
4263 | Code generated for the call to @code{__builtin_saveregs} appears at the | |
4264 | beginning of the function, as opposed to where the call to | |
4265 | @code{__builtin_saveregs} is written, regardless of what the code is. | |
4266 | This is because the registers must be saved before the function starts | |
4267 | to use them for its own purposes. | |
4268 | @c i rewrote the first sentence above to fix an overfull hbox. --mew | |
4269 | @c 10feb93 | |
4270 | ||
4271 | @findex __builtin_args_info | |
4272 | @item __builtin_args_info (@var{category}) | |
4273 | Use this built-in function to find the first anonymous arguments in | |
4274 | registers. | |
4275 | ||
4276 | In general, a machine may have several categories of registers used for | |
4277 | arguments, each for a particular category of data types. (For example, | |
4278 | on some machines, floating-point registers are used for floating-point | |
4279 | arguments while other arguments are passed in the general registers.) | |
4280 | To make non-varargs functions use the proper calling convention, you | |
4281 | have defined the @code{CUMULATIVE_ARGS} data type to record how many | |
4282 | registers in each category have been used so far | |
4283 | ||
4284 | @code{__builtin_args_info} accesses the same data structure of type | |
4285 | @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished | |
4286 | with it, with @var{category} specifying which word to access. Thus, the | |
4287 | value indicates the first unused register in a given category. | |
4288 | ||
4289 | Normally, you would use @code{__builtin_args_info} in the implementation | |
4290 | of @code{va_start}, accessing each category just once and storing the | |
4291 | value in the @code{va_list} object. This is because @code{va_list} will | |
4292 | have to update the values, and there is no way to alter the | |
4293 | values accessed by @code{__builtin_args_info}. | |
4294 | ||
4295 | @findex __builtin_next_arg | |
4296 | @item __builtin_next_arg (@var{lastarg}) | |
4297 | This is the equivalent of @code{__builtin_args_info}, for stack | |
4298 | arguments. It returns the address of the first anonymous stack | |
767094dd | 4299 | argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it |
feca2ed3 JW |
4300 | returns the address of the location above the first anonymous stack |
4301 | argument. Use it in @code{va_start} to initialize the pointer for | |
4302 | fetching arguments from the stack. Also use it in @code{va_start} to | |
4303 | verify that the second parameter @var{lastarg} is the last named argument | |
4304 | of the current function. | |
4305 | ||
4306 | @findex __builtin_classify_type | |
4307 | @item __builtin_classify_type (@var{object}) | |
4308 | Since each machine has its own conventions for which data types are | |
4309 | passed in which kind of register, your implementation of @code{va_arg} | |
4310 | has to embody these conventions. The easiest way to categorize the | |
4311 | specified data type is to use @code{__builtin_classify_type} together | |
4312 | with @code{sizeof} and @code{__alignof__}. | |
4313 | ||
4314 | @code{__builtin_classify_type} ignores the value of @var{object}, | |
4315 | considering only its data type. It returns an integer describing what | |
4316 | kind of type that is---integer, floating, pointer, structure, and so on. | |
4317 | ||
4318 | The file @file{typeclass.h} defines an enumeration that you can use to | |
4319 | interpret the values of @code{__builtin_classify_type}. | |
4320 | @end table | |
4321 | ||
4322 | These machine description macros help implement varargs: | |
4323 | ||
4324 | @table @code | |
4325 | @findex EXPAND_BUILTIN_SAVEREGS | |
d3707adb | 4326 | @item EXPAND_BUILTIN_SAVEREGS () |
feca2ed3 JW |
4327 | If defined, is a C expression that produces the machine-specific code |
4328 | for a call to @code{__builtin_saveregs}. This code will be moved to the | |
4329 | very beginning of the function, before any parameter access are made. | |
4330 | The return value of this function should be an RTX that contains the | |
4331 | value to use as the return of @code{__builtin_saveregs}. | |
4332 | ||
feca2ed3 | 4333 | @findex SETUP_INCOMING_VARARGS |
59d40964 | 4334 | @item SETUP_INCOMING_VARARGS (@var{args_so_far}, @var{mode}, @var{type}, @var{pretend_args_size}, @var{second_time}) |
feca2ed3 JW |
4335 | This macro offers an alternative to using @code{__builtin_saveregs} and |
4336 | defining the macro @code{EXPAND_BUILTIN_SAVEREGS}. Use it to store the | |
4337 | anonymous register arguments into the stack so that all the arguments | |
4338 | appear to have been passed consecutively on the stack. Once this is | |
4339 | done, you can use the standard implementation of varargs that works for | |
4340 | machines that pass all their arguments on the stack. | |
4341 | ||
4342 | The argument @var{args_so_far} is the @code{CUMULATIVE_ARGS} data | |
8760eaae | 4343 | structure, containing the values that are obtained after processing the |
feca2ed3 JW |
4344 | named arguments. The arguments @var{mode} and @var{type} describe the |
4345 | last named argument---its machine mode and its data type as a tree node. | |
4346 | ||
4347 | The macro implementation should do two things: first, push onto the | |
4348 | stack all the argument registers @emph{not} used for the named | |
4349 | arguments, and second, store the size of the data thus pushed into the | |
4350 | @code{int}-valued variable whose name is supplied as the argument | |
4351 | @var{pretend_args_size}. The value that you store here will serve as | |
4352 | additional offset for setting up the stack frame. | |
4353 | ||
4354 | Because you must generate code to push the anonymous arguments at | |
4355 | compile time without knowing their data types, | |
4356 | @code{SETUP_INCOMING_VARARGS} is only useful on machines that have just | |
4357 | a single category of argument register and use it uniformly for all data | |
4358 | types. | |
4359 | ||
4360 | If the argument @var{second_time} is nonzero, it means that the | |
4361 | arguments of the function are being analyzed for the second time. This | |
4362 | happens for an inline function, which is not actually compiled until the | |
4363 | end of the source file. The macro @code{SETUP_INCOMING_VARARGS} should | |
4364 | not generate any instructions in this case. | |
4365 | ||
4366 | @findex STRICT_ARGUMENT_NAMING | |
4367 | @item STRICT_ARGUMENT_NAMING | |
e5e809f4 JL |
4368 | Define this macro to be a nonzero value if the location where a function |
4369 | argument is passed depends on whether or not it is a named argument. | |
feca2ed3 JW |
4370 | |
4371 | This macro controls how the @var{named} argument to @code{FUNCTION_ARG} | |
e5e809f4 JL |
4372 | is set for varargs and stdarg functions. If this macro returns a |
4373 | nonzero value, the @var{named} argument is always true for named | |
4374 | arguments, and false for unnamed arguments. If it returns a value of | |
4375 | zero, but @code{SETUP_INCOMING_VARARGS} is defined, then all arguments | |
4376 | are treated as named. Otherwise, all named arguments except the last | |
4377 | are treated as named. | |
4378 | ||
4379 | You need not define this macro if it always returns zero. | |
9ab70a9b R |
4380 | |
4381 | @findex PRETEND_OUTGOING_VARARGS_NAMED | |
4382 | @item PRETEND_OUTGOING_VARARGS_NAMED | |
4383 | If you need to conditionally change ABIs so that one works with | |
4384 | @code{SETUP_INCOMING_VARARGS}, but the other works like neither | |
4385 | @code{SETUP_INCOMING_VARARGS} nor @code{STRICT_ARGUMENT_NAMING} was | |
4386 | defined, then define this macro to return nonzero if | |
4387 | @code{SETUP_INCOMING_VARARGS} is used, zero otherwise. | |
4388 | Otherwise, you should not define this macro. | |
feca2ed3 JW |
4389 | @end table |
4390 | ||
4391 | @node Trampolines | |
4392 | @section Trampolines for Nested Functions | |
4393 | @cindex trampolines for nested functions | |
4394 | @cindex nested functions, trampolines for | |
4395 | ||
4396 | A @dfn{trampoline} is a small piece of code that is created at run time | |
4397 | when the address of a nested function is taken. It normally resides on | |
4398 | the stack, in the stack frame of the containing function. These macros | |
a3a15b4d | 4399 | tell GCC how to generate code to allocate and initialize a |
feca2ed3 JW |
4400 | trampoline. |
4401 | ||
4402 | The instructions in the trampoline must do two things: load a constant | |
4403 | address into the static chain register, and jump to the real address of | |
4404 | the nested function. On CISC machines such as the m68k, this requires | |
4405 | two instructions, a move immediate and a jump. Then the two addresses | |
4406 | exist in the trampoline as word-long immediate operands. On RISC | |
4407 | machines, it is often necessary to load each address into a register in | |
4408 | two parts. Then pieces of each address form separate immediate | |
4409 | operands. | |
4410 | ||
4411 | The code generated to initialize the trampoline must store the variable | |
4412 | parts---the static chain value and the function address---into the | |
4413 | immediate operands of the instructions. On a CISC machine, this is | |
4414 | simply a matter of copying each address to a memory reference at the | |
4415 | proper offset from the start of the trampoline. On a RISC machine, it | |
4416 | may be necessary to take out pieces of the address and store them | |
4417 | separately. | |
4418 | ||
4419 | @table @code | |
4420 | @findex TRAMPOLINE_TEMPLATE | |
4421 | @item TRAMPOLINE_TEMPLATE (@var{file}) | |
4422 | A C statement to output, on the stream @var{file}, assembler code for a | |
4423 | block of data that contains the constant parts of a trampoline. This | |
4424 | code should not include a label---the label is taken care of | |
4425 | automatically. | |
4426 | ||
4427 | If you do not define this macro, it means no template is needed | |
4428 | for the target. Do not define this macro on systems where the block move | |
4429 | code to copy the trampoline into place would be larger than the code | |
4430 | to generate it on the spot. | |
4431 | ||
4432 | @findex TRAMPOLINE_SECTION | |
4433 | @item TRAMPOLINE_SECTION | |
4434 | The name of a subroutine to switch to the section in which the | |
4435 | trampoline template is to be placed (@pxref{Sections}). The default is | |
4436 | a value of @samp{readonly_data_section}, which places the trampoline in | |
4437 | the section containing read-only data. | |
4438 | ||
4439 | @findex TRAMPOLINE_SIZE | |
4440 | @item TRAMPOLINE_SIZE | |
4441 | A C expression for the size in bytes of the trampoline, as an integer. | |
4442 | ||
4443 | @findex TRAMPOLINE_ALIGNMENT | |
4444 | @item TRAMPOLINE_ALIGNMENT | |
4445 | Alignment required for trampolines, in bits. | |
4446 | ||
4447 | If you don't define this macro, the value of @code{BIGGEST_ALIGNMENT} | |
4448 | is used for aligning trampolines. | |
4449 | ||
4450 | @findex INITIALIZE_TRAMPOLINE | |
4451 | @item INITIALIZE_TRAMPOLINE (@var{addr}, @var{fnaddr}, @var{static_chain}) | |
4452 | A C statement to initialize the variable parts of a trampoline. | |
4453 | @var{addr} is an RTX for the address of the trampoline; @var{fnaddr} is | |
4454 | an RTX for the address of the nested function; @var{static_chain} is an | |
4455 | RTX for the static chain value that should be passed to the function | |
4456 | when it is called. | |
4457 | ||
b33493e3 AO |
4458 | @findex TRAMPOLINE_ADJUST_ADDRESS |
4459 | @item TRAMPOLINE_ADJUST_ADDRESS (@var{addr}) | |
4460 | A C statement that should perform any machine-specific adjustment in | |
4461 | the address of the trampoline. Its argument contains the address that | |
4462 | was passed to @code{INITIALIZE_TRAMPOLINE}. In case the address to be | |
4463 | used for a function call should be different from the address in which | |
4464 | the template was stored, the different address should be assigned to | |
4465 | @var{addr}. If this macro is not defined, @var{addr} will be used for | |
4466 | function calls. | |
4467 | ||
feca2ed3 JW |
4468 | @findex ALLOCATE_TRAMPOLINE |
4469 | @item ALLOCATE_TRAMPOLINE (@var{fp}) | |
4470 | A C expression to allocate run-time space for a trampoline. The | |
4471 | expression value should be an RTX representing a memory reference to the | |
4472 | space for the trampoline. | |
4473 | ||
08c148a8 NB |
4474 | @cindex @code{TARGET_ASM_FUNCTION_EPILOGUE} and trampolines |
4475 | @cindex @code{TARGET_ASM_FUNCTION_PROLOGUE} and trampolines | |
feca2ed3 JW |
4476 | If this macro is not defined, by default the trampoline is allocated as |
4477 | a stack slot. This default is right for most machines. The exceptions | |
4478 | are machines where it is impossible to execute instructions in the stack | |
4479 | area. On such machines, you may have to implement a separate stack, | |
08c148a8 NB |
4480 | using this macro in conjunction with @code{TARGET_ASM_FUNCTION_PROLOGUE} |
4481 | and @code{TARGET_ASM_FUNCTION_EPILOGUE}. | |
feca2ed3 JW |
4482 | |
4483 | @var{fp} points to a data structure, a @code{struct function}, which | |
4484 | describes the compilation status of the immediate containing function of | |
4485 | the function which the trampoline is for. Normally (when | |
4486 | @code{ALLOCATE_TRAMPOLINE} is not defined), the stack slot for the | |
4487 | trampoline is in the stack frame of this containing function. Other | |
4488 | allocation strategies probably must do something analogous with this | |
4489 | information. | |
4490 | @end table | |
4491 | ||
4492 | Implementing trampolines is difficult on many machines because they have | |
4493 | separate instruction and data caches. Writing into a stack location | |
4494 | fails to clear the memory in the instruction cache, so when the program | |
4495 | jumps to that location, it executes the old contents. | |
4496 | ||
4497 | Here are two possible solutions. One is to clear the relevant parts of | |
4498 | the instruction cache whenever a trampoline is set up. The other is to | |
4499 | make all trampolines identical, by having them jump to a standard | |
4500 | subroutine. The former technique makes trampoline execution faster; the | |
4501 | latter makes initialization faster. | |
4502 | ||
4503 | To clear the instruction cache when a trampoline is initialized, define | |
4504 | the following macros which describe the shape of the cache. | |
4505 | ||
4506 | @table @code | |
4507 | @findex INSN_CACHE_SIZE | |
4508 | @item INSN_CACHE_SIZE | |
4509 | The total size in bytes of the cache. | |
4510 | ||
4511 | @findex INSN_CACHE_LINE_WIDTH | |
4512 | @item INSN_CACHE_LINE_WIDTH | |
4513 | The length in bytes of each cache line. The cache is divided into cache | |
4514 | lines which are disjoint slots, each holding a contiguous chunk of data | |
4515 | fetched from memory. Each time data is brought into the cache, an | |
4516 | entire line is read at once. The data loaded into a cache line is | |
4517 | always aligned on a boundary equal to the line size. | |
4518 | ||
4519 | @findex INSN_CACHE_DEPTH | |
4520 | @item INSN_CACHE_DEPTH | |
4521 | The number of alternative cache lines that can hold any particular memory | |
4522 | location. | |
4523 | @end table | |
4524 | ||
4525 | Alternatively, if the machine has system calls or instructions to clear | |
4526 | the instruction cache directly, you can define the following macro. | |
4527 | ||
4528 | @table @code | |
4529 | @findex CLEAR_INSN_CACHE | |
aee96fe9 | 4530 | @item CLEAR_INSN_CACHE (@var{beg}, @var{end}) |
feca2ed3 JW |
4531 | If defined, expands to a C expression clearing the @emph{instruction |
4532 | cache} in the specified interval. If it is not defined, and the macro | |
aee96fe9 | 4533 | @code{INSN_CACHE_SIZE} is defined, some generic code is generated to clear the |
feca2ed3 | 4534 | cache. The definition of this macro would typically be a series of |
aee96fe9 | 4535 | @code{asm} statements. Both @var{beg} and @var{end} are both pointer |
feca2ed3 JW |
4536 | expressions. |
4537 | @end table | |
4538 | ||
4539 | To use a standard subroutine, define the following macro. In addition, | |
4540 | you must make sure that the instructions in a trampoline fill an entire | |
4541 | cache line with identical instructions, or else ensure that the | |
4542 | beginning of the trampoline code is always aligned at the same point in | |
4543 | its cache line. Look in @file{m68k.h} as a guide. | |
4544 | ||
4545 | @table @code | |
4546 | @findex TRANSFER_FROM_TRAMPOLINE | |
4547 | @item TRANSFER_FROM_TRAMPOLINE | |
4548 | Define this macro if trampolines need a special subroutine to do their | |
4549 | work. The macro should expand to a series of @code{asm} statements | |
161d7b59 | 4550 | which will be compiled with GCC@. They go in a library function named |
feca2ed3 JW |
4551 | @code{__transfer_from_trampoline}. |
4552 | ||
4553 | If you need to avoid executing the ordinary prologue code of a compiled | |
4554 | C function when you jump to the subroutine, you can do so by placing a | |
4555 | special label of your own in the assembler code. Use one @code{asm} | |
4556 | statement to generate an assembler label, and another to make the label | |
4557 | global. Then trampolines can use that label to jump directly to your | |
4558 | special assembler code. | |
4559 | @end table | |
4560 | ||
4561 | @node Library Calls | |
4562 | @section Implicit Calls to Library Routines | |
4563 | @cindex library subroutine names | |
4564 | @cindex @file{libgcc.a} | |
4565 | ||
4566 | @c prevent bad page break with this line | |
4567 | Here is an explanation of implicit calls to library routines. | |
4568 | ||
4569 | @table @code | |
4570 | @findex MULSI3_LIBCALL | |
4571 | @item MULSI3_LIBCALL | |
4572 | A C string constant giving the name of the function to call for | |
4573 | multiplication of one signed full-word by another. If you do not | |
4574 | define this macro, the default name is used, which is @code{__mulsi3}, | |
4575 | a function defined in @file{libgcc.a}. | |
4576 | ||
4577 | @findex DIVSI3_LIBCALL | |
4578 | @item DIVSI3_LIBCALL | |
4579 | A C string constant giving the name of the function to call for | |
4580 | division of one signed full-word by another. If you do not define | |
4581 | this macro, the default name is used, which is @code{__divsi3}, a | |
4582 | function defined in @file{libgcc.a}. | |
4583 | ||
4584 | @findex UDIVSI3_LIBCALL | |
4585 | @item UDIVSI3_LIBCALL | |
4586 | A C string constant giving the name of the function to call for | |
4587 | division of one unsigned full-word by another. If you do not define | |
4588 | this macro, the default name is used, which is @code{__udivsi3}, a | |
4589 | function defined in @file{libgcc.a}. | |
4590 | ||
4591 | @findex MODSI3_LIBCALL | |
4592 | @item MODSI3_LIBCALL | |
4593 | A C string constant giving the name of the function to call for the | |
4594 | remainder in division of one signed full-word by another. If you do | |
4595 | not define this macro, the default name is used, which is | |
4596 | @code{__modsi3}, a function defined in @file{libgcc.a}. | |
4597 | ||
4598 | @findex UMODSI3_LIBCALL | |
4599 | @item UMODSI3_LIBCALL | |
4600 | A C string constant giving the name of the function to call for the | |
4601 | remainder in division of one unsigned full-word by another. If you do | |
4602 | not define this macro, the default name is used, which is | |
4603 | @code{__umodsi3}, a function defined in @file{libgcc.a}. | |
4604 | ||
4605 | @findex MULDI3_LIBCALL | |
4606 | @item MULDI3_LIBCALL | |
4607 | A C string constant giving the name of the function to call for | |
4608 | multiplication of one signed double-word by another. If you do not | |
4609 | define this macro, the default name is used, which is @code{__muldi3}, | |
4610 | a function defined in @file{libgcc.a}. | |
4611 | ||
4612 | @findex DIVDI3_LIBCALL | |
4613 | @item DIVDI3_LIBCALL | |
4614 | A C string constant giving the name of the function to call for | |
4615 | division of one signed double-word by another. If you do not define | |
4616 | this macro, the default name is used, which is @code{__divdi3}, a | |
4617 | function defined in @file{libgcc.a}. | |
4618 | ||
4619 | @findex UDIVDI3_LIBCALL | |
4620 | @item UDIVDI3_LIBCALL | |
4621 | A C string constant giving the name of the function to call for | |
4622 | division of one unsigned full-word by another. If you do not define | |
4623 | this macro, the default name is used, which is @code{__udivdi3}, a | |
4624 | function defined in @file{libgcc.a}. | |
4625 | ||
4626 | @findex MODDI3_LIBCALL | |
4627 | @item MODDI3_LIBCALL | |
4628 | A C string constant giving the name of the function to call for the | |
4629 | remainder in division of one signed double-word by another. If you do | |
4630 | not define this macro, the default name is used, which is | |
4631 | @code{__moddi3}, a function defined in @file{libgcc.a}. | |
4632 | ||
4633 | @findex UMODDI3_LIBCALL | |
4634 | @item UMODDI3_LIBCALL | |
4635 | A C string constant giving the name of the function to call for the | |
4636 | remainder in division of one unsigned full-word by another. If you do | |
4637 | not define this macro, the default name is used, which is | |
4638 | @code{__umoddi3}, a function defined in @file{libgcc.a}. | |
4639 | ||
d8088c6f BS |
4640 | @findex DECLARE_LIBRARY_RENAMES |
4641 | @item DECLARE_LIBRARY_RENAMES | |
4642 | This macro, if defined, should expand to a piece of C code that will get | |
4643 | expanded when compiling functions for libgcc.a. It can be used to | |
4644 | provide alternate names for gcc's internal library functions if there | |
4645 | are ABI-mandated names that the compiler should provide. | |
4646 | ||
feca2ed3 JW |
4647 | @findex INIT_TARGET_OPTABS |
4648 | @item INIT_TARGET_OPTABS | |
4649 | Define this macro as a C statement that declares additional library | |
767094dd | 4650 | routines renames existing ones. @code{init_optabs} calls this macro after |
feca2ed3 JW |
4651 | initializing all the normal library routines. |
4652 | ||
c5c60e15 BS |
4653 | @findex FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison}) |
4654 | @item FLOAT_LIB_COMPARE_RETURNS_BOOL | |
4655 | Define this macro as a C statement that returns nonzero if a call to | |
4656 | the floating point comparison library function will return a boolean | |
4657 | value that indicates the result of the comparison. It should return | |
4658 | zero if one of gcc's own libgcc functions is called. | |
4659 | ||
4660 | Most ports don't need to define this macro. | |
4661 | ||
feca2ed3 JW |
4662 | @findex TARGET_EDOM |
4663 | @cindex @code{EDOM}, implicit usage | |
4664 | @item TARGET_EDOM | |
4665 | The value of @code{EDOM} on the target machine, as a C integer constant | |
a3a15b4d | 4666 | expression. If you don't define this macro, GCC does not attempt to |
feca2ed3 JW |
4667 | deposit the value of @code{EDOM} into @code{errno} directly. Look in |
4668 | @file{/usr/include/errno.h} to find the value of @code{EDOM} on your | |
4669 | system. | |
4670 | ||
4671 | If you do not define @code{TARGET_EDOM}, then compiled code reports | |
4672 | domain errors by calling the library function and letting it report the | |
4673 | error. If mathematical functions on your system use @code{matherr} when | |
4674 | there is an error, then you should leave @code{TARGET_EDOM} undefined so | |
4675 | that @code{matherr} is used normally. | |
4676 | ||
4677 | @findex GEN_ERRNO_RTX | |
4678 | @cindex @code{errno}, implicit usage | |
4679 | @item GEN_ERRNO_RTX | |
4680 | Define this macro as a C expression to create an rtl expression that | |
4681 | refers to the global ``variable'' @code{errno}. (On certain systems, | |
4682 | @code{errno} may not actually be a variable.) If you don't define this | |
4683 | macro, a reasonable default is used. | |
4684 | ||
4685 | @findex TARGET_MEM_FUNCTIONS | |
4686 | @cindex @code{bcopy}, implicit usage | |
4687 | @cindex @code{memcpy}, implicit usage | |
3bcd50fe | 4688 | @cindex @code{memmove}, implicit usage |
feca2ed3 JW |
4689 | @cindex @code{bzero}, implicit usage |
4690 | @cindex @code{memset}, implicit usage | |
4691 | @item TARGET_MEM_FUNCTIONS | |
5490d604 | 4692 | Define this macro if GCC should generate calls to the ISO C |
58c35fc2 JL |
4693 | (and System V) library functions @code{memcpy}, @code{memmove} and |
4694 | @code{memset} rather than the BSD functions @code{bcopy} and @code{bzero}. | |
feca2ed3 JW |
4695 | |
4696 | @findex LIBGCC_NEEDS_DOUBLE | |
4697 | @item LIBGCC_NEEDS_DOUBLE | |
7857f134 ZW |
4698 | Define this macro if @code{float} arguments cannot be passed to library |
4699 | routines (so they must be converted to @code{double}). This macro | |
4700 | affects both how library calls are generated and how the library | |
4701 | routines in @file{libgcc.a} accept their arguments. It is useful on | |
feca2ed3 JW |
4702 | machines where floating and fixed point arguments are passed |
4703 | differently, such as the i860. | |
4704 | ||
feca2ed3 JW |
4705 | @findex NEXT_OBJC_RUNTIME |
4706 | @item NEXT_OBJC_RUNTIME | |
2147b154 | 4707 | Define this macro to generate code for Objective-C message sending using |
feca2ed3 JW |
4708 | the calling convention of the NeXT system. This calling convention |
4709 | involves passing the object, the selector and the method arguments all | |
4710 | at once to the method-lookup library function. | |
4711 | ||
4712 | The default calling convention passes just the object and the selector | |
4713 | to the lookup function, which returns a pointer to the method. | |
4714 | @end table | |
4715 | ||
4716 | @node Addressing Modes | |
4717 | @section Addressing Modes | |
4718 | @cindex addressing modes | |
4719 | ||
4720 | @c prevent bad page break with this line | |
4721 | This is about addressing modes. | |
4722 | ||
4723 | @table @code | |
feca2ed3 | 4724 | @findex HAVE_PRE_INCREMENT |
feca2ed3 | 4725 | @findex HAVE_PRE_DECREMENT |
7a6bd5ae JL |
4726 | @findex HAVE_POST_INCREMENT |
4727 | @findex HAVE_POST_DECREMENT | |
feca2ed3 | 4728 | @item HAVE_PRE_INCREMENT |
feca2ed3 | 4729 | @itemx HAVE_PRE_DECREMENT |
7a6bd5ae JL |
4730 | @itemx HAVE_POST_INCREMENT |
4731 | @itemx HAVE_POST_DECREMENT | |
df2a54e9 | 4732 | A C expression that is nonzero if the machine supports pre-increment, |
7a6bd5ae | 4733 | pre-decrement, post-increment, or post-decrement addressing respectively. |
feca2ed3 | 4734 | |
864bcaa7 | 4735 | @findex HAVE_POST_MODIFY_DISP |
864bcaa7 JL |
4736 | @findex HAVE_PRE_MODIFY_DISP |
4737 | @item HAVE_PRE_MODIFY_DISP | |
7a6bd5ae | 4738 | @itemx HAVE_POST_MODIFY_DISP |
df2a54e9 | 4739 | A C expression that is nonzero if the machine supports pre- or |
7a6bd5ae JL |
4740 | post-address side-effect generation involving constants other than |
4741 | the size of the memory operand. | |
864bcaa7 JL |
4742 | |
4743 | @findex HAVE_POST_MODIFY_REG | |
864bcaa7 JL |
4744 | @findex HAVE_PRE_MODIFY_REG |
4745 | @item HAVE_PRE_MODIFY_REG | |
7a6bd5ae | 4746 | @itemx HAVE_POST_MODIFY_REG |
df2a54e9 | 4747 | A C expression that is nonzero if the machine supports pre- or |
7a6bd5ae | 4748 | post-address side-effect generation involving a register displacement. |
864bcaa7 | 4749 | |
feca2ed3 JW |
4750 | @findex CONSTANT_ADDRESS_P |
4751 | @item CONSTANT_ADDRESS_P (@var{x}) | |
4752 | A C expression that is 1 if the RTX @var{x} is a constant which | |
4753 | is a valid address. On most machines, this can be defined as | |
4754 | @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive | |
4755 | in which constant addresses are supported. | |
4756 | ||
4757 | @findex CONSTANT_P | |
4758 | @code{CONSTANT_P} accepts integer-values expressions whose values are | |
4759 | not explicitly known, such as @code{symbol_ref}, @code{label_ref}, and | |
4760 | @code{high} expressions and @code{const} arithmetic expressions, in | |
4761 | addition to @code{const_int} and @code{const_double} expressions. | |
4762 | ||
4763 | @findex MAX_REGS_PER_ADDRESS | |
4764 | @item MAX_REGS_PER_ADDRESS | |
4765 | A number, the maximum number of registers that can appear in a valid | |
4766 | memory address. Note that it is up to you to specify a value equal to | |
4767 | the maximum number that @code{GO_IF_LEGITIMATE_ADDRESS} would ever | |
4768 | accept. | |
4769 | ||
4770 | @findex GO_IF_LEGITIMATE_ADDRESS | |
4771 | @item GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label}) | |
4772 | A C compound statement with a conditional @code{goto @var{label};} | |
4773 | executed if @var{x} (an RTX) is a legitimate memory address on the | |
4774 | target machine for a memory operand of mode @var{mode}. | |
4775 | ||
4776 | It usually pays to define several simpler macros to serve as | |
4777 | subroutines for this one. Otherwise it may be too complicated to | |
4778 | understand. | |
4779 | ||
4780 | This macro must exist in two variants: a strict variant and a | |
4781 | non-strict one. The strict variant is used in the reload pass. It | |
4782 | must be defined so that any pseudo-register that has not been | |
4783 | allocated a hard register is considered a memory reference. In | |
4784 | contexts where some kind of register is required, a pseudo-register | |
4785 | with no hard register must be rejected. | |
4786 | ||
4787 | The non-strict variant is used in other passes. It must be defined to | |
4788 | accept all pseudo-registers in every context where some kind of | |
4789 | register is required. | |
4790 | ||
4791 | @findex REG_OK_STRICT | |
4792 | Compiler source files that want to use the strict variant of this | |
4793 | macro define the macro @code{REG_OK_STRICT}. You should use an | |
4794 | @code{#ifdef REG_OK_STRICT} conditional to define the strict variant | |
4795 | in that case and the non-strict variant otherwise. | |
4796 | ||
4797 | Subroutines to check for acceptable registers for various purposes (one | |
4798 | for base registers, one for index registers, and so on) are typically | |
4799 | among the subroutines used to define @code{GO_IF_LEGITIMATE_ADDRESS}. | |
4800 | Then only these subroutine macros need have two variants; the higher | |
bd819a4a | 4801 | levels of macros may be the same whether strict or not. |
feca2ed3 JW |
4802 | |
4803 | Normally, constant addresses which are the sum of a @code{symbol_ref} | |
4804 | and an integer are stored inside a @code{const} RTX to mark them as | |
4805 | constant. Therefore, there is no need to recognize such sums | |
4806 | specifically as legitimate addresses. Normally you would simply | |
4807 | recognize any @code{const} as legitimate. | |
4808 | ||
4809 | Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant | |
4810 | sums that are not marked with @code{const}. It assumes that a naked | |
4811 | @code{plus} indicates indexing. If so, then you @emph{must} reject such | |
4812 | naked constant sums as illegitimate addresses, so that none of them will | |
4813 | be given to @code{PRINT_OPERAND_ADDRESS}. | |
4814 | ||
fb49053f | 4815 | @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation |
feca2ed3 JW |
4816 | On some machines, whether a symbolic address is legitimate depends on |
4817 | the section that the address refers to. On these machines, define the | |
fb49053f RH |
4818 | target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information |
4819 | into the @code{symbol_ref}, and then check for it here. When you see a | |
feca2ed3 JW |
4820 | @code{const}, you will have to look inside it to find the |
4821 | @code{symbol_ref} in order to determine the section. @xref{Assembler | |
4822 | Format}. | |
4823 | ||
4824 | @findex saveable_obstack | |
4825 | The best way to modify the name string is by adding text to the | |
4826 | beginning, with suitable punctuation to prevent any ambiguity. Allocate | |
4827 | the new name in @code{saveable_obstack}. You will have to modify | |
4828 | @code{ASM_OUTPUT_LABELREF} to remove and decode the added text and | |
772c5265 RH |
4829 | output the name accordingly, and define @code{TARGET_STRIP_NAME_ENCODING} |
4830 | to access the original name string. | |
feca2ed3 JW |
4831 | |
4832 | You can check the information stored here into the @code{symbol_ref} in | |
4833 | the definitions of the macros @code{GO_IF_LEGITIMATE_ADDRESS} and | |
4834 | @code{PRINT_OPERAND_ADDRESS}. | |
4835 | ||
4836 | @findex REG_OK_FOR_BASE_P | |
4837 | @item REG_OK_FOR_BASE_P (@var{x}) | |
4838 | A C expression that is nonzero if @var{x} (assumed to be a @code{reg} | |
4839 | RTX) is valid for use as a base register. For hard registers, it | |
4840 | should always accept those which the hardware permits and reject the | |
4841 | others. Whether the macro accepts or rejects pseudo registers must be | |
4842 | controlled by @code{REG_OK_STRICT} as described above. This usually | |
4843 | requires two variant definitions, of which @code{REG_OK_STRICT} | |
4844 | controls the one actually used. | |
4845 | ||
861bb6c1 JL |
4846 | @findex REG_MODE_OK_FOR_BASE_P |
4847 | @item REG_MODE_OK_FOR_BASE_P (@var{x}, @var{mode}) | |
4848 | A C expression that is just like @code{REG_OK_FOR_BASE_P}, except that | |
4849 | that expression may examine the mode of the memory reference in | |
4850 | @var{mode}. You should define this macro if the mode of the memory | |
4851 | reference affects whether a register may be used as a base register. If | |
4852 | you define this macro, the compiler will use it instead of | |
4853 | @code{REG_OK_FOR_BASE_P}. | |
4854 | ||
feca2ed3 JW |
4855 | @findex REG_OK_FOR_INDEX_P |
4856 | @item REG_OK_FOR_INDEX_P (@var{x}) | |
4857 | A C expression that is nonzero if @var{x} (assumed to be a @code{reg} | |
4858 | RTX) is valid for use as an index register. | |
4859 | ||
4860 | The difference between an index register and a base register is that | |
4861 | the index register may be scaled. If an address involves the sum of | |
4862 | two registers, neither one of them scaled, then either one may be | |
4863 | labeled the ``base'' and the other the ``index''; but whichever | |
4864 | labeling is used must fit the machine's constraints of which registers | |
4865 | may serve in each capacity. The compiler will try both labelings, | |
4866 | looking for one that is valid, and will reload one or both registers | |
4867 | only if neither labeling works. | |
4868 | ||
b949ea8b JW |
4869 | @findex FIND_BASE_TERM |
4870 | @item FIND_BASE_TERM (@var{x}) | |
4871 | A C expression to determine the base term of address @var{x}. | |
4872 | This macro is used in only one place: `find_base_term' in alias.c. | |
4873 | ||
4874 | It is always safe for this macro to not be defined. It exists so | |
4875 | that alias analysis can understand machine-dependent addresses. | |
4876 | ||
4877 | The typical use of this macro is to handle addresses containing | |
161d7b59 | 4878 | a label_ref or symbol_ref within an UNSPEC@. |
b949ea8b | 4879 | |
feca2ed3 JW |
4880 | @findex LEGITIMIZE_ADDRESS |
4881 | @item LEGITIMIZE_ADDRESS (@var{x}, @var{oldx}, @var{mode}, @var{win}) | |
4882 | A C compound statement that attempts to replace @var{x} with a valid | |
4883 | memory address for an operand of mode @var{mode}. @var{win} will be a | |
4884 | C statement label elsewhere in the code; the macro definition may use | |
4885 | ||
4886 | @example | |
4887 | GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{win}); | |
4888 | @end example | |
4889 | ||
4890 | @noindent | |
4891 | to avoid further processing if the address has become legitimate. | |
4892 | ||
4893 | @findex break_out_memory_refs | |
4894 | @var{x} will always be the result of a call to @code{break_out_memory_refs}, | |
4895 | and @var{oldx} will be the operand that was given to that function to produce | |
4896 | @var{x}. | |
4897 | ||
4898 | The code generated by this macro should not alter the substructure of | |
4899 | @var{x}. If it transforms @var{x} into a more legitimate form, it | |
4900 | should assign @var{x} (which will always be a C variable) a new value. | |
4901 | ||
4902 | It is not necessary for this macro to come up with a legitimate | |
4903 | address. The compiler has standard ways of doing so in all cases. In | |
4904 | fact, it is safe for this macro to do nothing. But often a | |
4905 | machine-dependent strategy can generate better code. | |
4906 | ||
a9a2595b JR |
4907 | @findex LEGITIMIZE_RELOAD_ADDRESS |
4908 | @item LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win}) | |
4909 | A C compound statement that attempts to replace @var{x}, which is an address | |
4910 | that needs reloading, with a valid memory address for an operand of mode | |
4911 | @var{mode}. @var{win} will be a C statement label elsewhere in the code. | |
4912 | It is not necessary to define this macro, but it might be useful for | |
02f52e19 | 4913 | performance reasons. |
a9a2595b JR |
4914 | |
4915 | For example, on the i386, it is sometimes possible to use a single | |
4916 | reload register instead of two by reloading a sum of two pseudo | |
4917 | registers into a register. On the other hand, for number of RISC | |
4918 | processors offsets are limited so that often an intermediate address | |
4919 | needs to be generated in order to address a stack slot. By defining | |
aee96fe9 | 4920 | @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses |
a9a2595b JR |
4921 | generated for adjacent some stack slots can be made identical, and thus |
4922 | be shared. | |
4923 | ||
39bdfaa0 RH |
4924 | @emph{Note}: This macro should be used with caution. It is necessary |
4925 | to know something of how reload works in order to effectively use this, | |
4926 | and it is quite easy to produce macros that build in too much knowledge | |
4927 | of reload internals. | |
a9a2595b | 4928 | |
5f0c590d JL |
4929 | @emph{Note}: This macro must be able to reload an address created by a |
4930 | previous invocation of this macro. If it fails to handle such addresses | |
4931 | then the compiler may generate incorrect code or abort. | |
4932 | ||
a9a2595b | 4933 | @findex push_reload |
39bdfaa0 RH |
4934 | The macro definition should use @code{push_reload} to indicate parts that |
4935 | need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually | |
4936 | suitable to be passed unaltered to @code{push_reload}. | |
a9a2595b | 4937 | |
39bdfaa0 | 4938 | The code generated by this macro must not alter the substructure of |
a9a2595b JR |
4939 | @var{x}. If it transforms @var{x} into a more legitimate form, it |
4940 | should assign @var{x} (which will always be a C variable) a new value. | |
4941 | This also applies to parts that you change indirectly by calling | |
4942 | @code{push_reload}. | |
4943 | ||
39bdfaa0 RH |
4944 | @findex strict_memory_address_p |
4945 | The macro definition may use @code{strict_memory_address_p} to test if | |
4946 | the address has become legitimate. | |
4947 | ||
a9a2595b JR |
4948 | @findex copy_rtx |
4949 | If you want to change only a part of @var{x}, one standard way of doing | |
4950 | this is to use @code{copy_rtx}. Note, however, that is unshares only a | |
4951 | single level of rtl. Thus, if the part to be changed is not at the | |
c771326b | 4952 | top level, you'll need to replace first the top level. |
a9a2595b JR |
4953 | It is not necessary for this macro to come up with a legitimate |
4954 | address; but often a machine-dependent strategy can generate better code. | |
4955 | ||
feca2ed3 JW |
4956 | @findex GO_IF_MODE_DEPENDENT_ADDRESS |
4957 | @item GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label}) | |
4958 | A C statement or compound statement with a conditional @code{goto | |
4959 | @var{label};} executed if memory address @var{x} (an RTX) can have | |
4960 | different meanings depending on the machine mode of the memory | |
4961 | reference it is used for or if the address is valid for some modes | |
4962 | but not others. | |
4963 | ||
4964 | Autoincrement and autodecrement addresses typically have mode-dependent | |
4965 | effects because the amount of the increment or decrement is the size | |
4966 | of the operand being addressed. Some machines have other mode-dependent | |
4967 | addresses. Many RISC machines have no mode-dependent addresses. | |
4968 | ||
4969 | You may assume that @var{addr} is a valid address for the machine. | |
4970 | ||
4971 | @findex LEGITIMATE_CONSTANT_P | |
4972 | @item LEGITIMATE_CONSTANT_P (@var{x}) | |
4973 | A C expression that is nonzero if @var{x} is a legitimate constant for | |
4974 | an immediate operand on the target machine. You can assume that | |
4975 | @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact, | |
4976 | @samp{1} is a suitable definition for this macro on machines where | |
bd819a4a | 4977 | anything @code{CONSTANT_P} is valid. |
feca2ed3 JW |
4978 | @end table |
4979 | ||
4980 | @node Condition Code | |
4981 | @section Condition Code Status | |
4982 | @cindex condition code status | |
4983 | ||
4984 | @c prevent bad page break with this line | |
4985 | This describes the condition code status. | |
4986 | ||
4987 | @findex cc_status | |
4988 | The file @file{conditions.h} defines a variable @code{cc_status} to | |
4989 | describe how the condition code was computed (in case the interpretation of | |
4990 | the condition code depends on the instruction that it was set by). This | |
4991 | variable contains the RTL expressions on which the condition code is | |
4992 | currently based, and several standard flags. | |
4993 | ||
4994 | Sometimes additional machine-specific flags must be defined in the machine | |
4995 | description header file. It can also add additional machine-specific | |
4996 | information by defining @code{CC_STATUS_MDEP}. | |
4997 | ||
4998 | @table @code | |
4999 | @findex CC_STATUS_MDEP | |
5000 | @item CC_STATUS_MDEP | |
5001 | C code for a data type which is used for declaring the @code{mdep} | |
5002 | component of @code{cc_status}. It defaults to @code{int}. | |
5003 | ||
5004 | This macro is not used on machines that do not use @code{cc0}. | |
5005 | ||
5006 | @findex CC_STATUS_MDEP_INIT | |
5007 | @item CC_STATUS_MDEP_INIT | |
5008 | A C expression to initialize the @code{mdep} field to ``empty''. | |
5009 | The default definition does nothing, since most machines don't use | |
5010 | the field anyway. If you want to use the field, you should probably | |
5011 | define this macro to initialize it. | |
5012 | ||
5013 | This macro is not used on machines that do not use @code{cc0}. | |
5014 | ||
5015 | @findex NOTICE_UPDATE_CC | |
5016 | @item NOTICE_UPDATE_CC (@var{exp}, @var{insn}) | |
5017 | A C compound statement to set the components of @code{cc_status} | |
5018 | appropriately for an insn @var{insn} whose body is @var{exp}. It is | |
5019 | this macro's responsibility to recognize insns that set the condition | |
5020 | code as a byproduct of other activity as well as those that explicitly | |
5021 | set @code{(cc0)}. | |
5022 | ||
5023 | This macro is not used on machines that do not use @code{cc0}. | |
5024 | ||
5025 | If there are insns that do not set the condition code but do alter | |
5026 | other machine registers, this macro must check to see whether they | |
5027 | invalidate the expressions that the condition code is recorded as | |
5028 | reflecting. For example, on the 68000, insns that store in address | |
5029 | registers do not set the condition code, which means that usually | |
5030 | @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such | |
5031 | insns. But suppose that the previous insn set the condition code | |
5032 | based on location @samp{a4@@(102)} and the current insn stores a new | |
5033 | value in @samp{a4}. Although the condition code is not changed by | |
5034 | this, it will no longer be true that it reflects the contents of | |
5035 | @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter | |
5036 | @code{cc_status} in this case to say that nothing is known about the | |
5037 | condition code value. | |
5038 | ||
5039 | The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal | |
5040 | with the results of peephole optimization: insns whose patterns are | |
5041 | @code{parallel} RTXs containing various @code{reg}, @code{mem} or | |
5042 | constants which are just the operands. The RTL structure of these | |
5043 | insns is not sufficient to indicate what the insns actually do. What | |
5044 | @code{NOTICE_UPDATE_CC} should do when it sees one is just to run | |
5045 | @code{CC_STATUS_INIT}. | |
5046 | ||
5047 | A possible definition of @code{NOTICE_UPDATE_CC} is to call a function | |
5048 | that looks at an attribute (@pxref{Insn Attributes}) named, for example, | |
5049 | @samp{cc}. This avoids having detailed information about patterns in | |
5050 | two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}. | |
5051 | ||
5052 | @findex EXTRA_CC_MODES | |
5053 | @item EXTRA_CC_MODES | |
a5381466 ZW |
5054 | Condition codes are represented in registers by machine modes of class |
5055 | @code{MODE_CC}. By default, there is just one mode, @code{CCmode}, with | |
5056 | this class. If you need more such modes, create a file named | |
5057 | @file{@var{machine}-modes.def} in your @file{config/@var{machine}} | |
5058 | directory (@pxref{Back End, , Anatomy of a Target Back End}), containing | |
5059 | a list of these modes. Each entry in the list should be a call to the | |
5060 | macro @code{CC}. This macro takes one argument, which is the name of | |
5061 | the mode: it should begin with @samp{CC}. Do not put quotation marks | |
5062 | around the name, or include the trailing @samp{mode}; these are | |
5063 | automatically added. There should not be anything else in the file | |
5064 | except comments. | |
5065 | ||
5066 | A sample @file{@var{machine}-modes.def} file might look like this: | |
feca2ed3 | 5067 | |
feca2ed3 | 5068 | @smallexample |
a5381466 ZW |
5069 | CC (CC_NOOV) /* @r{Comparison only valid if there was no overflow.} */ |
5070 | CC (CCFP) /* @r{Floating point comparison that cannot trap.} */ | |
5071 | CC (CCFPE) /* @r{Floating point comparison that may trap.} */ | |
feca2ed3 JW |
5072 | @end smallexample |
5073 | ||
a5381466 ZW |
5074 | When you create this file, the macro @code{EXTRA_CC_MODES} is |
5075 | automatically defined by @command{configure}, with value @samp{1}. | |
5076 | ||
feca2ed3 JW |
5077 | @findex SELECT_CC_MODE |
5078 | @item SELECT_CC_MODE (@var{op}, @var{x}, @var{y}) | |
5079 | Returns a mode from class @code{MODE_CC} to be used when comparison | |
5080 | operation code @var{op} is applied to rtx @var{x} and @var{y}. For | |
981f6289 | 5081 | example, on the SPARC, @code{SELECT_CC_MODE} is defined as (see |
feca2ed3 JW |
5082 | @pxref{Jump Patterns} for a description of the reason for this |
5083 | definition) | |
5084 | ||
5085 | @smallexample | |
5086 | #define SELECT_CC_MODE(OP,X,Y) \ | |
5087 | (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \ | |
5088 | ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \ | |
5089 | : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \ | |
5090 | || GET_CODE (X) == NEG) \ | |
5091 | ? CC_NOOVmode : CCmode)) | |
5092 | @end smallexample | |
5093 | ||
5094 | You need not define this macro if @code{EXTRA_CC_MODES} is not defined. | |
5095 | ||
5096 | @findex CANONICALIZE_COMPARISON | |
5097 | @item CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1}) | |
8760eaae | 5098 | On some machines not all possible comparisons are defined, but you can |
feca2ed3 JW |
5099 | convert an invalid comparison into a valid one. For example, the Alpha |
5100 | does not have a @code{GT} comparison, but you can use an @code{LT} | |
5101 | comparison instead and swap the order of the operands. | |
5102 | ||
5103 | On such machines, define this macro to be a C statement to do any | |
5104 | required conversions. @var{code} is the initial comparison code | |
5105 | and @var{op0} and @var{op1} are the left and right operands of the | |
5106 | comparison, respectively. You should modify @var{code}, @var{op0}, and | |
5107 | @var{op1} as required. | |
5108 | ||
a3a15b4d | 5109 | GCC will not assume that the comparison resulting from this macro is |
feca2ed3 JW |
5110 | valid but will see if the resulting insn matches a pattern in the |
5111 | @file{md} file. | |
5112 | ||
5113 | You need not define this macro if it would never change the comparison | |
5114 | code or operands. | |
5115 | ||
5116 | @findex REVERSIBLE_CC_MODE | |
5117 | @item REVERSIBLE_CC_MODE (@var{mode}) | |
5118 | A C expression whose value is one if it is always safe to reverse a | |
5119 | comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE} | |
5120 | can ever return @var{mode} for a floating-point inequality comparison, | |
5121 | then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero. | |
5122 | ||
5123 | You need not define this macro if it would always returns zero or if the | |
5124 | floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}. | |
981f6289 | 5125 | For example, here is the definition used on the SPARC, where floating-point |
feca2ed3 JW |
5126 | inequality comparisons are always given @code{CCFPEmode}: |
5127 | ||
5128 | @smallexample | |
5129 | #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode) | |
5130 | @end smallexample | |
5131 | ||
9e7adcb3 JH |
5132 | @findex REVERSE_CONDITION (@var{code}, @var{mode}) |
5133 | A C expression whose value is reversed condition code of the @var{code} for | |
5134 | comparison done in CC_MODE @var{mode}. The macro is used only in case | |
5135 | @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case | |
5136 | machine has some non-standard way how to reverse certain conditionals. For | |
5137 | instance in case all floating point conditions are non-trapping, compiler may | |
5138 | freely convert unordered compares to ordered one. Then definition may look | |
5139 | like: | |
5140 | ||
5141 | @smallexample | |
5142 | #define REVERSE_CONDITION(CODE, MODE) \ | |
c771326b | 5143 | ((MODE) != CCFPmode ? reverse_condition (CODE) \ |
9e7adcb3 JH |
5144 | : reverse_condition_maybe_unordered (CODE)) |
5145 | @end smallexample | |
5146 | ||
7e6d8ba1 AH |
5147 | @findex REVERSE_CONDEXEC_PREDICATES_P |
5148 | @item REVERSE_CONDEXEC_PREDICATES_P (@var{code1}, @var{code2}) | |
5149 | A C expression that returns true if the conditional execution predicate | |
5150 | @var{code1} is the inverse of @var{code2} and vice versa. Define this to | |
5151 | return 0 if the target has conditional execution predicates that cannot be | |
a43f528e AH |
5152 | reversed safely. If no expansion is specified, this macro is defined as |
5153 | follows: | |
7e6d8ba1 AH |
5154 | |
5155 | @smallexample | |
aee96fe9 JM |
5156 | #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \ |
5157 | ((x) == reverse_condition (y)) | |
7e6d8ba1 AH |
5158 | @end smallexample |
5159 | ||
feca2ed3 JW |
5160 | @end table |
5161 | ||
5162 | @node Costs | |
5163 | @section Describing Relative Costs of Operations | |
5164 | @cindex costs of instructions | |
5165 | @cindex relative costs | |
5166 | @cindex speed of instructions | |
5167 | ||
5168 | These macros let you describe the relative speed of various operations | |
5169 | on the target machine. | |
5170 | ||
5171 | @table @code | |
5172 | @findex CONST_COSTS | |
5173 | @item CONST_COSTS (@var{x}, @var{code}, @var{outer_code}) | |
5174 | A part of a C @code{switch} statement that describes the relative costs | |
5175 | of constant RTL expressions. It must contain @code{case} labels for | |
5176 | expression codes @code{const_int}, @code{const}, @code{symbol_ref}, | |
5177 | @code{label_ref} and @code{const_double}. Each case must ultimately | |
5178 | reach a @code{return} statement to return the relative cost of the use | |
5179 | of that kind of constant value in an expression. The cost may depend on | |
5180 | the precise value of the constant, which is available for examination in | |
5181 | @var{x}, and the rtx code of the expression in which it is contained, | |
5182 | found in @var{outer_code}. | |
5183 | ||
5184 | @var{code} is the expression code---redundant, since it can be | |
5185 | obtained with @code{GET_CODE (@var{x})}. | |
5186 | ||
5187 | @findex RTX_COSTS | |
5188 | @findex COSTS_N_INSNS | |
5189 | @item RTX_COSTS (@var{x}, @var{code}, @var{outer_code}) | |
5190 | Like @code{CONST_COSTS} but applies to nonconstant RTL expressions. | |
5191 | This can be used, for example, to indicate how costly a multiply | |
5192 | instruction is. In writing this macro, you can use the construct | |
5193 | @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast | |
5194 | instructions. @var{outer_code} is the code of the expression in which | |
5195 | @var{x} is contained. | |
5196 | ||
5197 | This macro is optional; do not define it if the default cost assumptions | |
5198 | are adequate for the target machine. | |
5199 | ||
8625fab5 KG |
5200 | @findex DEFAULT_RTX_COSTS |
5201 | @item DEFAULT_RTX_COSTS (@var{x}, @var{code}, @var{outer_code}) | |
5202 | This macro, if defined, is called for any case not handled by the | |
5203 | @code{RTX_COSTS} or @code{CONST_COSTS} macros. This eliminates the need | |
5204 | to put case labels into the macro, but the code, or any functions it | |
5205 | calls, must assume that the RTL in @var{x} could be of any type that has | |
5206 | not already been handled. The arguments are the same as for | |
5207 | @code{RTX_COSTS}, and the macro should execute a return statement giving | |
5208 | the cost of any RTL expressions that it can handle. The default cost | |
5209 | calculation is used for any RTL for which this macro does not return a | |
5210 | value. | |
5211 | ||
5212 | This macro is optional; do not define it if the default cost assumptions | |
02f52e19 | 5213 | are adequate for the target machine. |
8625fab5 | 5214 | |
feca2ed3 JW |
5215 | @findex ADDRESS_COST |
5216 | @item ADDRESS_COST (@var{address}) | |
5217 | An expression giving the cost of an addressing mode that contains | |
5218 | @var{address}. If not defined, the cost is computed from | |
5219 | the @var{address} expression and the @code{CONST_COSTS} values. | |
5220 | ||
5221 | For most CISC machines, the default cost is a good approximation of the | |
5222 | true cost of the addressing mode. However, on RISC machines, all | |
5223 | instructions normally have the same length and execution time. Hence | |
5224 | all addresses will have equal costs. | |
5225 | ||
5226 | In cases where more than one form of an address is known, the form with | |
5227 | the lowest cost will be used. If multiple forms have the same, lowest, | |
5228 | cost, the one that is the most complex will be used. | |
5229 | ||
5230 | For example, suppose an address that is equal to the sum of a register | |
5231 | and a constant is used twice in the same basic block. When this macro | |
5232 | is not defined, the address will be computed in a register and memory | |
5233 | references will be indirect through that register. On machines where | |
5234 | the cost of the addressing mode containing the sum is no higher than | |
5235 | that of a simple indirect reference, this will produce an additional | |
5236 | instruction and possibly require an additional register. Proper | |
5237 | specification of this macro eliminates this overhead for such machines. | |
5238 | ||
5239 | Similar use of this macro is made in strength reduction of loops. | |
5240 | ||
5241 | @var{address} need not be valid as an address. In such a case, the cost | |
5242 | is not relevant and can be any value; invalid addresses need not be | |
5243 | assigned a different cost. | |
5244 | ||
5245 | On machines where an address involving more than one register is as | |
5246 | cheap as an address computation involving only one register, defining | |
5247 | @code{ADDRESS_COST} to reflect this can cause two registers to be live | |
5248 | over a region of code where only one would have been if | |
5249 | @code{ADDRESS_COST} were not defined in that manner. This effect should | |
5250 | be considered in the definition of this macro. Equivalent costs should | |
5251 | probably only be given to addresses with different numbers of registers | |
5252 | on machines with lots of registers. | |
5253 | ||
5254 | This macro will normally either not be defined or be defined as a | |
5255 | constant. | |
5256 | ||
5257 | @findex REGISTER_MOVE_COST | |
e56b4594 AO |
5258 | @item REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to}) |
5259 | A C expression for the cost of moving data of mode @var{mode} from a | |
5260 | register in class @var{from} to one in class @var{to}. The classes are | |
5261 | expressed using the enumeration values such as @code{GENERAL_REGS}. A | |
5262 | value of 2 is the default; other values are interpreted relative to | |
5263 | that. | |
feca2ed3 JW |
5264 | |
5265 | It is not required that the cost always equal 2 when @var{from} is the | |
5266 | same as @var{to}; on some machines it is expensive to move between | |
5267 | registers if they are not general registers. | |
5268 | ||
5269 | If reload sees an insn consisting of a single @code{set} between two | |
5270 | hard registers, and if @code{REGISTER_MOVE_COST} applied to their | |
5271 | classes returns a value of 2, reload does not check to ensure that the | |
5272 | constraints of the insn are met. Setting a cost of other than 2 will | |
5273 | allow reload to verify that the constraints are met. You should do this | |
5274 | if the @samp{mov@var{m}} pattern's constraints do not allow such copying. | |
5275 | ||
5276 | @findex MEMORY_MOVE_COST | |
cbd5b9a2 KR |
5277 | @item MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in}) |
5278 | A C expression for the cost of moving data of mode @var{mode} between a | |
5279 | register of class @var{class} and memory; @var{in} is zero if the value | |
df2a54e9 | 5280 | is to be written to memory, nonzero if it is to be read in. This cost |
473fe49b KR |
5281 | is relative to those in @code{REGISTER_MOVE_COST}. If moving between |
5282 | registers and memory is more expensive than between two registers, you | |
5283 | should define this macro to express the relative cost. | |
5284 | ||
a3a15b4d | 5285 | If you do not define this macro, GCC uses a default cost of 4 plus |
38e01259 | 5286 | the cost of copying via a secondary reload register, if one is |
473fe49b KR |
5287 | needed. If your machine requires a secondary reload register to copy |
5288 | between memory and a register of @var{class} but the reload mechanism is | |
5289 | more complex than copying via an intermediate, define this macro to | |
5290 | reflect the actual cost of the move. | |
5291 | ||
a3a15b4d | 5292 | GCC defines the function @code{memory_move_secondary_cost} if |
473fe49b KR |
5293 | secondary reloads are needed. It computes the costs due to copying via |
5294 | a secondary register. If your machine copies from memory using a | |
5295 | secondary register in the conventional way but the default base value of | |
5296 | 4 is not correct for your machine, define this macro to add some other | |
5297 | value to the result of that function. The arguments to that function | |
5298 | are the same as to this macro. | |
cbd5b9a2 | 5299 | |
feca2ed3 JW |
5300 | @findex BRANCH_COST |
5301 | @item BRANCH_COST | |
5302 | A C expression for the cost of a branch instruction. A value of 1 is | |
5303 | the default; other values are interpreted relative to that. | |
5304 | @end table | |
5305 | ||
5306 | Here are additional macros which do not specify precise relative costs, | |
a3a15b4d | 5307 | but only that certain actions are more expensive than GCC would |
feca2ed3 JW |
5308 | ordinarily expect. |
5309 | ||
5310 | @table @code | |
5311 | @findex SLOW_BYTE_ACCESS | |
5312 | @item SLOW_BYTE_ACCESS | |
5313 | Define this macro as a C expression which is nonzero if accessing less | |
e979f9e8 | 5314 | than a word of memory (i.e.@: a @code{char} or a @code{short}) is no |
feca2ed3 JW |
5315 | faster than accessing a word of memory, i.e., if such access |
5316 | require more than one instruction or if there is no difference in cost | |
5317 | between byte and (aligned) word loads. | |
5318 | ||
5319 | When this macro is not defined, the compiler will access a field by | |
5320 | finding the smallest containing object; when it is defined, a fullword | |
5321 | load will be used if alignment permits. Unless bytes accesses are | |
5322 | faster than word accesses, using word accesses is preferable since it | |
5323 | may eliminate subsequent memory access if subsequent accesses occur to | |
5324 | other fields in the same word of the structure, but to different bytes. | |
5325 | ||
feca2ed3 | 5326 | @findex SLOW_UNALIGNED_ACCESS |
5fad8ebf DE |
5327 | @item SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment}) |
5328 | Define this macro to be the value 1 if memory accesses described by the | |
5329 | @var{mode} and @var{alignment} parameters have a cost many times greater | |
5330 | than aligned accesses, for example if they are emulated in a trap | |
5331 | handler. | |
feca2ed3 | 5332 | |
df2a54e9 JM |
5333 | When this macro is nonzero, the compiler will act as if |
5334 | @code{STRICT_ALIGNMENT} were nonzero when generating code for block | |
feca2ed3 | 5335 | moves. This can cause significantly more instructions to be produced. |
df2a54e9 | 5336 | Therefore, do not set this macro nonzero if unaligned accesses only add a |
feca2ed3 JW |
5337 | cycle or two to the time for a memory access. |
5338 | ||
6be57663 | 5339 | If the value of this macro is always zero, it need not be defined. If |
df2a54e9 JM |
5340 | this macro is defined, it should produce a nonzero value when |
5341 | @code{STRICT_ALIGNMENT} is nonzero. | |
feca2ed3 JW |
5342 | |
5343 | @findex DONT_REDUCE_ADDR | |
5344 | @item DONT_REDUCE_ADDR | |
5345 | Define this macro to inhibit strength reduction of memory addresses. | |
5346 | (On some machines, such strength reduction seems to do harm rather | |
5347 | than good.) | |
5348 | ||
5349 | @findex MOVE_RATIO | |
5350 | @item MOVE_RATIO | |
9862dea9 | 5351 | The threshold of number of scalar memory-to-memory move insns, @emph{below} |
c5c76735 | 5352 | which a sequence of insns should be generated instead of a |
feca2ed3 JW |
5353 | string move insn or a library call. Increasing the value will always |
5354 | make code faster, but eventually incurs high cost in increased code size. | |
5355 | ||
c5c76735 JL |
5356 | Note that on machines where the corresponding move insn is a |
5357 | @code{define_expand} that emits a sequence of insns, this macro counts | |
5358 | the number of such sequences. | |
9862dea9 | 5359 | |
feca2ed3 JW |
5360 | If you don't define this, a reasonable default is used. |
5361 | ||
fbe1758d AM |
5362 | @findex MOVE_BY_PIECES_P |
5363 | @item MOVE_BY_PIECES_P (@var{size}, @var{alignment}) | |
5364 | A C expression used to determine whether @code{move_by_pieces} will be used to | |
5365 | copy a chunk of memory, or whether some other block move mechanism | |
6e01bd94 | 5366 | will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less |
fbe1758d AM |
5367 | than @code{MOVE_RATIO}. |
5368 | ||
5369 | @findex MOVE_MAX_PIECES | |
5370 | @item MOVE_MAX_PIECES | |
5371 | A C expression used by @code{move_by_pieces} to determine the largest unit | |
6e01bd94 | 5372 | a load or store used to copy memory is. Defaults to @code{MOVE_MAX}. |
fbe1758d | 5373 | |
78762e3b RS |
5374 | @findex CLEAR_RATIO |
5375 | @item CLEAR_RATIO | |
5376 | The threshold of number of scalar move insns, @emph{below} which a sequence | |
5377 | of insns should be generated to clear memory instead of a string clear insn | |
5378 | or a library call. Increasing the value will always make code faster, but | |
5379 | eventually incurs high cost in increased code size. | |
5380 | ||
5381 | If you don't define this, a reasonable default is used. | |
5382 | ||
5383 | @findex CLEAR_BY_PIECES_P | |
5384 | @item CLEAR_BY_PIECES_P (@var{size}, @var{alignment}) | |
5385 | A C expression used to determine whether @code{clear_by_pieces} will be used | |
5386 | to clear a chunk of memory, or whether some other block clear mechanism | |
5387 | will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less | |
5388 | than @code{CLEAR_RATIO}. | |
5389 | ||
fbe1758d AM |
5390 | @findex USE_LOAD_POST_INCREMENT |
5391 | @item USE_LOAD_POST_INCREMENT (@var{mode}) | |
6e01bd94 MH |
5392 | A C expression used to determine whether a load postincrement is a good |
5393 | thing to use for a given mode. Defaults to the value of | |
5394 | @code{HAVE_POST_INCREMENT}. | |
5395 | ||
5396 | @findex USE_LOAD_POST_DECREMENT | |
5397 | @item USE_LOAD_POST_DECREMENT (@var{mode}) | |
5398 | A C expression used to determine whether a load postdecrement is a good | |
5399 | thing to use for a given mode. Defaults to the value of | |
5400 | @code{HAVE_POST_DECREMENT}. | |
fbe1758d AM |
5401 | |
5402 | @findex USE_LOAD_PRE_INCREMENT | |
5403 | @item USE_LOAD_PRE_INCREMENT (@var{mode}) | |
6e01bd94 MH |
5404 | A C expression used to determine whether a load preincrement is a good |
5405 | thing to use for a given mode. Defaults to the value of | |
5406 | @code{HAVE_PRE_INCREMENT}. | |
5407 | ||
5408 | @findex USE_LOAD_PRE_DECREMENT | |
5409 | @item USE_LOAD_PRE_DECREMENT (@var{mode}) | |
5410 | A C expression used to determine whether a load predecrement is a good | |
5411 | thing to use for a given mode. Defaults to the value of | |
5412 | @code{HAVE_PRE_DECREMENT}. | |
fbe1758d AM |
5413 | |
5414 | @findex USE_STORE_POST_INCREMENT | |
5415 | @item USE_STORE_POST_INCREMENT (@var{mode}) | |
6e01bd94 MH |
5416 | A C expression used to determine whether a store postincrement is a good |
5417 | thing to use for a given mode. Defaults to the value of | |
5418 | @code{HAVE_POST_INCREMENT}. | |
5419 | ||
5420 | @findex USE_STORE_POST_DECREMENT | |
5421 | @item USE_STORE_POST_DECREMENT (@var{mode}) | |
c771326b | 5422 | A C expression used to determine whether a store postdecrement is a good |
6e01bd94 MH |
5423 | thing to use for a given mode. Defaults to the value of |
5424 | @code{HAVE_POST_DECREMENT}. | |
fbe1758d AM |
5425 | |
5426 | @findex USE_STORE_PRE_INCREMENT | |
5427 | @item USE_STORE_PRE_INCREMENT (@var{mode}) | |
6e01bd94 MH |
5428 | This macro is used to determine whether a store preincrement is a good |
5429 | thing to use for a given mode. Defaults to the value of | |
5430 | @code{HAVE_PRE_INCREMENT}. | |
5431 | ||
5432 | @findex USE_STORE_PRE_DECREMENT | |
5433 | @item USE_STORE_PRE_DECREMENT (@var{mode}) | |
5434 | This macro is used to determine whether a store predecrement is a good | |
5435 | thing to use for a given mode. Defaults to the value of | |
5436 | @code{HAVE_PRE_DECREMENT}. | |
fbe1758d | 5437 | |
feca2ed3 JW |
5438 | @findex NO_FUNCTION_CSE |
5439 | @item NO_FUNCTION_CSE | |
5440 | Define this macro if it is as good or better to call a constant | |
5441 | function address than to call an address kept in a register. | |
5442 | ||
5443 | @findex NO_RECURSIVE_FUNCTION_CSE | |
5444 | @item NO_RECURSIVE_FUNCTION_CSE | |
5445 | Define this macro if it is as good or better for a function to call | |
5446 | itself with an explicit address than to call an address kept in a | |
5447 | register. | |
feca2ed3 JW |
5448 | @end table |
5449 | ||
c237e94a ZW |
5450 | @node Scheduling |
5451 | @section Adjusting the Instruction Scheduler | |
5452 | ||
5453 | The instruction scheduler may need a fair amount of machine-specific | |
5454 | adjustment in order to produce good code. GCC provides several target | |
5455 | hooks for this purpose. It is usually enough to define just a few of | |
5456 | them: try the first ones in this list first. | |
5457 | ||
5458 | @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void) | |
fae15c93 VM |
5459 | This hook returns the maximum number of instructions that can ever |
5460 | issue at the same time on the target machine. The default is one. | |
5461 | Although the insn scheduler can define itself the possibility of issue | |
5462 | an insn on the same cycle, the value can serve as an additional | |
5463 | constraint to issue insns on the same simulated processor cycle (see | |
5464 | hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}). | |
5465 | This value must be constant over the entire compilation. If you need | |
5466 | it to vary depending on what the instructions are, you must use | |
c237e94a | 5467 | @samp{TARGET_SCHED_VARIABLE_ISSUE}. |
fae15c93 | 5468 | |
4226378a PK |
5469 | For the automaton based pipeline interface, you could define this hook |
5470 | to return the value of the macro @code{MAX_DFA_ISSUE_RATE}. | |
c237e94a ZW |
5471 | @end deftypefn |
5472 | ||
5473 | @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more}) | |
5474 | This hook is executed by the scheduler after it has scheduled an insn | |
5475 | from the ready list. It should return the number of insns which can | |
5476 | still be issued in the current cycle. Normally this is | |
5477 | @samp{@w{@var{more} - 1}}. You should define this hook if some insns | |
5478 | take more machine resources than others, so that fewer insns can follow | |
5479 | them in the same cycle. @var{file} is either a null pointer, or a stdio | |
5480 | stream to write any debug output to. @var{verbose} is the verbose level | |
5481 | provided by @option{-fsched-verbose-@var{n}}. @var{insn} is the | |
5482 | instruction that was scheduled. | |
5483 | @end deftypefn | |
5484 | ||
5485 | @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost}) | |
fae15c93 VM |
5486 | This function corrects the value of @var{cost} based on the |
5487 | relationship between @var{insn} and @var{dep_insn} through the | |
5488 | dependence @var{link}. It should return the new value. The default | |
5489 | is to make no adjustment to @var{cost}. This can be used for example | |
5490 | to specify to the scheduler using the traditional pipeline description | |
c237e94a | 5491 | that an output- or anti-dependence does not incur the same cost as a |
fae15c93 VM |
5492 | data-dependence. If the scheduler using the automaton based pipeline |
5493 | description, the cost of anti-dependence is zero and the cost of | |
5494 | output-dependence is maximum of one and the difference of latency | |
5495 | times of the first and the second insns. If these values are not | |
5496 | acceptable, you could use the hook to modify them too. See also | |
5497 | @pxref{Automaton pipeline description}. | |
c237e94a ZW |
5498 | @end deftypefn |
5499 | ||
5500 | @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority}) | |
5501 | This hook adjusts the integer scheduling priority @var{priority} of | |
5502 | @var{insn}. It should return the new priority. Reduce the priority to | |
5503 | execute @var{insn} earlier, increase the priority to execute @var{insn} | |
5504 | later. Do not define this hook if you do not need to adjust the | |
5505 | scheduling priorities of insns. | |
5506 | @end deftypefn | |
5507 | ||
5508 | @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock}) | |
5509 | This hook is executed by the scheduler after it has scheduled the ready | |
5510 | list, to allow the machine description to reorder it (for example to | |
5511 | combine two small instructions together on @samp{VLIW} machines). | |
5512 | @var{file} is either a null pointer, or a stdio stream to write any | |
5513 | debug output to. @var{verbose} is the verbose level provided by | |
5514 | @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready | |
5515 | list of instructions that are ready to be scheduled. @var{n_readyp} is | |
5516 | a pointer to the number of elements in the ready list. The scheduler | |
5517 | reads the ready list in reverse order, starting with | |
5518 | @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock} | |
5519 | is the timer tick of the scheduler. You may modify the ready list and | |
5520 | the number of ready insns. The return value is the number of insns that | |
5521 | can issue this cycle; normally this is just @code{issue_rate}. See also | |
5522 | @samp{TARGET_SCHED_REORDER2}. | |
5523 | @end deftypefn | |
5524 | ||
5525 | @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock}) | |
5526 | Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That | |
5527 | function is called whenever the scheduler starts a new cycle. This one | |
5528 | is called once per iteration over a cycle, immediately after | |
5529 | @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and | |
5530 | return the number of insns to be scheduled in the same cycle. Defining | |
5531 | this hook can be useful if there are frequent situations where | |
5532 | scheduling one insn causes other insns to become ready in the same | |
5533 | cycle. These other insns can then be taken into account properly. | |
5534 | @end deftypefn | |
5535 | ||
5536 | @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready}) | |
5537 | This hook is executed by the scheduler at the beginning of each block of | |
5538 | instructions that are to be scheduled. @var{file} is either a null | |
5539 | pointer, or a stdio stream to write any debug output to. @var{verbose} | |
5540 | is the verbose level provided by @option{-fsched-verbose-@var{n}}. | |
5541 | @var{max_ready} is the maximum number of insns in the current scheduling | |
5542 | region that can be live at the same time. This can be used to allocate | |
5543 | scratch space if it is needed, e.g. by @samp{TARGET_SCHED_REORDER}. | |
5544 | @end deftypefn | |
5545 | ||
5546 | @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose}) | |
5547 | This hook is executed by the scheduler at the end of each block of | |
5548 | instructions that are to be scheduled. It can be used to perform | |
5549 | cleanup of any actions done by the other scheduling hooks. @var{file} | |
5550 | is either a null pointer, or a stdio stream to write any debug output | |
5551 | to. @var{verbose} is the verbose level provided by | |
5552 | @option{-fsched-verbose-@var{n}}. | |
5553 | @end deftypefn | |
5554 | ||
fae15c93 VM |
5555 | @deftypefn {Target Hook} int TARGET_SCHED_USE_DFA_PIPELINE_INTERFACE (void) |
5556 | This hook is called many times during insn scheduling. If the hook | |
5557 | returns nonzero, the automaton based pipeline description is used for | |
5558 | insn scheduling. Otherwise the traditional pipeline description is | |
5559 | used. The default is usage of the traditional pipeline description. | |
5560 | ||
5561 | You should also remember that to simplify the insn scheduler sources | |
5562 | an empty traditional pipeline description interface is generated even | |
5563 | if there is no a traditional pipeline description in the @file{.md} | |
5564 | file. The same is true for the automaton based pipeline description. | |
5565 | That means that you should be accurate in defining the hook. | |
5566 | @end deftypefn | |
5567 | ||
5568 | @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void) | |
5569 | The hook returns an RTL insn. The automaton state used in the | |
5570 | pipeline hazard recognizer is changed as if the insn were scheduled | |
5571 | when the new simulated processor cycle starts. Usage of the hook may | |
5572 | simplify the automaton pipeline description for some @acronym{VLIW} | |
5573 | processors. If the hook is defined, it is used only for the automaton | |
5574 | based pipeline description. The default is not to change the state | |
5575 | when the new simulated processor cycle starts. | |
5576 | @end deftypefn | |
5577 | ||
5578 | @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void) | |
5579 | The hook can be used to initialize data used by the previous hook. | |
5580 | @end deftypefn | |
5581 | ||
5582 | @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void) | |
5583 | The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used | |
5584 | to changed the state as if the insn were scheduled when the new | |
5585 | simulated processor cycle finishes. | |
5586 | @end deftypefn | |
5587 | ||
5588 | @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void) | |
5589 | The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but | |
5590 | used to initialize data used by the previous hook. | |
5591 | @end deftypefn | |
5592 | ||
5593 | @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void) | |
5594 | This hook controls better choosing an insn from the ready insn queue | |
5595 | for the @acronym{DFA}-based insn scheduler. Usually the scheduler | |
5596 | chooses the first insn from the queue. If the hook returns a positive | |
5597 | value, an additional scheduler code tries all permutations of | |
5598 | @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()} | |
5599 | subsequent ready insns to choose an insn whose issue will result in | |
5600 | maximal number of issued insns on the same cycle. For the | |
5601 | @acronym{VLIW} processor, the code could actually solve the problem of | |
5602 | packing simple insns into the @acronym{VLIW} insn. Of course, if the | |
5603 | rules of @acronym{VLIW} packing are described in the automaton. | |
5604 | ||
5605 | This code also could be used for superscalar @acronym{RISC} | |
5606 | processors. Let us consider a superscalar @acronym{RISC} processor | |
5607 | with 3 pipelines. Some insns can be executed in pipelines @var{A} or | |
5608 | @var{B}, some insns can be executed only in pipelines @var{B} or | |
5609 | @var{C}, and one insn can be executed in pipeline @var{B}. The | |
5610 | processor may issue the 1st insn into @var{A} and the 2nd one into | |
5611 | @var{B}. In this case, the 3rd insn will wait for freeing @var{B} | |
5612 | until the next cycle. If the scheduler issues the 3rd insn the first, | |
5613 | the processor could issue all 3 insns per cycle. | |
5614 | ||
5615 | Actually this code demonstrates advantages of the automaton based | |
5616 | pipeline hazard recognizer. We try quickly and easy many insn | |
5617 | schedules to choose the best one. | |
5618 | ||
5619 | The default is no multipass scheduling. | |
5620 | @end deftypefn | |
5621 | ||
5622 | @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_BUBBLES (void) | |
5623 | The @acronym{DFA}-based scheduler could take the insertion of nop | |
5624 | operations for better insn scheduling into account. It can be done | |
5625 | only if the multi-pass insn scheduling works (see hook | |
5626 | @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD}). | |
5627 | ||
5628 | Let us consider a @acronym{VLIW} processor insn with 3 slots. Each | |
5629 | insn can be placed only in one of the three slots. We have 3 ready | |
5630 | insns @var{A}, @var{B}, and @var{C}. @var{A} and @var{C} can be | |
5631 | placed only in the 1st slot, @var{B} can be placed only in the 3rd | |
5632 | slot. We described the automaton which does not permit empty slot | |
5633 | gaps between insns (usually such description is simpler). Without | |
5634 | this code the scheduler would place each insn in 3 separate | |
5635 | @acronym{VLIW} insns. If the scheduler places a nop insn into the 2nd | |
5636 | slot, it could place the 3 insns into 2 @acronym{VLIW} insns. What is | |
5637 | the nop insn is returned by hook @samp{TARGET_SCHED_DFA_BUBBLE}. Hook | |
5638 | @samp{TARGET_SCHED_INIT_DFA_BUBBLES} can be used to initialize or | |
5639 | create the nop insns. | |
5640 | ||
5641 | You should remember that the scheduler does not insert the nop insns. | |
5642 | It is not wise because of the following optimizations. The scheduler | |
5643 | only considers such possibility to improve the result schedule. The | |
5644 | nop insns should be inserted lately, e.g. on the final phase. | |
5645 | @end deftypefn | |
5646 | ||
5647 | @deftypefn {Target Hook} rtx TARGET_SCHED_DFA_BUBBLE (int @var{index}) | |
5648 | This hook @samp{FIRST_CYCLE_MULTIPASS_SCHEDULING} is used to insert | |
5649 | nop operations for better insn scheduling when @acronym{DFA}-based | |
5650 | scheduler makes multipass insn scheduling (see also description of | |
5651 | hook @samp{TARGET_SCHED_INIT_DFA_BUBBLES}). This hook | |
5652 | returns a nop insn with given @var{index}. The indexes start with | |
5653 | zero. The hook should return @code{NULL} if there are no more nop | |
5654 | insns with indexes greater than given index. | |
5655 | @end deftypefn | |
5656 | ||
5657 | Macros in the following table are generated by the program | |
5658 | @file{genattr} and can be useful for writing the hooks. | |
5659 | ||
5660 | @table @code | |
5661 | @findex TRADITIONAL_PIPELINE_INTERFACE | |
5662 | @item TRADITIONAL_PIPELINE_INTERFACE | |
5663 | The macro definition is generated if there is a traditional pipeline | |
5664 | description in @file{.md} file. You should also remember that to | |
5665 | simplify the insn scheduler sources an empty traditional pipeline | |
5666 | description interface is generated even if there is no a traditional | |
5667 | pipeline description in the @file{.md} file. The macro can be used to | |
5668 | distinguish the two types of the traditional interface. | |
5669 | ||
5670 | @findex DFA_PIPELINE_INTERFACE | |
5671 | @item DFA_PIPELINE_INTERFACE | |
5672 | The macro definition is generated if there is an automaton pipeline | |
5673 | description in @file{.md} file. You should also remember that to | |
5674 | simplify the insn scheduler sources an empty automaton pipeline | |
5675 | description interface is generated even if there is no an automaton | |
5676 | pipeline description in the @file{.md} file. The macro can be used to | |
5677 | distinguish the two types of the automaton interface. | |
5678 | ||
5679 | @findex MAX_DFA_ISSUE_RATE | |
5680 | @item MAX_DFA_ISSUE_RATE | |
5681 | The macro definition is generated in the automaton based pipeline | |
5682 | description interface. Its value is calculated from the automaton | |
5683 | based pipeline description and is equal to maximal number of all insns | |
5684 | described in constructions @samp{define_insn_reservation} which can be | |
5685 | issued on the same processor cycle. | |
5686 | ||
5687 | @end table | |
5688 | ||
feca2ed3 JW |
5689 | @node Sections |
5690 | @section Dividing the Output into Sections (Texts, Data, @dots{}) | |
5691 | @c the above section title is WAY too long. maybe cut the part between | |
5692 | @c the (...)? --mew 10feb93 | |
5693 | ||
5694 | An object file is divided into sections containing different types of | |
5695 | data. In the most common case, there are three sections: the @dfn{text | |
5696 | section}, which holds instructions and read-only data; the @dfn{data | |
5697 | section}, which holds initialized writable data; and the @dfn{bss | |
5698 | section}, which holds uninitialized data. Some systems have other kinds | |
5699 | of sections. | |
5700 | ||
5701 | The compiler must tell the assembler when to switch sections. These | |
5702 | macros control what commands to output to tell the assembler this. You | |
5703 | can also define additional sections. | |
5704 | ||
5705 | @table @code | |
5706 | @findex TEXT_SECTION_ASM_OP | |
5707 | @item TEXT_SECTION_ASM_OP | |
047c1c92 HPN |
5708 | A C expression whose value is a string, including spacing, containing the |
5709 | assembler operation that should precede instructions and read-only data. | |
5710 | Normally @code{"\t.text"} is right. | |
feca2ed3 | 5711 | |
33c09f2f RH |
5712 | @findex TEXT_SECTION |
5713 | @item TEXT_SECTION | |
5714 | A C statement that switches to the default section containing instructions. | |
5715 | Normally this is not needed, as simply defining @code{TEXT_SECTION_ASM_OP} | |
5716 | is enough. The MIPS port uses this to sort all functions after all data | |
5717 | declarations. | |
5718 | ||
194734e9 JH |
5719 | @findex HOT_TEXT_SECTION_NAME |
5720 | @item HOT_TEXT_SECTION_NAME | |
5721 | If defined, a C string constant for the name of the section containing most | |
5722 | frequently executed functions of the program. If not defined, GCC will provide | |
5723 | a default definition if the target supports named sections. | |
5724 | ||
5725 | @findex UNLIKELY_EXECUTED_TEXT_SECTION_NAME | |
5726 | @item UNLIKELY_EXECUTED_TEXT_SECTION_NAME | |
5727 | If defined, a C string constant for the name of the section containing unlikely | |
5728 | executed functions in the program. | |
5729 | ||
feca2ed3 JW |
5730 | @findex DATA_SECTION_ASM_OP |
5731 | @item DATA_SECTION_ASM_OP | |
047c1c92 HPN |
5732 | A C expression whose value is a string, including spacing, containing the |
5733 | assembler operation to identify the following data as writable initialized | |
5734 | data. Normally @code{"\t.data"} is right. | |
feca2ed3 | 5735 | |
d48bc59a RH |
5736 | @findex READONLY_DATA_SECTION_ASM_OP |
5737 | @item READONLY_DATA_SECTION_ASM_OP | |
5738 | A C expression whose value is a string, including spacing, containing the | |
5739 | assembler operation to identify the following data as read-only initialized | |
5740 | data. | |
5741 | ||
5742 | @findex READONLY_DATA_SECTION | |
5743 | @item READONLY_DATA_SECTION | |
5744 | A macro naming a function to call to switch to the proper section for | |
5745 | read-only data. The default is to use @code{READONLY_DATA_SECTION_ASM_OP} | |
5746 | if defined, else fall back to @code{text_section}. | |
5747 | ||
5748 | The most common definition will be @code{data_section}, if the target | |
5749 | does not have a special read-only data section, and does not put data | |
5750 | in the text section. | |
5751 | ||
feca2ed3 JW |
5752 | @findex SHARED_SECTION_ASM_OP |
5753 | @item SHARED_SECTION_ASM_OP | |
047c1c92 HPN |
5754 | If defined, a C expression whose value is a string, including spacing, |
5755 | containing the assembler operation to identify the following data as | |
5756 | shared data. If not defined, @code{DATA_SECTION_ASM_OP} will be used. | |
feca2ed3 JW |
5757 | |
5758 | @findex BSS_SECTION_ASM_OP | |
5759 | @item BSS_SECTION_ASM_OP | |
047c1c92 HPN |
5760 | If defined, a C expression whose value is a string, including spacing, |
5761 | containing the assembler operation to identify the following data as | |
5762 | uninitialized global data. If not defined, and neither | |
5763 | @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined, | |
5764 | uninitialized global data will be output in the data section if | |
630d3d5a | 5765 | @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be |
047c1c92 | 5766 | used. |
feca2ed3 JW |
5767 | |
5768 | @findex SHARED_BSS_SECTION_ASM_OP | |
5769 | @item SHARED_BSS_SECTION_ASM_OP | |
047c1c92 HPN |
5770 | If defined, a C expression whose value is a string, including spacing, |
5771 | containing the assembler operation to identify the following data as | |
5772 | uninitialized global shared data. If not defined, and | |
5773 | @code{BSS_SECTION_ASM_OP} is, the latter will be used. | |
feca2ed3 JW |
5774 | |
5775 | @findex INIT_SECTION_ASM_OP | |
5776 | @item INIT_SECTION_ASM_OP | |
047c1c92 HPN |
5777 | If defined, a C expression whose value is a string, including spacing, |
5778 | containing the assembler operation to identify the following data as | |
5779 | initialization code. If not defined, GCC will assume such a section does | |
5780 | not exist. | |
feca2ed3 | 5781 | |
1b2dd04a AO |
5782 | @findex FINI_SECTION_ASM_OP |
5783 | @item FINI_SECTION_ASM_OP | |
047c1c92 HPN |
5784 | If defined, a C expression whose value is a string, including spacing, |
5785 | containing the assembler operation to identify the following data as | |
5786 | finalization code. If not defined, GCC will assume such a section does | |
5787 | not exist. | |
1b2dd04a AO |
5788 | |
5789 | @findex CRT_CALL_STATIC_FUNCTION | |
cea3bd3e RH |
5790 | @item CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function}) |
5791 | If defined, an ASM statement that switches to a different section | |
5792 | via @var{section_op}, calls @var{function}, and switches back to | |
5793 | the text section. This is used in @file{crtstuff.c} if | |
5794 | @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls | |
5795 | to initialization and finalization functions from the init and fini | |
5796 | sections. By default, this macro uses a simple function call. Some | |
1b2dd04a AO |
5797 | ports need hand-crafted assembly code to avoid dependencies on |
5798 | registers initialized in the function prologue or to ensure that | |
5799 | constant pools don't end up too far way in the text section. | |
5800 | ||
cea3bd3e RH |
5801 | @findex FORCE_CODE_SECTION_ALIGN |
5802 | @item FORCE_CODE_SECTION_ALIGN | |
5803 | If defined, an ASM statement that aligns a code section to some | |
5804 | arbitrary boundary. This is used to force all fragments of the | |
5805 | @code{.init} and @code{.fini} sections to have to same alignment | |
5806 | and thus prevent the linker from having to add any padding. | |
5807 | ||
feca2ed3 JW |
5808 | @findex EXTRA_SECTIONS |
5809 | @findex in_text | |
5810 | @findex in_data | |
5811 | @item EXTRA_SECTIONS | |
5812 | A list of names for sections other than the standard two, which are | |
5813 | @code{in_text} and @code{in_data}. You need not define this macro | |
5814 | on a system with no other sections (that GCC needs to use). | |
5815 | ||
5816 | @findex EXTRA_SECTION_FUNCTIONS | |
5817 | @findex text_section | |
5818 | @findex data_section | |
5819 | @item EXTRA_SECTION_FUNCTIONS | |
5820 | One or more functions to be defined in @file{varasm.c}. These | |
5821 | functions should do jobs analogous to those of @code{text_section} and | |
5822 | @code{data_section}, for your additional sections. Do not define this | |
5823 | macro if you do not define @code{EXTRA_SECTIONS}. | |
5824 | ||
feca2ed3 JW |
5825 | @findex JUMP_TABLES_IN_TEXT_SECTION |
5826 | @item JUMP_TABLES_IN_TEXT_SECTION | |
df2a54e9 | 5827 | Define this macro to be an expression with a nonzero value if jump |
75197b37 BS |
5828 | tables (for @code{tablejump} insns) should be output in the text |
5829 | section, along with the assembler instructions. Otherwise, the | |
5830 | readonly data section is used. | |
feca2ed3 JW |
5831 | |
5832 | This macro is irrelevant if there is no separate readonly data section. | |
feca2ed3 JW |
5833 | @end table |
5834 | ||
ae46c4e0 RH |
5835 | @deftypefn {Target Hook} void TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align}) |
5836 | Switches to the appropriate section for output of @var{exp}. You can | |
5837 | assume that @var{exp} is either a @code{VAR_DECL} node or a constant of | |
5838 | some sort. @var{reloc} indicates whether the initial value of @var{exp} | |
5839 | requires link-time relocations. Bit 0 is set when variable contains | |
5840 | local relocations only, while bit 1 is set for global relocations. | |
5841 | Select the section by calling @code{data_section} or one of the | |
5842 | alternatives for other sections. @var{align} is the constant alignment | |
5843 | in bits. | |
5844 | ||
5845 | The default version of this function takes care of putting read-only | |
5846 | variables in @code{readonly_data_section}. | |
5847 | @end deftypefn | |
5848 | ||
5849 | @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc}) | |
5850 | Build up a unique section name, expressed as a @code{STRING_CST} node, | |
5851 | and assign it to @samp{DECL_SECTION_NAME (@var{decl})}. | |
5852 | As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether | |
5853 | the initial value of @var{exp} requires link-time relocations. | |
5854 | ||
5855 | The default version of this function appends the symbol name to the | |
5856 | ELF section name that would normally be used for the symbol. For | |
5857 | example, the function @code{foo} would be placed in @code{.text.foo}. | |
5858 | Whatever the actual target object format, this is often good enough. | |
5859 | @end deftypefn | |
5860 | ||
b64a1b53 RH |
5861 | @deftypefn {Target Hook} void TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align}) |
5862 | Switches to the appropriate section for output of constant pool entry | |
5863 | @var{x} in @var{mode}. You can assume that @var{x} is some kind of | |
5864 | constant in RTL@. The argument @var{mode} is redundant except in the | |
5865 | case of a @code{const_int} rtx. Select the section by calling | |
5866 | @code{readonly_data_section} or one of the alternatives for other | |
5867 | sections. @var{align} is the constant alignment in bits. | |
5868 | ||
5869 | The default version of this function takes care of putting symbolic | |
5870 | constants in @code{flag_pic} mode in @code{data_section} and everything | |
5871 | else in @code{readonly_data_section}. | |
5872 | @end deftypefn | |
5873 | ||
fb49053f RH |
5874 | @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, int @var{new_decl_p}) |
5875 | Define this hook if references to a symbol or a constant must be | |
5876 | treated differently depending on something about the variable or | |
5877 | function named by the symbol (such as what section it is in). | |
5878 | ||
5879 | The hook is executed under two circumstances. One is immediately after | |
5880 | the rtl for @var{decl} that represents a variable or a function has been | |
5881 | created and stored in @code{DECL_RTL(@var{decl})}. The value of the rtl | |
5882 | will be a @code{mem} whose address is a @code{symbol_ref}. The other is | |
5883 | immediately after the rtl for @var{decl} that represents a constant has | |
5884 | been created and stored in @code{TREE_CST_RTL (@var{decl})}. The macro | |
5885 | is called once for each distinct constant in a source file. | |
5886 | ||
5887 | The @var{new_decl_p} argument will be true if this is the first time | |
5888 | that @code{ENCODE_SECTION_INFO} has been invoked on this decl. It will | |
5889 | be false for subsequent invocations, which will happen for duplicate | |
5890 | declarations. Whether or not anything must be done for the duplicate | |
5891 | declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}. | |
5892 | ||
5893 | @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO} | |
5894 | The usual thing for this hook to do is to record a flag in the | |
5895 | @code{symbol_ref} (such as @code{SYMBOL_REF_FLAG}) or to store a | |
5896 | modified name string in the @code{symbol_ref} (if one bit is not | |
5897 | enough information). | |
5898 | @end deftypefn | |
5899 | ||
772c5265 RH |
5900 | @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name) |
5901 | Decode @var{name} and return the real name part, sans | |
5902 | the characters that @code{TARGET_ENCODE_SECTION_INFO} | |
5903 | may have added. | |
5904 | @end deftypefn | |
5905 | ||
47754fd5 RH |
5906 | @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp}) |
5907 | Returns true if @var{exp} should be placed into a ``small data'' section. | |
5908 | The default version of this hook always returns false. | |
5909 | @end deftypefn | |
5910 | ||
e2a6476e DE |
5911 | @deftypevar {Target Hook} bool TARGET_HAVE_SRODATA_SECTION |
5912 | Contains the value true if the target places read-only | |
5913 | ``small data'' into a separate section. The default value is false. | |
5914 | @end deftypevar | |
5915 | ||
47754fd5 RH |
5916 | @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp}) |
5917 | Returns true if @var{exp} names an object for which name resolution | |
5918 | rules must resolve to the current ``module'' (dynamic shared library | |
5919 | or executable image). | |
5920 | ||
5921 | The default version of this hook implements the name resolution rules | |
5922 | for ELF, which has a looser model of global name binding than other | |
5923 | currently supported object file formats. | |
5924 | @end deftypefn | |
5925 | ||
e2a6476e DE |
5926 | @deftypevar {Target Hook} bool TARGET_HAVE_TLS |
5927 | Contains the value true if the target supports thread-local storage. | |
5928 | The default value is false. | |
5929 | @end deftypevar | |
5930 | ||
5931 | ||
feca2ed3 JW |
5932 | @node PIC |
5933 | @section Position Independent Code | |
5934 | @cindex position independent code | |
5935 | @cindex PIC | |
5936 | ||
5937 | This section describes macros that help implement generation of position | |
5938 | independent code. Simply defining these macros is not enough to | |
5939 | generate valid PIC; you must also add support to the macros | |
5940 | @code{GO_IF_LEGITIMATE_ADDRESS} and @code{PRINT_OPERAND_ADDRESS}, as | |
5941 | well as @code{LEGITIMIZE_ADDRESS}. You must modify the definition of | |
5942 | @samp{movsi} to do something appropriate when the source operand | |
5943 | contains a symbolic address. You may also need to alter the handling of | |
5944 | switch statements so that they use relative addresses. | |
5945 | @c i rearranged the order of the macros above to try to force one of | |
5946 | @c them to the next line, to eliminate an overfull hbox. --mew 10feb93 | |
5947 | ||
5948 | @table @code | |
5949 | @findex PIC_OFFSET_TABLE_REGNUM | |
5950 | @item PIC_OFFSET_TABLE_REGNUM | |
5951 | The register number of the register used to address a table of static | |
5952 | data addresses in memory. In some cases this register is defined by a | |
161d7b59 | 5953 | processor's ``application binary interface'' (ABI)@. When this macro |
feca2ed3 JW |
5954 | is defined, RTL is generated for this register once, as with the stack |
5955 | pointer and frame pointer registers. If this macro is not defined, it | |
5956 | is up to the machine-dependent files to allocate such a register (if | |
003b9f78 | 5957 | necessary). Note that this register must be fixed when in use (e.g.@: |
12beba6f | 5958 | when @code{flag_pic} is true). |
feca2ed3 JW |
5959 | |
5960 | @findex PIC_OFFSET_TABLE_REG_CALL_CLOBBERED | |
5961 | @item PIC_OFFSET_TABLE_REG_CALL_CLOBBERED | |
5962 | Define this macro if the register defined by | |
5963 | @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define | |
ed4db1ee | 5964 | this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined. |
feca2ed3 JW |
5965 | |
5966 | @findex FINALIZE_PIC | |
5967 | @item FINALIZE_PIC | |
5968 | By generating position-independent code, when two different programs (A | |
5969 | and B) share a common library (libC.a), the text of the library can be | |
5970 | shared whether or not the library is linked at the same address for both | |
5971 | programs. In some of these environments, position-independent code | |
5972 | requires not only the use of different addressing modes, but also | |
5973 | special code to enable the use of these addressing modes. | |
5974 | ||
5975 | The @code{FINALIZE_PIC} macro serves as a hook to emit these special | |
5976 | codes once the function is being compiled into assembly code, but not | |
5977 | before. (It is not done before, because in the case of compiling an | |
5978 | inline function, it would lead to multiple PIC prologues being | |
5979 | included in functions which used inline functions and were compiled to | |
5980 | assembly language.) | |
5981 | ||
5982 | @findex LEGITIMATE_PIC_OPERAND_P | |
5983 | @item LEGITIMATE_PIC_OPERAND_P (@var{x}) | |
5984 | A C expression that is nonzero if @var{x} is a legitimate immediate | |
5985 | operand on the target machine when generating position independent code. | |
5986 | You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not | |
5987 | check this. You can also assume @var{flag_pic} is true, so you need not | |
5988 | check it either. You need not define this macro if all constants | |
5989 | (including @code{SYMBOL_REF}) can be immediate operands when generating | |
5990 | position independent code. | |
5991 | @end table | |
5992 | ||
5993 | @node Assembler Format | |
5994 | @section Defining the Output Assembler Language | |
5995 | ||
5996 | This section describes macros whose principal purpose is to describe how | |
648c546a | 5997 | to write instructions in assembler language---rather than what the |
feca2ed3 JW |
5998 | instructions do. |
5999 | ||
6000 | @menu | |
6001 | * File Framework:: Structural information for the assembler file. | |
6002 | * Data Output:: Output of constants (numbers, strings, addresses). | |
6003 | * Uninitialized Data:: Output of uninitialized variables. | |
6004 | * Label Output:: Output and generation of labels. | |
6005 | * Initialization:: General principles of initialization | |
6006 | and termination routines. | |
6007 | * Macros for Initialization:: | |
6008 | Specific macros that control the handling of | |
6009 | initialization and termination routines. | |
6010 | * Instruction Output:: Output of actual instructions. | |
6011 | * Dispatch Tables:: Output of jump tables. | |
6012 | * Exception Region Output:: Output of exception region code. | |
6013 | * Alignment Output:: Pseudo ops for alignment and skipping data. | |
6014 | @end menu | |
6015 | ||
6016 | @node File Framework | |
6017 | @subsection The Overall Framework of an Assembler File | |
6018 | @cindex assembler format | |
6019 | @cindex output of assembler code | |
6020 | ||
6021 | @c prevent bad page break with this line | |
6022 | This describes the overall framework of an assembler file. | |
6023 | ||
6024 | @table @code | |
6025 | @findex ASM_FILE_START | |
6026 | @item ASM_FILE_START (@var{stream}) | |
6027 | A C expression which outputs to the stdio stream @var{stream} | |
6028 | some appropriate text to go at the start of an assembler file. | |
6029 | ||
6030 | Normally this macro is defined to output a line containing | |
6031 | @samp{#NO_APP}, which is a comment that has no effect on most | |
6032 | assemblers but tells the GNU assembler that it can save time by not | |
6033 | checking for certain assembler constructs. | |
6034 | ||
6035 | On systems that use SDB, it is necessary to output certain commands; | |
6036 | see @file{attasm.h}. | |
6037 | ||
6038 | @findex ASM_FILE_END | |
6039 | @item ASM_FILE_END (@var{stream}) | |
6040 | A C expression which outputs to the stdio stream @var{stream} | |
6041 | some appropriate text to go at the end of an assembler file. | |
6042 | ||
6043 | If this macro is not defined, the default is to output nothing | |
6044 | special at the end of the file. Most systems don't require any | |
6045 | definition. | |
6046 | ||
6047 | On systems that use SDB, it is necessary to output certain commands; | |
6048 | see @file{attasm.h}. | |
6049 | ||
feca2ed3 JW |
6050 | @findex ASM_COMMENT_START |
6051 | @item ASM_COMMENT_START | |
6052 | A C string constant describing how to begin a comment in the target | |
6053 | assembler language. The compiler assumes that the comment will end at | |
6054 | the end of the line. | |
6055 | ||
6056 | @findex ASM_APP_ON | |
6057 | @item ASM_APP_ON | |
6058 | A C string constant for text to be output before each @code{asm} | |
6059 | statement or group of consecutive ones. Normally this is | |
6060 | @code{"#APP"}, which is a comment that has no effect on most | |
6061 | assemblers but tells the GNU assembler that it must check the lines | |
6062 | that follow for all valid assembler constructs. | |
6063 | ||
6064 | @findex ASM_APP_OFF | |
6065 | @item ASM_APP_OFF | |
6066 | A C string constant for text to be output after each @code{asm} | |
6067 | statement or group of consecutive ones. Normally this is | |
6068 | @code{"#NO_APP"}, which tells the GNU assembler to resume making the | |
6069 | time-saving assumptions that are valid for ordinary compiler output. | |
6070 | ||
6071 | @findex ASM_OUTPUT_SOURCE_FILENAME | |
6072 | @item ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name}) | |
6073 | A C statement to output COFF information or DWARF debugging information | |
6074 | which indicates that filename @var{name} is the current source file to | |
6075 | the stdio stream @var{stream}. | |
6076 | ||
6077 | This macro need not be defined if the standard form of output | |
6078 | for the file format in use is appropriate. | |
6079 | ||
e9a25f70 | 6080 | @findex OUTPUT_QUOTED_STRING |
8760eaae | 6081 | @item OUTPUT_QUOTED_STRING (@var{stream}, @var{string}) |
e9a25f70 JL |
6082 | A C statement to output the string @var{string} to the stdio stream |
6083 | @var{stream}. If you do not call the function @code{output_quoted_string} | |
a3a15b4d | 6084 | in your config files, GCC will only call it to output filenames to |
e9a25f70 JL |
6085 | the assembler source. So you can use it to canonicalize the format |
6086 | of the filename using this macro. | |
6087 | ||
feca2ed3 JW |
6088 | @findex ASM_OUTPUT_SOURCE_LINE |
6089 | @item ASM_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}) | |
6090 | A C statement to output DBX or SDB debugging information before code | |
6091 | for line number @var{line} of the current source file to the | |
6092 | stdio stream @var{stream}. | |
6093 | ||
6094 | This macro need not be defined if the standard form of debugging | |
6095 | information for the debugger in use is appropriate. | |
6096 | ||
6097 | @findex ASM_OUTPUT_IDENT | |
6098 | @item ASM_OUTPUT_IDENT (@var{stream}, @var{string}) | |
6099 | A C statement to output something to the assembler file to handle a | |
6100 | @samp{#ident} directive containing the text @var{string}. If this | |
6101 | macro is not defined, nothing is output for a @samp{#ident} directive. | |
6102 | ||
feca2ed3 JW |
6103 | @findex OBJC_PROLOGUE |
6104 | @item OBJC_PROLOGUE | |
6105 | A C statement to output any assembler statements which are required to | |
2147b154 SS |
6106 | precede any Objective-C object definitions or message sending. The |
6107 | statement is executed only when compiling an Objective-C program. | |
feca2ed3 JW |
6108 | @end table |
6109 | ||
7c262518 RH |
6110 | @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align}) |
6111 | Output assembly directives to switch to section @var{name}. The section | |
6112 | should have attributes as specified by @var{flags}, which is a bit mask | |
6113 | of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align} | |
df2a54e9 | 6114 | is nonzero, it contains an alignment in bytes to be used for the section, |
f282ffb3 | 6115 | otherwise some target default should be used. Only targets that must |
7c262518 RH |
6116 | specify an alignment within the section directive need pay attention to |
6117 | @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}. | |
6118 | @end deftypefn | |
6119 | ||
6120 | @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS | |
6121 | This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}. | |
6122 | @end deftypefn | |
6123 | ||
6124 | @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc}) | |
6125 | Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION} | |
6126 | based on a variable or function decl, a section name, and whether or not the | |
6127 | declaration's initializer may contain runtime relocations. @var{decl} may be | |
6128 | null, in which case read-write data should be assumed. | |
6129 | ||
6130 | The default version if this function handles choosing code vs data, | |
6131 | read-only vs read-write data, and @code{flag_pic}. You should only | |
6132 | need to override this if your target has special flags that might be | |
6133 | set via @code{__attribute__}. | |
6134 | @end deftypefn | |
6135 | ||
feca2ed3 JW |
6136 | @need 2000 |
6137 | @node Data Output | |
6138 | @subsection Output of Data | |
6139 | ||
301d03af RS |
6140 | |
6141 | @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP | |
6142 | @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP | |
6143 | @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP | |
6144 | @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP | |
6145 | @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP | |
6146 | @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP | |
6147 | @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP | |
6148 | @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP | |
6149 | @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP | |
6150 | These hooks specify assembly directives for creating certain kinds | |
6151 | of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a | |
6152 | byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an | |
6153 | aligned two-byte object, and so on. Any of the hooks may be | |
6154 | @code{NULL}, indicating that no suitable directive is available. | |
6155 | ||
6156 | The compiler will print these strings at the start of a new line, | |
6157 | followed immediately by the object's initial value. In most cases, | |
6158 | the string should contain a tab, a pseudo-op, and then another tab. | |
6159 | @end deftypevr | |
6160 | ||
6161 | @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p}) | |
6162 | The @code{assemble_integer} function uses this hook to output an | |
6163 | integer object. @var{x} is the object's value, @var{size} is its size | |
6164 | in bytes and @var{aligned_p} indicates whether it is aligned. The | |
6165 | function should return @code{true} if it was able to output the | |
6166 | object. If it returns false, @code{assemble_integer} will try to | |
6167 | split the object into smaller parts. | |
6168 | ||
6169 | The default implementation of this hook will use the | |
6170 | @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false} | |
6171 | when the relevant string is @code{NULL}. | |
6172 | @end deftypefn | |
feca2ed3 JW |
6173 | |
6174 | @table @code | |
422be3c3 AO |
6175 | @findex OUTPUT_ADDR_CONST_EXTRA |
6176 | @item OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail}) | |
6177 | A C statement to recognize @var{rtx} patterns that | |
6178 | @code{output_addr_const} can't deal with, and output assembly code to | |
6179 | @var{stream} corresponding to the pattern @var{x}. This may be used to | |
6180 | allow machine-dependent @code{UNSPEC}s to appear within constants. | |
6181 | ||
6182 | If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must | |
6183 | @code{goto fail}, so that a standard error message is printed. If it | |
6184 | prints an error message itself, by calling, for example, | |
6185 | @code{output_operand_lossage}, it may just complete normally. | |
6186 | ||
feca2ed3 JW |
6187 | @findex ASM_OUTPUT_ASCII |
6188 | @item ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len}) | |
6189 | A C statement to output to the stdio stream @var{stream} an assembler | |
6190 | instruction to assemble a string constant containing the @var{len} | |
6191 | bytes at @var{ptr}. @var{ptr} will be a C expression of type | |
6192 | @code{char *} and @var{len} a C expression of type @code{int}. | |
6193 | ||
6194 | If the assembler has a @code{.ascii} pseudo-op as found in the | |
6195 | Berkeley Unix assembler, do not define the macro | |
6196 | @code{ASM_OUTPUT_ASCII}. | |
6197 | ||
67231816 RH |
6198 | @findex ASM_OUTPUT_FDESC |
6199 | @item ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n}) | |
6200 | A C statement to output word @var{n} of a function descriptor for | |
6201 | @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS} | |
6202 | is defined, and is otherwise unused. | |
6203 | ||
861bb6c1 JL |
6204 | @findex CONSTANT_POOL_BEFORE_FUNCTION |
6205 | @item CONSTANT_POOL_BEFORE_FUNCTION | |
6206 | You may define this macro as a C expression. You should define the | |
df2a54e9 | 6207 | expression to have a nonzero value if GCC should output the constant |
861bb6c1 | 6208 | pool for a function before the code for the function, or a zero value if |
a3a15b4d JL |
6209 | GCC should output the constant pool after the function. If you do |
6210 | not define this macro, the usual case, GCC will output the constant | |
861bb6c1 JL |
6211 | pool before the function. |
6212 | ||
feca2ed3 | 6213 | @findex ASM_OUTPUT_POOL_PROLOGUE |
8760eaae | 6214 | @item ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size}) |
feca2ed3 JW |
6215 | A C statement to output assembler commands to define the start of the |
6216 | constant pool for a function. @var{funname} is a string giving | |
6217 | the name of the function. Should the return type of the function | |
6218 | be required, it can be obtained via @var{fundecl}. @var{size} | |
6219 | is the size, in bytes, of the constant pool that will be written | |
6220 | immediately after this call. | |
6221 | ||
6222 | If no constant-pool prefix is required, the usual case, this macro need | |
6223 | not be defined. | |
6224 | ||
6225 | @findex ASM_OUTPUT_SPECIAL_POOL_ENTRY | |
6226 | @item ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto}) | |
6227 | A C statement (with or without semicolon) to output a constant in the | |
6228 | constant pool, if it needs special treatment. (This macro need not do | |
6229 | anything for RTL expressions that can be output normally.) | |
6230 | ||
6231 | The argument @var{file} is the standard I/O stream to output the | |
6232 | assembler code on. @var{x} is the RTL expression for the constant to | |
6233 | output, and @var{mode} is the machine mode (in case @var{x} is a | |
6234 | @samp{const_int}). @var{align} is the required alignment for the value | |
6235 | @var{x}; you should output an assembler directive to force this much | |
6236 | alignment. | |
6237 | ||
6238 | The argument @var{labelno} is a number to use in an internal label for | |
6239 | the address of this pool entry. The definition of this macro is | |
6240 | responsible for outputting the label definition at the proper place. | |
6241 | Here is how to do this: | |
6242 | ||
6243 | @example | |
6244 | ASM_OUTPUT_INTERNAL_LABEL (@var{file}, "LC", @var{labelno}); | |
6245 | @end example | |
6246 | ||
6247 | When you output a pool entry specially, you should end with a | |
6248 | @code{goto} to the label @var{jumpto}. This will prevent the same pool | |
6249 | entry from being output a second time in the usual manner. | |
6250 | ||
6251 | You need not define this macro if it would do nothing. | |
6252 | ||
861bb6c1 JL |
6253 | @findex CONSTANT_AFTER_FUNCTION_P |
6254 | @item CONSTANT_AFTER_FUNCTION_P (@var{exp}) | |
6255 | Define this macro as a C expression which is nonzero if the constant | |
6256 | @var{exp}, of type @code{tree}, should be output after the code for a | |
6257 | function. The compiler will normally output all constants before the | |
161d7b59 | 6258 | function; you need not define this macro if this is OK@. |
861bb6c1 JL |
6259 | |
6260 | @findex ASM_OUTPUT_POOL_EPILOGUE | |
6261 | @item ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size}) | |
6262 | A C statement to output assembler commands to at the end of the constant | |
6263 | pool for a function. @var{funname} is a string giving the name of the | |
6264 | function. Should the return type of the function be required, you can | |
6265 | obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the | |
a3a15b4d | 6266 | constant pool that GCC wrote immediately before this call. |
861bb6c1 JL |
6267 | |
6268 | If no constant-pool epilogue is required, the usual case, you need not | |
6269 | define this macro. | |
6270 | ||
feca2ed3 JW |
6271 | @findex IS_ASM_LOGICAL_LINE_SEPARATOR |
6272 | @item IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}) | |
6273 | Define this macro as a C expression which is nonzero if @var{C} is | |
6274 | used as a logical line separator by the assembler. | |
6275 | ||
6276 | If you do not define this macro, the default is that only | |
6277 | the character @samp{;} is treated as a logical line separator. | |
feca2ed3 JW |
6278 | @end table |
6279 | ||
8ca83838 | 6280 | @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN |
baed53ac | 6281 | @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN |
17b53c33 NB |
6282 | These target hooks are C string constants, describing the syntax in the |
6283 | assembler for grouping arithmetic expressions. If not overridden, they | |
6284 | default to normal parentheses, which is correct for most assemblers. | |
8ca83838 | 6285 | @end deftypevr |
17b53c33 | 6286 | |
feca2ed3 JW |
6287 | These macros are provided by @file{real.h} for writing the definitions |
6288 | of @code{ASM_OUTPUT_DOUBLE} and the like: | |
6289 | ||
6290 | @table @code | |
6291 | @item REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l}) | |
6292 | @itemx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l}) | |
6293 | @itemx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l}) | |
6294 | @findex REAL_VALUE_TO_TARGET_SINGLE | |
6295 | @findex REAL_VALUE_TO_TARGET_DOUBLE | |
6296 | @findex REAL_VALUE_TO_TARGET_LONG_DOUBLE | |
6297 | These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the target's | |
b216cd4a ZW |
6298 | floating point representation, and store its bit pattern in the variable |
6299 | @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE}, this variable should | |
6300 | be a simple @code{long int}. For the others, it should be an array of | |
6301 | @code{long int}. The number of elements in this array is determined by | |
6302 | the size of the desired target floating point data type: 32 bits of it | |
6303 | go in each @code{long int} array element. Each array element holds 32 | |
6304 | bits of the result, even if @code{long int} is wider than 32 bits on the | |
6305 | host machine. | |
feca2ed3 JW |
6306 | |
6307 | The array element values are designed so that you can print them out | |
6308 | using @code{fprintf} in the order they should appear in the target | |
6309 | machine's memory. | |
6310 | ||
4b67a274 | 6311 | @item REAL_VALUE_TO_DECIMAL (@var{x}, @var{string}, @var{digits}) |
feca2ed3 JW |
6312 | @findex REAL_VALUE_TO_DECIMAL |
6313 | This macro converts @var{x}, of type @code{REAL_VALUE_TYPE}, to a | |
6314 | decimal number and stores it as a string into @var{string}. | |
6315 | You must pass, as @var{string}, the address of a long enough block | |
6316 | of space to hold the result. | |
6317 | ||
4b67a274 RH |
6318 | The argument @var{digits} is the number of decimal digits to print, |
6319 | or @minus{}1 to indicate ``enough'', i.e. @code{DECIMAL_DIG} for | |
6320 | for the target. | |
feca2ed3 JW |
6321 | @end table |
6322 | ||
6323 | @node Uninitialized Data | |
6324 | @subsection Output of Uninitialized Variables | |
6325 | ||
6326 | Each of the macros in this section is used to do the whole job of | |
6327 | outputting a single uninitialized variable. | |
6328 | ||
6329 | @table @code | |
6330 | @findex ASM_OUTPUT_COMMON | |
6331 | @item ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
6332 | A C statement (sans semicolon) to output to the stdio stream | |
6333 | @var{stream} the assembler definition of a common-label named | |
6334 | @var{name} whose size is @var{size} bytes. The variable @var{rounded} | |
6335 | is the size rounded up to whatever alignment the caller wants. | |
6336 | ||
6337 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
6338 | output the name itself; before and after that, output the additional | |
6339 | assembler syntax for defining the name, and a newline. | |
6340 | ||
6341 | This macro controls how the assembler definitions of uninitialized | |
6342 | common global variables are output. | |
6343 | ||
6344 | @findex ASM_OUTPUT_ALIGNED_COMMON | |
6345 | @item ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment}) | |
6346 | Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a | |
6347 | separate, explicit argument. If you define this macro, it is used in | |
6348 | place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in | |
6349 | handling the required alignment of the variable. The alignment is specified | |
6350 | as the number of bits. | |
6351 | ||
e9a25f70 JL |
6352 | @findex ASM_OUTPUT_ALIGNED_DECL_COMMON |
6353 | @item ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) | |
6354 | Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the | |
6355 | variable to be output, if there is one, or @code{NULL_TREE} if there | |
8760eaae | 6356 | is no corresponding variable. If you define this macro, GCC will use it |
e9a25f70 JL |
6357 | in place of both @code{ASM_OUTPUT_COMMON} and |
6358 | @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see | |
6359 | the variable's decl in order to chose what to output. | |
6360 | ||
feca2ed3 JW |
6361 | @findex ASM_OUTPUT_SHARED_COMMON |
6362 | @item ASM_OUTPUT_SHARED_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
6363 | If defined, it is similar to @code{ASM_OUTPUT_COMMON}, except that it | |
6364 | is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_COMMON} | |
6365 | will be used. | |
6366 | ||
6367 | @findex ASM_OUTPUT_BSS | |
6368 | @item ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded}) | |
6369 | A C statement (sans semicolon) to output to the stdio stream | |
6370 | @var{stream} the assembler definition of uninitialized global @var{decl} named | |
6371 | @var{name} whose size is @var{size} bytes. The variable @var{rounded} | |
6372 | is the size rounded up to whatever alignment the caller wants. | |
6373 | ||
6374 | Try to use function @code{asm_output_bss} defined in @file{varasm.c} when | |
6375 | defining this macro. If unable, use the expression | |
6376 | @code{assemble_name (@var{stream}, @var{name})} to output the name itself; | |
6377 | before and after that, output the additional assembler syntax for defining | |
6378 | the name, and a newline. | |
6379 | ||
6380 | This macro controls how the assembler definitions of uninitialized global | |
6381 | variables are output. This macro exists to properly support languages like | |
aee96fe9 | 6382 | C++ which do not have @code{common} data. However, this macro currently |
feca2ed3 JW |
6383 | is not defined for all targets. If this macro and |
6384 | @code{ASM_OUTPUT_ALIGNED_BSS} are not defined then @code{ASM_OUTPUT_COMMON} | |
e9a25f70 JL |
6385 | or @code{ASM_OUTPUT_ALIGNED_COMMON} or |
6386 | @code{ASM_OUTPUT_ALIGNED_DECL_COMMON} is used. | |
feca2ed3 JW |
6387 | |
6388 | @findex ASM_OUTPUT_ALIGNED_BSS | |
6389 | @item ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) | |
6390 | Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a | |
6391 | separate, explicit argument. If you define this macro, it is used in | |
6392 | place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in | |
6393 | handling the required alignment of the variable. The alignment is specified | |
6394 | as the number of bits. | |
6395 | ||
6396 | Try to use function @code{asm_output_aligned_bss} defined in file | |
6397 | @file{varasm.c} when defining this macro. | |
6398 | ||
6399 | @findex ASM_OUTPUT_SHARED_BSS | |
6400 | @item ASM_OUTPUT_SHARED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded}) | |
6401 | If defined, it is similar to @code{ASM_OUTPUT_BSS}, except that it | |
6402 | is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_BSS} | |
6403 | will be used. | |
6404 | ||
6405 | @findex ASM_OUTPUT_LOCAL | |
6406 | @item ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
6407 | A C statement (sans semicolon) to output to the stdio stream | |
6408 | @var{stream} the assembler definition of a local-common-label named | |
6409 | @var{name} whose size is @var{size} bytes. The variable @var{rounded} | |
6410 | is the size rounded up to whatever alignment the caller wants. | |
6411 | ||
6412 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
6413 | output the name itself; before and after that, output the additional | |
6414 | assembler syntax for defining the name, and a newline. | |
6415 | ||
6416 | This macro controls how the assembler definitions of uninitialized | |
6417 | static variables are output. | |
6418 | ||
6419 | @findex ASM_OUTPUT_ALIGNED_LOCAL | |
6420 | @item ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment}) | |
6421 | Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a | |
6422 | separate, explicit argument. If you define this macro, it is used in | |
6423 | place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in | |
6424 | handling the required alignment of the variable. The alignment is specified | |
6425 | as the number of bits. | |
6426 | ||
e9a25f70 JL |
6427 | @findex ASM_OUTPUT_ALIGNED_DECL_LOCAL |
6428 | @item ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment}) | |
6429 | Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the | |
6430 | variable to be output, if there is one, or @code{NULL_TREE} if there | |
8760eaae | 6431 | is no corresponding variable. If you define this macro, GCC will use it |
e9a25f70 JL |
6432 | in place of both @code{ASM_OUTPUT_DECL} and |
6433 | @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see | |
6434 | the variable's decl in order to chose what to output. | |
6435 | ||
feca2ed3 JW |
6436 | @findex ASM_OUTPUT_SHARED_LOCAL |
6437 | @item ASM_OUTPUT_SHARED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded}) | |
6438 | If defined, it is similar to @code{ASM_OUTPUT_LOCAL}, except that it | |
6439 | is used when @var{name} is shared. If not defined, @code{ASM_OUTPUT_LOCAL} | |
6440 | will be used. | |
6441 | @end table | |
6442 | ||
6443 | @node Label Output | |
6444 | @subsection Output and Generation of Labels | |
6445 | ||
6446 | @c prevent bad page break with this line | |
6447 | This is about outputting labels. | |
6448 | ||
6449 | @table @code | |
6450 | @findex ASM_OUTPUT_LABEL | |
6451 | @findex assemble_name | |
6452 | @item ASM_OUTPUT_LABEL (@var{stream}, @var{name}) | |
6453 | A C statement (sans semicolon) to output to the stdio stream | |
6454 | @var{stream} the assembler definition of a label named @var{name}. | |
6455 | Use the expression @code{assemble_name (@var{stream}, @var{name})} to | |
6456 | output the name itself; before and after that, output the additional | |
4ad5e05d KG |
6457 | assembler syntax for defining the name, and a newline. A default |
6458 | definition of this macro is provided which is correct for most systems. | |
feca2ed3 | 6459 | |
2be2ac70 ZW |
6460 | @findex SIZE_ASM_OP |
6461 | @item SIZE_ASM_OP | |
6462 | A C string containing the appropriate assembler directive to specify the | |
6463 | size of a symbol, without any arguments. On systems that use ELF, the | |
6464 | default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other | |
6465 | systems, the default is not to define this macro. | |
6466 | ||
6467 | Define this macro only if it is correct to use the default definitions | |
6468 | of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE} | |
6469 | for your system. If you need your own custom definitions of those | |
6470 | macros, or if you do not need explicit symbol sizes at all, do not | |
6471 | define this macro. | |
6472 | ||
6473 | @findex ASM_OUTPUT_SIZE_DIRECTIVE | |
6474 | @item ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size}) | |
6475 | A C statement (sans semicolon) to output to the stdio stream | |
6476 | @var{stream} a directive telling the assembler that the size of the | |
6477 | symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}. | |
6478 | If you define @code{SIZE_ASM_OP}, a default definition of this macro is | |
6479 | provided. | |
6480 | ||
6481 | @findex ASM_OUTPUT_MEASURED_SIZE | |
99086d59 | 6482 | @item ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name}) |
2be2ac70 ZW |
6483 | A C statement (sans semicolon) to output to the stdio stream |
6484 | @var{stream} a directive telling the assembler to calculate the size of | |
99086d59 ZW |
6485 | the symbol @var{name} by subtracting its address from the current |
6486 | address. | |
6487 | ||
6488 | If you define @code{SIZE_ASM_OP}, a default definition of this macro is | |
6489 | provided. The default assumes that the assembler recognizes a special | |
6490 | @samp{.} symbol as referring to the current address, and can calculate | |
6491 | the difference between this and another symbol. If your assembler does | |
6492 | not recognize @samp{.} or cannot do calculations with it, you will need | |
6493 | to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique. | |
2be2ac70 ZW |
6494 | |
6495 | @findex TYPE_ASM_OP | |
6496 | @item TYPE_ASM_OP | |
6497 | A C string containing the appropriate assembler directive to specify the | |
6498 | type of a symbol, without any arguments. On systems that use ELF, the | |
6499 | default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other | |
6500 | systems, the default is not to define this macro. | |
6501 | ||
6502 | Define this macro only if it is correct to use the default definition of | |
6503 | @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own | |
6504 | custom definition of this macro, or if you do not need explicit symbol | |
6505 | types at all, do not define this macro. | |
6506 | ||
6507 | @findex TYPE_OPERAND_FMT | |
6508 | @item TYPE_OPERAND_FMT | |
6509 | A C string which specifies (using @code{printf} syntax) the format of | |
6510 | the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the | |
6511 | default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems, | |
6512 | the default is not to define this macro. | |
6513 | ||
6514 | Define this macro only if it is correct to use the default definition of | |
6515 | @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own | |
6516 | custom definition of this macro, or if you do not need explicit symbol | |
6517 | types at all, do not define this macro. | |
6518 | ||
6519 | @findex ASM_OUTPUT_TYPE_DIRECTIVE | |
6520 | @item ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type}) | |
6521 | A C statement (sans semicolon) to output to the stdio stream | |
6522 | @var{stream} a directive telling the assembler that the type of the | |
6523 | symbol @var{name} is @var{type}. @var{type} is a C string; currently, | |
6524 | that string is always either @samp{"function"} or @samp{"object"}, but | |
6525 | you should not count on this. | |
6526 | ||
6527 | If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default | |
6528 | definition of this macro is provided. | |
6529 | ||
feca2ed3 JW |
6530 | @findex ASM_DECLARE_FUNCTION_NAME |
6531 | @item ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl}) | |
6532 | A C statement (sans semicolon) to output to the stdio stream | |
6533 | @var{stream} any text necessary for declaring the name @var{name} of a | |
6534 | function which is being defined. This macro is responsible for | |
6535 | outputting the label definition (perhaps using | |
6536 | @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the | |
6537 | @code{FUNCTION_DECL} tree node representing the function. | |
6538 | ||
6539 | If this macro is not defined, then the function name is defined in the | |
6540 | usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). | |
6541 | ||
2be2ac70 ZW |
6542 | You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition |
6543 | of this macro. | |
6544 | ||
feca2ed3 JW |
6545 | @findex ASM_DECLARE_FUNCTION_SIZE |
6546 | @item ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl}) | |
6547 | A C statement (sans semicolon) to output to the stdio stream | |
6548 | @var{stream} any text necessary for declaring the size of a function | |
6549 | which is being defined. The argument @var{name} is the name of the | |
6550 | function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node | |
6551 | representing the function. | |
6552 | ||
6553 | If this macro is not defined, then the function size is not defined. | |
6554 | ||
2be2ac70 ZW |
6555 | You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition |
6556 | of this macro. | |
6557 | ||
feca2ed3 JW |
6558 | @findex ASM_DECLARE_OBJECT_NAME |
6559 | @item ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl}) | |
6560 | A C statement (sans semicolon) to output to the stdio stream | |
6561 | @var{stream} any text necessary for declaring the name @var{name} of an | |
6562 | initialized variable which is being defined. This macro must output the | |
6563 | label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument | |
6564 | @var{decl} is the @code{VAR_DECL} tree node representing the variable. | |
6565 | ||
6566 | If this macro is not defined, then the variable name is defined in the | |
6567 | usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}). | |
6568 | ||
2be2ac70 ZW |
6569 | You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or |
6570 | @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro. | |
6571 | ||
1cb36a98 RH |
6572 | @findex ASM_DECLARE_REGISTER_GLOBAL |
6573 | @item ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name}) | |
6574 | A C statement (sans semicolon) to output to the stdio stream | |
6575 | @var{stream} any text necessary for claiming a register @var{regno} | |
6576 | for a global variable @var{decl} with name @var{name}. | |
6577 | ||
6578 | If you don't define this macro, that is equivalent to defining it to do | |
6579 | nothing. | |
6580 | ||
feca2ed3 JW |
6581 | @findex ASM_FINISH_DECLARE_OBJECT |
6582 | @item ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend}) | |
6583 | A C statement (sans semicolon) to finish up declaring a variable name | |
6584 | once the compiler has processed its initializer fully and thus has had a | |
6585 | chance to determine the size of an array when controlled by an | |
6586 | initializer. This is used on systems where it's necessary to declare | |
6587 | something about the size of the object. | |
6588 | ||
6589 | If you don't define this macro, that is equivalent to defining it to do | |
6590 | nothing. | |
6591 | ||
2be2ac70 ZW |
6592 | You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or |
6593 | @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro. | |
b65d23aa | 6594 | @end table |
2be2ac70 | 6595 | |
5eb99654 KG |
6596 | @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name}) |
6597 | This target hook is a function to output to the stdio stream | |
feca2ed3 | 6598 | @var{stream} some commands that will make the label @var{name} global; |
5eb99654 | 6599 | that is, available for reference from other files. |
feca2ed3 | 6600 | |
5eb99654 KG |
6601 | The default implementation relies on a proper definition of |
6602 | @code{GLOBAL_ASM_OP}. | |
b65d23aa | 6603 | @end deftypefn |
072cdaed | 6604 | |
b65d23aa | 6605 | @table @code |
feca2ed3 | 6606 | @findex ASM_WEAKEN_LABEL |
79c4e63f | 6607 | @item ASM_WEAKEN_LABEL (@var{stream}, @var{name}) |
feca2ed3 JW |
6608 | A C statement (sans semicolon) to output to the stdio stream |
6609 | @var{stream} some commands that will make the label @var{name} weak; | |
6610 | that is, available for reference from other files but only used if | |
6611 | no other definition is available. Use the expression | |
6612 | @code{assemble_name (@var{stream}, @var{name})} to output the name | |
6613 | itself; before and after that, output the additional assembler syntax | |
6614 | for making that name weak, and a newline. | |
6615 | ||
79c4e63f AM |
6616 | If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not |
6617 | support weak symbols and you should not define the @code{SUPPORTS_WEAK} | |
6618 | macro. | |
6619 | ||
6620 | @findex ASM_WEAKEN_DECL | |
6621 | @item ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value}) | |
6622 | Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and | |
6623 | @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function | |
6624 | or variable decl. If @var{value} is not @code{NULL}, this C statement | |
6625 | should output to the stdio stream @var{stream} assembler code which | |
6626 | defines (equates) the weak symbol @var{name} to have the value | |
6627 | @var{value}. If @var{value} is @code{NULL}, it should output commands | |
6628 | to make @var{name} weak. | |
feca2ed3 JW |
6629 | |
6630 | @findex SUPPORTS_WEAK | |
6631 | @item SUPPORTS_WEAK | |
6632 | A C expression which evaluates to true if the target supports weak symbols. | |
6633 | ||
6634 | If you don't define this macro, @file{defaults.h} provides a default | |
79c4e63f AM |
6635 | definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL} |
6636 | is defined, the default definition is @samp{1}; otherwise, it is | |
6637 | @samp{0}. Define this macro if you want to control weak symbol support | |
6638 | with a compiler flag such as @option{-melf}. | |
feca2ed3 JW |
6639 | |
6640 | @findex MAKE_DECL_ONE_ONLY (@var{decl}) | |
6641 | @item MAKE_DECL_ONE_ONLY | |
6642 | A C statement (sans semicolon) to mark @var{decl} to be emitted as a | |
6643 | public symbol such that extra copies in multiple translation units will | |
6644 | be discarded by the linker. Define this macro if your object file | |
6645 | format provides support for this concept, such as the @samp{COMDAT} | |
6646 | section flags in the Microsoft Windows PE/COFF format, and this support | |
6647 | requires changes to @var{decl}, such as putting it in a separate section. | |
6648 | ||
e9a25f70 JL |
6649 | @findex SUPPORTS_ONE_ONLY |
6650 | @item SUPPORTS_ONE_ONLY | |
feca2ed3 JW |
6651 | A C expression which evaluates to true if the target supports one-only |
6652 | semantics. | |
6653 | ||
6654 | If you don't define this macro, @file{varasm.c} provides a default | |
6655 | definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default | |
6656 | definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if | |
e9a25f70 | 6657 | you want to control one-only symbol support with a compiler flag, or if |
feca2ed3 JW |
6658 | setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to |
6659 | be emitted as one-only. | |
6660 | ||
93638d7a AM |
6661 | @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility}) |
6662 | This target hook is a function to output to @var{asm_out_file} some | |
6663 | commands that will make the symbol(s) associated with @var{decl} have | |
6664 | hidden, protected or internal visibility as specified by @var{visibility}. | |
6665 | @end deftypefn | |
6666 | ||
feca2ed3 JW |
6667 | @findex ASM_OUTPUT_EXTERNAL |
6668 | @item ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name}) | |
6669 | A C statement (sans semicolon) to output to the stdio stream | |
6670 | @var{stream} any text necessary for declaring the name of an external | |
6671 | symbol named @var{name} which is referenced in this compilation but | |
6672 | not defined. The value of @var{decl} is the tree node for the | |
6673 | declaration. | |
6674 | ||
6675 | This macro need not be defined if it does not need to output anything. | |
6676 | The GNU assembler and most Unix assemblers don't require anything. | |
6677 | ||
6678 | @findex ASM_OUTPUT_EXTERNAL_LIBCALL | |
6679 | @item ASM_OUTPUT_EXTERNAL_LIBCALL (@var{stream}, @var{symref}) | |
6680 | A C statement (sans semicolon) to output on @var{stream} an assembler | |
6681 | pseudo-op to declare a library function name external. The name of the | |
6682 | library function is given by @var{symref}, which has type @code{rtx} and | |
6683 | is a @code{symbol_ref}. | |
6684 | ||
6685 | This macro need not be defined if it does not need to output anything. | |
6686 | The GNU assembler and most Unix assemblers don't require anything. | |
6687 | ||
6688 | @findex ASM_OUTPUT_LABELREF | |
6689 | @item ASM_OUTPUT_LABELREF (@var{stream}, @var{name}) | |
6690 | A C statement (sans semicolon) to output to the stdio stream | |
6691 | @var{stream} a reference in assembler syntax to a label named | |
6692 | @var{name}. This should add @samp{_} to the front of the name, if that | |
6693 | is customary on your operating system, as it is in most Berkeley Unix | |
6694 | systems. This macro is used in @code{assemble_name}. | |
6695 | ||
99c8c61c AO |
6696 | @findex ASM_OUTPUT_SYMBOL_REF |
6697 | @item ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym}) | |
6698 | A C statement (sans semicolon) to output a reference to | |
2f0b7af6 | 6699 | @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name} |
99c8c61c AO |
6700 | will be used to output the name of the symbol. This macro may be used |
6701 | to modify the way a symbol is referenced depending on information | |
fb49053f | 6702 | encoded by @code{TARGET_ENCODE_SECTION_INFO}. |
99c8c61c | 6703 | |
2f0b7af6 GK |
6704 | @findex ASM_OUTPUT_LABEL_REF |
6705 | @item ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf}) | |
6706 | A C statement (sans semicolon) to output a reference to @var{buf}, the | |
4226378a | 6707 | result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined, |
2f0b7af6 GK |
6708 | @code{assemble_name} will be used to output the name of the symbol. |
6709 | This macro is not used by @code{output_asm_label}, or the @code{%l} | |
6710 | specifier that calls it; the intention is that this macro should be set | |
4226378a PK |
6711 | when it is necessary to output a label differently when its address is |
6712 | being taken. | |
2f0b7af6 | 6713 | |
feca2ed3 JW |
6714 | @findex ASM_OUTPUT_INTERNAL_LABEL |
6715 | @item ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{prefix}, @var{num}) | |
6716 | A C statement to output to the stdio stream @var{stream} a label whose | |
6717 | name is made from the string @var{prefix} and the number @var{num}. | |
6718 | ||
6719 | It is absolutely essential that these labels be distinct from the labels | |
6720 | used for user-level functions and variables. Otherwise, certain programs | |
6721 | will have name conflicts with internal labels. | |
6722 | ||
6723 | It is desirable to exclude internal labels from the symbol table of the | |
6724 | object file. Most assemblers have a naming convention for labels that | |
6725 | should be excluded; on many systems, the letter @samp{L} at the | |
6726 | beginning of a label has this effect. You should find out what | |
6727 | convention your system uses, and follow it. | |
6728 | ||
6729 | The usual definition of this macro is as follows: | |
6730 | ||
6731 | @example | |
6732 | fprintf (@var{stream}, "L%s%d:\n", @var{prefix}, @var{num}) | |
6733 | @end example | |
6734 | ||
8215347e JW |
6735 | @findex ASM_OUTPUT_DEBUG_LABEL |
6736 | @item ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num}) | |
6737 | A C statement to output to the stdio stream @var{stream} a debug info | |
6738 | label whose name is made from the string @var{prefix} and the number | |
6739 | @var{num}. This is useful for VLIW targets, where debug info labels | |
6740 | may need to be treated differently than branch target labels. On some | |
6741 | systems, branch target labels must be at the beginning of instruction | |
6742 | bundles, but debug info labels can occur in the middle of instruction | |
6743 | bundles. | |
6744 | ||
6745 | If this macro is not defined, then @code{ASM_OUTPUT_INTERNAL_LABEL} will be | |
6746 | used. | |
6747 | ||
feca2ed3 JW |
6748 | @findex ASM_GENERATE_INTERNAL_LABEL |
6749 | @item ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num}) | |
6750 | A C statement to store into the string @var{string} a label whose name | |
6751 | is made from the string @var{prefix} and the number @var{num}. | |
6752 | ||
6753 | This string, when output subsequently by @code{assemble_name}, should | |
6754 | produce the output that @code{ASM_OUTPUT_INTERNAL_LABEL} would produce | |
6755 | with the same @var{prefix} and @var{num}. | |
6756 | ||
6757 | If the string begins with @samp{*}, then @code{assemble_name} will | |
6758 | output the rest of the string unchanged. It is often convenient for | |
6759 | @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the | |
6760 | string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets | |
6761 | to output the string, and may change it. (Of course, | |
6762 | @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so | |
6763 | you should know what it does on your machine.) | |
6764 | ||
6765 | @findex ASM_FORMAT_PRIVATE_NAME | |
6766 | @item ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number}) | |
6767 | A C expression to assign to @var{outvar} (which is a variable of type | |
6768 | @code{char *}) a newly allocated string made from the string | |
6769 | @var{name} and the number @var{number}, with some suitable punctuation | |
6770 | added. Use @code{alloca} to get space for the string. | |
6771 | ||
6772 | The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to | |
6773 | produce an assembler label for an internal static variable whose name is | |
6774 | @var{name}. Therefore, the string must be such as to result in valid | |
6775 | assembler code. The argument @var{number} is different each time this | |
6776 | macro is executed; it prevents conflicts between similarly-named | |
6777 | internal static variables in different scopes. | |
6778 | ||
6779 | Ideally this string should not be a valid C identifier, to prevent any | |
6780 | conflict with the user's own symbols. Most assemblers allow periods | |
6781 | or percent signs in assembler symbols; putting at least one of these | |
6782 | between the name and the number will suffice. | |
6783 | ||
6784 | @findex ASM_OUTPUT_DEF | |
6785 | @item ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value}) | |
6786 | A C statement to output to the stdio stream @var{stream} assembler code | |
6787 | which defines (equates) the symbol @var{name} to have the value @var{value}. | |
6788 | ||
203cb4ef | 6789 | @findex SET_ASM_OP |
aee96fe9 | 6790 | If @code{SET_ASM_OP} is defined, a default definition is provided which is |
feca2ed3 | 6791 | correct for most systems. |
810e3c45 | 6792 | |
e4faf1eb | 6793 | @findex ASM_OUTPUT_DEF_FROM_DECLS |
8760eaae | 6794 | @item ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value}) |
e4faf1eb | 6795 | A C statement to output to the stdio stream @var{stream} assembler code |
3b7a2e58 | 6796 | which defines (equates) the symbol whose tree node is @var{decl_of_name} |
e4faf1eb NC |
6797 | to have the value of the tree node @var{decl_of_value}. This macro will |
6798 | be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if | |
6799 | the tree nodes are available. | |
6800 | ||
203cb4ef | 6801 | @findex SET_ASM_OP |
aee96fe9 | 6802 | If @code{SET_ASM_OP} is defined, a default definition is provided which is |
956d6950 JL |
6803 | correct for most systems. |
6804 | ||
810e3c45 JM |
6805 | @findex ASM_OUTPUT_WEAK_ALIAS |
6806 | @item ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value}) | |
6807 | A C statement to output to the stdio stream @var{stream} assembler code | |
6808 | which defines (equates) the weak symbol @var{name} to have the value | |
3aa8ab7b L |
6809 | @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as |
6810 | an undefined weak symbol. | |
810e3c45 JM |
6811 | |
6812 | Define this macro if the target only supports weak aliases; define | |
aee96fe9 | 6813 | @code{ASM_OUTPUT_DEF} instead if possible. |
810e3c45 | 6814 | |
feca2ed3 JW |
6815 | @findex OBJC_GEN_METHOD_LABEL |
6816 | @item OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name}) | |
6817 | Define this macro to override the default assembler names used for | |
2147b154 | 6818 | Objective-C methods. |
feca2ed3 JW |
6819 | |
6820 | The default name is a unique method number followed by the name of the | |
6821 | class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of | |
6822 | the category is also included in the assembler name (e.g.@: | |
6823 | @samp{_1_Foo_Bar}). | |
6824 | ||
6825 | These names are safe on most systems, but make debugging difficult since | |
6826 | the method's selector is not present in the name. Therefore, particular | |
6827 | systems define other ways of computing names. | |
6828 | ||
6829 | @var{buf} is an expression of type @code{char *} which gives you a | |
6830 | buffer in which to store the name; its length is as long as | |
6831 | @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus | |
6832 | 50 characters extra. | |
6833 | ||
6834 | The argument @var{is_inst} specifies whether the method is an instance | |
6835 | method or a class method; @var{class_name} is the name of the class; | |
59d42021 | 6836 | @var{cat_name} is the name of the category (or @code{NULL} if the method is not |
feca2ed3 JW |
6837 | in a category); and @var{sel_name} is the name of the selector. |
6838 | ||
6839 | On systems where the assembler can handle quoted names, you can use this | |
6840 | macro to provide more human-readable names. | |
28df0b5a | 6841 | |
f60b945b SS |
6842 | @findex ASM_DECLARE_CLASS_REFERENCE |
6843 | @item ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name}) | |
6844 | A C statement (sans semicolon) to output to the stdio stream | |
6845 | @var{stream} commands to declare that the label @var{name} is an | |
6846 | Objective-C class reference. This is only needed for targets whose | |
6847 | linkers have special support for NeXT-style runtimes. | |
6848 | ||
28df0b5a SS |
6849 | @findex ASM_DECLARE_UNRESOLVED_REFERENCE |
6850 | @item ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name}) | |
6851 | A C statement (sans semicolon) to output to the stdio stream | |
6852 | @var{stream} commands to declare that the label @var{name} is an | |
6853 | unresolved Objective-C class reference. This is only needed for targets | |
6854 | whose linkers have special support for NeXT-style runtimes. | |
feca2ed3 JW |
6855 | @end table |
6856 | ||
6857 | @node Initialization | |
6858 | @subsection How Initialization Functions Are Handled | |
6859 | @cindex initialization routines | |
6860 | @cindex termination routines | |
6861 | @cindex constructors, output of | |
6862 | @cindex destructors, output of | |
6863 | ||
6864 | The compiled code for certain languages includes @dfn{constructors} | |
6865 | (also called @dfn{initialization routines})---functions to initialize | |
6866 | data in the program when the program is started. These functions need | |
6867 | to be called before the program is ``started''---that is to say, before | |
6868 | @code{main} is called. | |
6869 | ||
6870 | Compiling some languages generates @dfn{destructors} (also called | |
6871 | @dfn{termination routines}) that should be called when the program | |
6872 | terminates. | |
6873 | ||
6874 | To make the initialization and termination functions work, the compiler | |
6875 | must output something in the assembler code to cause those functions to | |
6876 | be called at the appropriate time. When you port the compiler to a new | |
6877 | system, you need to specify how to do this. | |
6878 | ||
6879 | There are two major ways that GCC currently supports the execution of | |
6880 | initialization and termination functions. Each way has two variants. | |
6881 | Much of the structure is common to all four variations. | |
6882 | ||
6883 | @findex __CTOR_LIST__ | |
6884 | @findex __DTOR_LIST__ | |
6885 | The linker must build two lists of these functions---a list of | |
6886 | initialization functions, called @code{__CTOR_LIST__}, and a list of | |
6887 | termination functions, called @code{__DTOR_LIST__}. | |
6888 | ||
6889 | Each list always begins with an ignored function pointer (which may hold | |
6890 | 0, @minus{}1, or a count of the function pointers after it, depending on | |
6891 | the environment). This is followed by a series of zero or more function | |
6892 | pointers to constructors (or destructors), followed by a function | |
6893 | pointer containing zero. | |
6894 | ||
6895 | Depending on the operating system and its executable file format, either | |
6896 | @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup | |
6897 | time and exit time. Constructors are called in reverse order of the | |
6898 | list; destructors in forward order. | |
6899 | ||
6900 | The best way to handle static constructors works only for object file | |
6901 | formats which provide arbitrarily-named sections. A section is set | |
6902 | aside for a list of constructors, and another for a list of destructors. | |
6903 | Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each | |
6904 | object file that defines an initialization function also puts a word in | |
6905 | the constructor section to point to that function. The linker | |
6906 | accumulates all these words into one contiguous @samp{.ctors} section. | |
6907 | Termination functions are handled similarly. | |
6908 | ||
2cc07db4 RH |
6909 | This method will be chosen as the default by @file{target-def.h} if |
6910 | @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not | |
f282ffb3 | 6911 | support arbitrary sections, but does support special designated |
2cc07db4 RH |
6912 | constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP} |
6913 | and @code{DTORS_SECTION_ASM_OP} to achieve the same effect. | |
feca2ed3 JW |
6914 | |
6915 | When arbitrary sections are available, there are two variants, depending | |
6916 | upon how the code in @file{crtstuff.c} is called. On systems that | |
2cc07db4 | 6917 | support a @dfn{.init} section which is executed at program startup, |
feca2ed3 JW |
6918 | parts of @file{crtstuff.c} are compiled into that section. The |
6919 | program is linked by the @code{gcc} driver like this: | |
6920 | ||
6921 | @example | |
2cc07db4 | 6922 | ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o |
feca2ed3 JW |
6923 | @end example |
6924 | ||
2cc07db4 RH |
6925 | The prologue of a function (@code{__init}) appears in the @code{.init} |
6926 | section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise | |
6927 | for the function @code{__fini} in the @dfn{.fini} section. Normally these | |
6928 | files are provided by the operating system or by the GNU C library, but | |
6929 | are provided by GCC for a few targets. | |
6930 | ||
6931 | The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets) | |
6932 | compiled from @file{crtstuff.c}. They contain, among other things, code | |
6933 | fragments within the @code{.init} and @code{.fini} sections that branch | |
6934 | to routines in the @code{.text} section. The linker will pull all parts | |
6935 | of a section together, which results in a complete @code{__init} function | |
6936 | that invokes the routines we need at startup. | |
feca2ed3 JW |
6937 | |
6938 | To use this variant, you must define the @code{INIT_SECTION_ASM_OP} | |
6939 | macro properly. | |
6940 | ||
2cc07db4 RH |
6941 | If no init section is available, when GCC compiles any function called |
6942 | @code{main} (or more accurately, any function designated as a program | |
6943 | entry point by the language front end calling @code{expand_main_function}), | |
6944 | it inserts a procedure call to @code{__main} as the first executable code | |
6945 | after the function prologue. The @code{__main} function is defined | |
6946 | in @file{libgcc2.c} and runs the global constructors. | |
feca2ed3 JW |
6947 | |
6948 | In file formats that don't support arbitrary sections, there are again | |
6949 | two variants. In the simplest variant, the GNU linker (GNU @code{ld}) | |
6950 | and an `a.out' format must be used. In this case, | |
2cc07db4 | 6951 | @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs} |
feca2ed3 JW |
6952 | entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__}, |
6953 | and with the address of the void function containing the initialization | |
6954 | code as its value. The GNU linker recognizes this as a request to add | |
2cc07db4 | 6955 | the value to a @dfn{set}; the values are accumulated, and are eventually |
feca2ed3 JW |
6956 | placed in the executable as a vector in the format described above, with |
6957 | a leading (ignored) count and a trailing zero element. | |
2cc07db4 | 6958 | @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init |
feca2ed3 JW |
6959 | section is available, the absence of @code{INIT_SECTION_ASM_OP} causes |
6960 | the compilation of @code{main} to call @code{__main} as above, starting | |
6961 | the initialization process. | |
6962 | ||
6963 | The last variant uses neither arbitrary sections nor the GNU linker. | |
6964 | This is preferable when you want to do dynamic linking and when using | |
161d7b59 | 6965 | file formats which the GNU linker does not support, such as `ECOFF'@. In |
2cc07db4 RH |
6966 | this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and |
6967 | termination functions are recognized simply by their names. This requires | |
6968 | an extra program in the linkage step, called @command{collect2}. This program | |
6969 | pretends to be the linker, for use with GCC; it does its job by running | |
6970 | the ordinary linker, but also arranges to include the vectors of | |
6971 | initialization and termination functions. These functions are called | |
6972 | via @code{__main} as described above. In order to use this method, | |
6973 | @code{use_collect2} must be defined in the target in @file{config.gcc}. | |
feca2ed3 JW |
6974 | |
6975 | @ifinfo | |
6976 | The following section describes the specific macros that control and | |
6977 | customize the handling of initialization and termination functions. | |
6978 | @end ifinfo | |
6979 | ||
6980 | @node Macros for Initialization | |
6981 | @subsection Macros Controlling Initialization Routines | |
6982 | ||
6983 | Here are the macros that control how the compiler handles initialization | |
6984 | and termination functions: | |
6985 | ||
6986 | @table @code | |
6987 | @findex INIT_SECTION_ASM_OP | |
6988 | @item INIT_SECTION_ASM_OP | |
047c1c92 HPN |
6989 | If defined, a C string constant, including spacing, for the assembler |
6990 | operation to identify the following data as initialization code. If not | |
6991 | defined, GCC will assume such a section does not exist. When you are | |
6992 | using special sections for initialization and termination functions, this | |
6993 | macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to | |
6994 | run the initialization functions. | |
feca2ed3 JW |
6995 | |
6996 | @item HAS_INIT_SECTION | |
6997 | @findex HAS_INIT_SECTION | |
6998 | If defined, @code{main} will not call @code{__main} as described above. | |
2cc07db4 RH |
6999 | This macro should be defined for systems that control start-up code |
7000 | on a symbol-by-symbol basis, such as OSF/1, and should not | |
7001 | be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}. | |
feca2ed3 JW |
7002 | |
7003 | @item LD_INIT_SWITCH | |
7004 | @findex LD_INIT_SWITCH | |
7005 | If defined, a C string constant for a switch that tells the linker that | |
7006 | the following symbol is an initialization routine. | |
7007 | ||
7008 | @item LD_FINI_SWITCH | |
7009 | @findex LD_FINI_SWITCH | |
7010 | If defined, a C string constant for a switch that tells the linker that | |
7011 | the following symbol is a finalization routine. | |
7012 | ||
414e05cf RE |
7013 | @item COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func}) |
7014 | If defined, a C statement that will write a function that can be | |
7015 | automatically called when a shared library is loaded. The function | |
7016 | should call @var{func}, which takes no arguments. If not defined, and | |
7017 | the object format requires an explicit initialization function, then a | |
172270b3 | 7018 | function called @code{_GLOBAL__DI} will be generated. |
414e05cf RE |
7019 | |
7020 | This function and the following one are used by collect2 when linking a | |
f282ffb3 | 7021 | shared library that needs constructors or destructors, or has DWARF2 |
414e05cf RE |
7022 | exception tables embedded in the code. |
7023 | ||
7024 | @item COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func}) | |
7025 | If defined, a C statement that will write a function that can be | |
7026 | automatically called when a shared library is unloaded. The function | |
7027 | should call @var{func}, which takes no arguments. If not defined, and | |
7028 | the object format requires an explicit finalization function, then a | |
172270b3 | 7029 | function called @code{_GLOBAL__DD} will be generated. |
414e05cf | 7030 | |
feca2ed3 JW |
7031 | @item INVOKE__main |
7032 | @findex INVOKE__main | |
7033 | If defined, @code{main} will call @code{__main} despite the presence of | |
7034 | @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems | |
7035 | where the init section is not actually run automatically, but is still | |
7036 | useful for collecting the lists of constructors and destructors. | |
7037 | ||
ea4f1fce JO |
7038 | @item SUPPORTS_INIT_PRIORITY |
7039 | @findex SUPPORTS_INIT_PRIORITY | |
7040 | If nonzero, the C++ @code{init_priority} attribute is supported and the | |
7041 | compiler should emit instructions to control the order of initialization | |
7042 | of objects. If zero, the compiler will issue an error message upon | |
7043 | encountering an @code{init_priority} attribute. | |
2cc07db4 RH |
7044 | @end table |
7045 | ||
7046 | @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS | |
7047 | This value is true if the target supports some ``native'' method of | |
7048 | collecting constructors and destructors to be run at startup and exit. | |
7049 | It is false if we must use @command{collect2}. | |
7050 | @end deftypefn | |
7051 | ||
7052 | @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority}) | |
7053 | If defined, a function that outputs assembler code to arrange to call | |
7054 | the function referenced by @var{symbol} at initialization time. | |
ea4f1fce | 7055 | |
2cc07db4 RH |
7056 | Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking |
7057 | no arguments and with no return value. If the target supports initialization | |
7058 | priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY}; | |
7059 | otherwise it must be @code{DEFAULT_INIT_PRIORITY}. | |
7060 | ||
14976c58 | 7061 | If this macro is not defined by the target, a suitable default will |
2cc07db4 RH |
7062 | be chosen if (1) the target supports arbitrary section names, (2) the |
7063 | target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2} | |
7064 | is not defined. | |
7065 | @end deftypefn | |
7066 | ||
7067 | @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority}) | |
7068 | This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination | |
feca2ed3 | 7069 | functions rather than initialization functions. |
2cc07db4 | 7070 | @end deftypefn |
14686fcd | 7071 | |
2cc07db4 RH |
7072 | If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine |
7073 | generated for the generated object file will have static linkage. | |
feca2ed3 | 7074 | |
2cc07db4 RH |
7075 | If your system uses @command{collect2} as the means of processing |
7076 | constructors, then that program normally uses @command{nm} to scan | |
7077 | an object file for constructor functions to be called. | |
14686fcd JL |
7078 | |
7079 | On certain kinds of systems, you can define these macros to make | |
2cc07db4 | 7080 | @command{collect2} work faster (and, in some cases, make it work at all): |
feca2ed3 JW |
7081 | |
7082 | @table @code | |
7083 | @findex OBJECT_FORMAT_COFF | |
7084 | @item OBJECT_FORMAT_COFF | |
7085 | Define this macro if the system uses COFF (Common Object File Format) | |
2cc07db4 | 7086 | object files, so that @command{collect2} can assume this format and scan |
feca2ed3 JW |
7087 | object files directly for dynamic constructor/destructor functions. |
7088 | ||
7089 | @findex OBJECT_FORMAT_ROSE | |
7090 | @item OBJECT_FORMAT_ROSE | |
7091 | Define this macro if the system uses ROSE format object files, so that | |
2cc07db4 | 7092 | @command{collect2} can assume this format and scan object files directly |
feca2ed3 JW |
7093 | for dynamic constructor/destructor functions. |
7094 | ||
2cc07db4 RH |
7095 | These macros are effective only in a native compiler; @command{collect2} as |
7096 | part of a cross compiler always uses @command{nm} for the target machine. | |
feca2ed3 JW |
7097 | |
7098 | @findex REAL_NM_FILE_NAME | |
7099 | @item REAL_NM_FILE_NAME | |
7100 | Define this macro as a C string constant containing the file name to use | |
2cc07db4 RH |
7101 | to execute @command{nm}. The default is to search the path normally for |
7102 | @command{nm}. | |
feca2ed3 JW |
7103 | |
7104 | If your system supports shared libraries and has a program to list the | |
7105 | dynamic dependencies of a given library or executable, you can define | |
7106 | these macros to enable support for running initialization and | |
7107 | termination functions in shared libraries: | |
7108 | ||
7109 | @findex LDD_SUFFIX | |
7110 | @item LDD_SUFFIX | |
2cc07db4 RH |
7111 | Define this macro to a C string constant containing the name of the program |
7112 | which lists dynamic dependencies, like @command{"ldd"} under SunOS 4. | |
feca2ed3 JW |
7113 | |
7114 | @findex PARSE_LDD_OUTPUT | |
aee96fe9 | 7115 | @item PARSE_LDD_OUTPUT (@var{ptr}) |
feca2ed3 | 7116 | Define this macro to be C code that extracts filenames from the output |
aee96fe9 | 7117 | of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable |
feca2ed3 JW |
7118 | of type @code{char *} that points to the beginning of a line of output |
7119 | from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the | |
aee96fe9 JM |
7120 | code must advance @var{ptr} to the beginning of the filename on that |
7121 | line. Otherwise, it must set @var{ptr} to @code{NULL}. | |
feca2ed3 JW |
7122 | @end table |
7123 | ||
7124 | @node Instruction Output | |
7125 | @subsection Output of Assembler Instructions | |
7126 | ||
7127 | @c prevent bad page break with this line | |
7128 | This describes assembler instruction output. | |
7129 | ||
7130 | @table @code | |
7131 | @findex REGISTER_NAMES | |
7132 | @item REGISTER_NAMES | |
7133 | A C initializer containing the assembler's names for the machine | |
7134 | registers, each one as a C string constant. This is what translates | |
7135 | register numbers in the compiler into assembler language. | |
7136 | ||
7137 | @findex ADDITIONAL_REGISTER_NAMES | |
7138 | @item ADDITIONAL_REGISTER_NAMES | |
7139 | If defined, a C initializer for an array of structures containing a name | |
7140 | and a register number. This macro defines additional names for hard | |
7141 | registers, thus allowing the @code{asm} option in declarations to refer | |
7142 | to registers using alternate names. | |
7143 | ||
7144 | @findex ASM_OUTPUT_OPCODE | |
7145 | @item ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr}) | |
7146 | Define this macro if you are using an unusual assembler that | |
7147 | requires different names for the machine instructions. | |
7148 | ||
7149 | The definition is a C statement or statements which output an | |
7150 | assembler instruction opcode to the stdio stream @var{stream}. The | |
7151 | macro-operand @var{ptr} is a variable of type @code{char *} which | |
7152 | points to the opcode name in its ``internal'' form---the form that is | |
7153 | written in the machine description. The definition should output the | |
7154 | opcode name to @var{stream}, performing any translation you desire, and | |
7155 | increment the variable @var{ptr} to point at the end of the opcode | |
7156 | so that it will not be output twice. | |
7157 | ||
7158 | In fact, your macro definition may process less than the entire opcode | |
7159 | name, or more than the opcode name; but if you want to process text | |
7160 | that includes @samp{%}-sequences to substitute operands, you must take | |
7161 | care of the substitution yourself. Just be sure to increment | |
7162 | @var{ptr} over whatever text should not be output normally. | |
7163 | ||
37bef197 | 7164 | @findex recog_data.operand |
feca2ed3 | 7165 | If you need to look at the operand values, they can be found as the |
37bef197 | 7166 | elements of @code{recog_data.operand}. |
feca2ed3 JW |
7167 | |
7168 | If the macro definition does nothing, the instruction is output | |
7169 | in the usual way. | |
7170 | ||
7171 | @findex FINAL_PRESCAN_INSN | |
7172 | @item FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands}) | |
7173 | If defined, a C statement to be executed just prior to the output of | |
7174 | assembler code for @var{insn}, to modify the extracted operands so | |
7175 | they will be output differently. | |
7176 | ||
7177 | Here the argument @var{opvec} is the vector containing the operands | |
7178 | extracted from @var{insn}, and @var{noperands} is the number of | |
7179 | elements of the vector which contain meaningful data for this insn. | |
7180 | The contents of this vector are what will be used to convert the insn | |
7181 | template into assembler code, so you can change the assembler output | |
7182 | by changing the contents of the vector. | |
7183 | ||
7184 | This macro is useful when various assembler syntaxes share a single | |
7185 | file of instruction patterns; by defining this macro differently, you | |
7186 | can cause a large class of instructions to be output differently (such | |
7187 | as with rearranged operands). Naturally, variations in assembler | |
7188 | syntax affecting individual insn patterns ought to be handled by | |
7189 | writing conditional output routines in those patterns. | |
7190 | ||
7191 | If this macro is not defined, it is equivalent to a null statement. | |
7192 | ||
7193 | @findex FINAL_PRESCAN_LABEL | |
7194 | @item FINAL_PRESCAN_LABEL | |
7195 | If defined, @code{FINAL_PRESCAN_INSN} will be called on each | |
7196 | @code{CODE_LABEL}. In that case, @var{opvec} will be a null pointer and | |
7197 | @var{noperands} will be zero. | |
7198 | ||
7199 | @findex PRINT_OPERAND | |
7200 | @item PRINT_OPERAND (@var{stream}, @var{x}, @var{code}) | |
7201 | A C compound statement to output to stdio stream @var{stream} the | |
7202 | assembler syntax for an instruction operand @var{x}. @var{x} is an | |
7203 | RTL expression. | |
7204 | ||
7205 | @var{code} is a value that can be used to specify one of several ways | |
7206 | of printing the operand. It is used when identical operands must be | |
7207 | printed differently depending on the context. @var{code} comes from | |
7208 | the @samp{%} specification that was used to request printing of the | |
7209 | operand. If the specification was just @samp{%@var{digit}} then | |
7210 | @var{code} is 0; if the specification was @samp{%@var{ltr} | |
7211 | @var{digit}} then @var{code} is the ASCII code for @var{ltr}. | |
7212 | ||
7213 | @findex reg_names | |
7214 | If @var{x} is a register, this macro should print the register's name. | |
7215 | The names can be found in an array @code{reg_names} whose type is | |
7216 | @code{char *[]}. @code{reg_names} is initialized from | |
7217 | @code{REGISTER_NAMES}. | |
7218 | ||
7219 | When the machine description has a specification @samp{%@var{punct}} | |
7220 | (a @samp{%} followed by a punctuation character), this macro is called | |
7221 | with a null pointer for @var{x} and the punctuation character for | |
7222 | @var{code}. | |
7223 | ||
7224 | @findex PRINT_OPERAND_PUNCT_VALID_P | |
7225 | @item PRINT_OPERAND_PUNCT_VALID_P (@var{code}) | |
7226 | A C expression which evaluates to true if @var{code} is a valid | |
7227 | punctuation character for use in the @code{PRINT_OPERAND} macro. If | |
7228 | @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no | |
7229 | punctuation characters (except for the standard one, @samp{%}) are used | |
7230 | in this way. | |
7231 | ||
7232 | @findex PRINT_OPERAND_ADDRESS | |
7233 | @item PRINT_OPERAND_ADDRESS (@var{stream}, @var{x}) | |
7234 | A C compound statement to output to stdio stream @var{stream} the | |
7235 | assembler syntax for an instruction operand that is a memory reference | |
7236 | whose address is @var{x}. @var{x} is an RTL expression. | |
7237 | ||
fb49053f | 7238 | @cindex @code{TARGET_ENCODE_SECTION_INFO} usage |
feca2ed3 | 7239 | On some machines, the syntax for a symbolic address depends on the |
fb49053f RH |
7240 | section that the address refers to. On these machines, define the hook |
7241 | @code{TARGET_ENCODE_SECTION_INFO} to store the information into the | |
feca2ed3 JW |
7242 | @code{symbol_ref}, and then check for it here. @xref{Assembler Format}. |
7243 | ||
7244 | @findex DBR_OUTPUT_SEQEND | |
7245 | @findex dbr_sequence_length | |
7246 | @item DBR_OUTPUT_SEQEND(@var{file}) | |
7247 | A C statement, to be executed after all slot-filler instructions have | |
7248 | been output. If necessary, call @code{dbr_sequence_length} to | |
7249 | determine the number of slots filled in a sequence (zero if not | |
7250 | currently outputting a sequence), to decide how many no-ops to output, | |
7251 | or whatever. | |
7252 | ||
7253 | Don't define this macro if it has nothing to do, but it is helpful in | |
7254 | reading assembly output if the extent of the delay sequence is made | |
e979f9e8 | 7255 | explicit (e.g.@: with white space). |
feca2ed3 JW |
7256 | |
7257 | @findex final_sequence | |
7258 | Note that output routines for instructions with delay slots must be | |
e979f9e8 JM |
7259 | prepared to deal with not being output as part of a sequence |
7260 | (i.e.@: when the scheduling pass is not run, or when no slot fillers could be | |
feca2ed3 JW |
7261 | found.) The variable @code{final_sequence} is null when not |
7262 | processing a sequence, otherwise it contains the @code{sequence} rtx | |
7263 | being output. | |
7264 | ||
7265 | @findex REGISTER_PREFIX | |
7266 | @findex LOCAL_LABEL_PREFIX | |
7267 | @findex USER_LABEL_PREFIX | |
7268 | @findex IMMEDIATE_PREFIX | |
7269 | @findex asm_fprintf | |
7270 | @item REGISTER_PREFIX | |
7271 | @itemx LOCAL_LABEL_PREFIX | |
7272 | @itemx USER_LABEL_PREFIX | |
7273 | @itemx IMMEDIATE_PREFIX | |
7274 | If defined, C string expressions to be used for the @samp{%R}, @samp{%L}, | |
7275 | @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see | |
7276 | @file{final.c}). These are useful when a single @file{md} file must | |
7277 | support multiple assembler formats. In that case, the various @file{tm.h} | |
7278 | files can define these macros differently. | |
7279 | ||
fe0503ea NC |
7280 | @item ASM_FPRINTF_EXTENSIONS(@var{file}, @var{argptr}, @var{format}) |
7281 | @findex ASM_FPRINTF_EXTENSIONS | |
3b7a2e58 | 7282 | If defined this macro should expand to a series of @code{case} |
fe0503ea NC |
7283 | statements which will be parsed inside the @code{switch} statement of |
7284 | the @code{asm_fprintf} function. This allows targets to define extra | |
7285 | printf formats which may useful when generating their assembler | |
3b7a2e58 | 7286 | statements. Note that upper case letters are reserved for future |
fe0503ea NC |
7287 | generic extensions to asm_fprintf, and so are not available to target |
7288 | specific code. The output file is given by the parameter @var{file}. | |
7289 | The varargs input pointer is @var{argptr} and the rest of the format | |
7290 | string, starting the character after the one that is being switched | |
7291 | upon, is pointed to by @var{format}. | |
7292 | ||
feca2ed3 JW |
7293 | @findex ASSEMBLER_DIALECT |
7294 | @item ASSEMBLER_DIALECT | |
7295 | If your target supports multiple dialects of assembler language (such as | |
7296 | different opcodes), define this macro as a C expression that gives the | |
7297 | numeric index of the assembler language dialect to use, with zero as the | |
7298 | first variant. | |
7299 | ||
7300 | If this macro is defined, you may use constructs of the form | |
c237e94a | 7301 | @smallexample |
f282ffb3 | 7302 | @samp{@{option0|option1|option2@dots{}@}} |
c237e94a ZW |
7303 | @end smallexample |
7304 | @noindent | |
7305 | in the output templates of patterns (@pxref{Output Template}) or in the | |
7306 | first argument of @code{asm_fprintf}. This construct outputs | |
7307 | @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of | |
7308 | @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters | |
7309 | within these strings retain their usual meaning. If there are fewer | |
7310 | alternatives within the braces than the value of | |
7311 | @code{ASSEMBLER_DIALECT}, the construct outputs nothing. | |
feca2ed3 JW |
7312 | |
7313 | If you do not define this macro, the characters @samp{@{}, @samp{|} and | |
7314 | @samp{@}} do not have any special meaning when used in templates or | |
7315 | operands to @code{asm_fprintf}. | |
7316 | ||
7317 | Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX}, | |
7318 | @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express | |
e5e809f4 | 7319 | the variations in assembler language syntax with that mechanism. Define |
feca2ed3 JW |
7320 | @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax |
7321 | if the syntax variant are larger and involve such things as different | |
7322 | opcodes or operand order. | |
7323 | ||
7324 | @findex ASM_OUTPUT_REG_PUSH | |
7325 | @item ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno}) | |
7326 | A C expression to output to @var{stream} some assembler code | |
7327 | which will push hard register number @var{regno} onto the stack. | |
7328 | The code need not be optimal, since this macro is used only when | |
7329 | profiling. | |
7330 | ||
7331 | @findex ASM_OUTPUT_REG_POP | |
7332 | @item ASM_OUTPUT_REG_POP (@var{stream}, @var{regno}) | |
7333 | A C expression to output to @var{stream} some assembler code | |
7334 | which will pop hard register number @var{regno} off of the stack. | |
7335 | The code need not be optimal, since this macro is used only when | |
7336 | profiling. | |
7337 | @end table | |
7338 | ||
7339 | @node Dispatch Tables | |
7340 | @subsection Output of Dispatch Tables | |
7341 | ||
7342 | @c prevent bad page break with this line | |
7343 | This concerns dispatch tables. | |
7344 | ||
7345 | @table @code | |
7346 | @cindex dispatch table | |
7347 | @findex ASM_OUTPUT_ADDR_DIFF_ELT | |
33f7f353 | 7348 | @item ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel}) |
feca2ed3 JW |
7349 | A C statement to output to the stdio stream @var{stream} an assembler |
7350 | pseudo-instruction to generate a difference between two labels. | |
7351 | @var{value} and @var{rel} are the numbers of two internal labels. The | |
7352 | definitions of these labels are output using | |
7353 | @code{ASM_OUTPUT_INTERNAL_LABEL}, and they must be printed in the same | |
7354 | way here. For example, | |
7355 | ||
7356 | @example | |
7357 | fprintf (@var{stream}, "\t.word L%d-L%d\n", | |
7358 | @var{value}, @var{rel}) | |
7359 | @end example | |
7360 | ||
7361 | You must provide this macro on machines where the addresses in a | |
f0523f02 | 7362 | dispatch table are relative to the table's own address. If defined, GCC |
161d7b59 | 7363 | will also use this macro on all machines when producing PIC@. |
aee96fe9 | 7364 | @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the |
33f7f353 | 7365 | mode and flags can be read. |
feca2ed3 JW |
7366 | |
7367 | @findex ASM_OUTPUT_ADDR_VEC_ELT | |
7368 | @item ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value}) | |
7369 | This macro should be provided on machines where the addresses | |
7370 | in a dispatch table are absolute. | |
7371 | ||
7372 | The definition should be a C statement to output to the stdio stream | |
7373 | @var{stream} an assembler pseudo-instruction to generate a reference to | |
7374 | a label. @var{value} is the number of an internal label whose | |
7375 | definition is output using @code{ASM_OUTPUT_INTERNAL_LABEL}. | |
7376 | For example, | |
7377 | ||
7378 | @example | |
7379 | fprintf (@var{stream}, "\t.word L%d\n", @var{value}) | |
7380 | @end example | |
7381 | ||
7382 | @findex ASM_OUTPUT_CASE_LABEL | |
7383 | @item ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table}) | |
7384 | Define this if the label before a jump-table needs to be output | |
7385 | specially. The first three arguments are the same as for | |
7386 | @code{ASM_OUTPUT_INTERNAL_LABEL}; the fourth argument is the | |
7387 | jump-table which follows (a @code{jump_insn} containing an | |
7388 | @code{addr_vec} or @code{addr_diff_vec}). | |
7389 | ||
7390 | This feature is used on system V to output a @code{swbeg} statement | |
7391 | for the table. | |
7392 | ||
7393 | If this macro is not defined, these labels are output with | |
7394 | @code{ASM_OUTPUT_INTERNAL_LABEL}. | |
7395 | ||
7396 | @findex ASM_OUTPUT_CASE_END | |
7397 | @item ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table}) | |
7398 | Define this if something special must be output at the end of a | |
7399 | jump-table. The definition should be a C statement to be executed | |
7400 | after the assembler code for the table is written. It should write | |
7401 | the appropriate code to stdio stream @var{stream}. The argument | |
7402 | @var{table} is the jump-table insn, and @var{num} is the label-number | |
7403 | of the preceding label. | |
7404 | ||
7405 | If this macro is not defined, nothing special is output at the end of | |
7406 | the jump-table. | |
7407 | @end table | |
7408 | ||
02f52e19 | 7409 | @node Exception Region Output |
feca2ed3 JW |
7410 | @subsection Assembler Commands for Exception Regions |
7411 | ||
7412 | @c prevent bad page break with this line | |
7413 | ||
7414 | This describes commands marking the start and the end of an exception | |
7415 | region. | |
7416 | ||
7417 | @table @code | |
7c262518 RH |
7418 | @findex EH_FRAME_SECTION_NAME |
7419 | @item EH_FRAME_SECTION_NAME | |
7420 | If defined, a C string constant for the name of the section containing | |
7421 | exception handling frame unwind information. If not defined, GCC will | |
7422 | provide a default definition if the target supports named sections. | |
7423 | @file{crtstuff.c} uses this macro to switch to the appropriate section. | |
0021b564 JM |
7424 | |
7425 | You should define this symbol if your target supports DWARF 2 frame | |
7426 | unwind information and the default definition does not work. | |
7427 | ||
02c9b1ca RH |
7428 | @findex EH_FRAME_IN_DATA_SECTION |
7429 | @item EH_FRAME_IN_DATA_SECTION | |
7430 | If defined, DWARF 2 frame unwind information will be placed in the | |
7431 | data section even though the target supports named sections. This | |
7432 | might be necessary, for instance, if the system linker does garbage | |
7433 | collection and sections cannot be marked as not to be collected. | |
7434 | ||
7435 | Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is | |
7436 | also defined. | |
7437 | ||
feca2ed3 JW |
7438 | @findex MASK_RETURN_ADDR |
7439 | @item MASK_RETURN_ADDR | |
aee96fe9 | 7440 | An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so |
feca2ed3 | 7441 | that it does not contain any extraneous set bits in it. |
0021b564 JM |
7442 | |
7443 | @findex DWARF2_UNWIND_INFO | |
7444 | @item DWARF2_UNWIND_INFO | |
7445 | Define this macro to 0 if your target supports DWARF 2 frame unwind | |
7446 | information, but it does not yet work with exception handling. | |
7447 | Otherwise, if your target supports this information (if it defines | |
7448 | @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP} | |
7449 | or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of | |
7450 | 1. | |
7451 | ||
7452 | If this macro is defined to 1, the DWARF 2 unwinder will be the default | |
aee96fe9 | 7453 | exception handling mechanism; otherwise, @code{setjmp}/@code{longjmp} will be used by |
0021b564 JM |
7454 | default. |
7455 | ||
7456 | If this macro is defined to anything, the DWARF 2 unwinder will be used | |
aee96fe9 | 7457 | instead of inline unwinders and @code{__unwind_function} in the non-@code{setjmp} case. |
0021b564 | 7458 | |
27c35f4b HPN |
7459 | @findex DWARF_CIE_DATA_ALIGNMENT |
7460 | @item DWARF_CIE_DATA_ALIGNMENT | |
7461 | This macro need only be defined if the target might save registers in the | |
7462 | function prologue at an offset to the stack pointer that is not aligned to | |
7463 | @code{UNITS_PER_WORD}. The definition should be the negative minimum | |
7464 | alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive | |
7465 | minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if | |
7466 | the target supports DWARF 2 frame unwind information. | |
7467 | ||
feca2ed3 JW |
7468 | @end table |
7469 | ||
07c9d2eb SS |
7470 | @deftypefn {Target Hook} void TARGET_ASM_EXCEPTION_SECTION () |
7471 | If defined, a function that switches to the section in which the main | |
7472 | exception table is to be placed (@pxref{Sections}). The default is a | |
7473 | function that switches to a section named @code{.gcc_except_table} on | |
7474 | machines that support named sections via | |
7475 | @code{TARGET_ASM_NAMED_SECTION}, otherwise if @option{-fpic} or | |
7476 | @option{-fPIC} is in effect, the @code{data_section}, otherwise the | |
7477 | @code{readonly_data_section}. | |
7478 | @end deftypefn | |
7479 | ||
7480 | @deftypefn {Target Hook} void TARGET_ASM_EH_FRAME_SECTION () | |
7481 | If defined, a function that switches to the section in which the DWARF 2 | |
7482 | frame unwind information to be placed (@pxref{Sections}). The default | |
7483 | is a function that outputs a standard GAS section directive, if | |
7484 | @code{EH_FRAME_SECTION_NAME} is defined, or else a data section | |
7485 | directive followed by a synthetic label. | |
7486 | @end deftypefn | |
7487 | ||
7606e68f SS |
7488 | @deftypevar {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO |
7489 | Contains the value true if the target should add a zero word onto the | |
7490 | end of a Dwarf-2 frame info section when used for exception handling. | |
7491 | Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and | |
7492 | true otherwise. | |
7493 | @end deftypevar | |
7494 | ||
feca2ed3 JW |
7495 | @node Alignment Output |
7496 | @subsection Assembler Commands for Alignment | |
7497 | ||
7498 | @c prevent bad page break with this line | |
7499 | This describes commands for alignment. | |
7500 | ||
7501 | @table @code | |
247a370b JH |
7502 | @findex JUMP_ALIGN |
7503 | @item JUMP_ALIGN (@var{label}) | |
7504 | The alignment (log base 2) to put in front of @var{label}, which is | |
f710504c | 7505 | a common destination of jumps and has no fallthru incoming edge. |
25e22dc0 JH |
7506 | |
7507 | This macro need not be defined if you don't want any special alignment | |
7508 | to be done at such a time. Most machine descriptions do not currently | |
7509 | define the macro. | |
efa3896a | 7510 | |
3446405d JH |
7511 | Unless it's necessary to inspect the @var{label} parameter, it is better |
7512 | to set the variable @var{align_jumps} in the target's | |
c21cd8b1 | 7513 | @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's |
247a370b JH |
7514 | selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation. |
7515 | ||
7516 | @findex LABEL_ALIGN_AFTER_BARRIER | |
7517 | @item LABEL_ALIGN_AFTER_BARRIER (@var{label}) | |
7518 | The alignment (log base 2) to put in front of @var{label}, which follows | |
7519 | a @code{BARRIER}. | |
7520 | ||
7521 | This macro need not be defined if you don't want any special alignment | |
7522 | to be done at such a time. Most machine descriptions do not currently | |
7523 | define the macro. | |
3446405d | 7524 | |
efa3896a GK |
7525 | @findex LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP |
7526 | @item LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP | |
02f52e19 | 7527 | The maximum number of bytes to skip when applying |
efa3896a GK |
7528 | @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if |
7529 | @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. | |
7530 | ||
fc470718 R |
7531 | @findex LOOP_ALIGN |
7532 | @item LOOP_ALIGN (@var{label}) | |
7533 | The alignment (log base 2) to put in front of @var{label}, which follows | |
aee96fe9 | 7534 | a @code{NOTE_INSN_LOOP_BEG} note. |
feca2ed3 JW |
7535 | |
7536 | This macro need not be defined if you don't want any special alignment | |
7537 | to be done at such a time. Most machine descriptions do not currently | |
7538 | define the macro. | |
7539 | ||
efa3896a | 7540 | Unless it's necessary to inspect the @var{label} parameter, it is better |
aee96fe9 | 7541 | to set the variable @code{align_loops} in the target's |
c21cd8b1 | 7542 | @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's |
aee96fe9 | 7543 | selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation. |
efa3896a GK |
7544 | |
7545 | @findex LOOP_ALIGN_MAX_SKIP | |
7546 | @item LOOP_ALIGN_MAX_SKIP | |
7547 | The maximum number of bytes to skip when applying @code{LOOP_ALIGN}. | |
7548 | This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. | |
7549 | ||
fc470718 R |
7550 | @findex LABEL_ALIGN |
7551 | @item LABEL_ALIGN (@var{label}) | |
7552 | The alignment (log base 2) to put in front of @var{label}. | |
aee96fe9 | 7553 | If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment, |
fc470718 R |
7554 | the maximum of the specified values is used. |
7555 | ||
efa3896a | 7556 | Unless it's necessary to inspect the @var{label} parameter, it is better |
aee96fe9 | 7557 | to set the variable @code{align_labels} in the target's |
c21cd8b1 | 7558 | @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's |
aee96fe9 | 7559 | selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation. |
efa3896a GK |
7560 | |
7561 | @findex LABEL_ALIGN_MAX_SKIP | |
7562 | @item LABEL_ALIGN_MAX_SKIP | |
7563 | The maximum number of bytes to skip when applying @code{LABEL_ALIGN}. | |
7564 | This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined. | |
7565 | ||
feca2ed3 JW |
7566 | @findex ASM_OUTPUT_SKIP |
7567 | @item ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes}) | |
7568 | A C statement to output to the stdio stream @var{stream} an assembler | |
7569 | instruction to advance the location counter by @var{nbytes} bytes. | |
7570 | Those bytes should be zero when loaded. @var{nbytes} will be a C | |
7571 | expression of type @code{int}. | |
7572 | ||
7573 | @findex ASM_NO_SKIP_IN_TEXT | |
7574 | @item ASM_NO_SKIP_IN_TEXT | |
7575 | Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the | |
556e0f21 | 7576 | text section because it fails to put zeros in the bytes that are skipped. |
feca2ed3 JW |
7577 | This is true on many Unix systems, where the pseudo--op to skip bytes |
7578 | produces no-op instructions rather than zeros when used in the text | |
7579 | section. | |
7580 | ||
7581 | @findex ASM_OUTPUT_ALIGN | |
7582 | @item ASM_OUTPUT_ALIGN (@var{stream}, @var{power}) | |
7583 | A C statement to output to the stdio stream @var{stream} an assembler | |
7584 | command to advance the location counter to a multiple of 2 to the | |
7585 | @var{power} bytes. @var{power} will be a C expression of type @code{int}. | |
26f63a77 | 7586 | |
8e16ab99 SF |
7587 | @findex ASM_OUTPUT_ALIGN_WITH_NOP |
7588 | @item ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power}) | |
7589 | Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used | |
7590 | for padding, if necessary. | |
7591 | ||
26f63a77 JL |
7592 | @findex ASM_OUTPUT_MAX_SKIP_ALIGN |
7593 | @item ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip}) | |
7594 | A C statement to output to the stdio stream @var{stream} an assembler | |
7595 | command to advance the location counter to a multiple of 2 to the | |
7596 | @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to | |
7597 | satisfy the alignment request. @var{power} and @var{max_skip} will be | |
7598 | a C expression of type @code{int}. | |
feca2ed3 JW |
7599 | @end table |
7600 | ||
7601 | @need 3000 | |
7602 | @node Debugging Info | |
7603 | @section Controlling Debugging Information Format | |
7604 | ||
7605 | @c prevent bad page break with this line | |
7606 | This describes how to specify debugging information. | |
7607 | ||
7608 | @menu | |
7609 | * All Debuggers:: Macros that affect all debugging formats uniformly. | |
7610 | * DBX Options:: Macros enabling specific options in DBX format. | |
7611 | * DBX Hooks:: Hook macros for varying DBX format. | |
7612 | * File Names and DBX:: Macros controlling output of file names in DBX format. | |
7613 | * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats. | |
5f98259a | 7614 | * VMS Debug:: Macros for VMS debug format. |
feca2ed3 JW |
7615 | @end menu |
7616 | ||
7617 | @node All Debuggers | |
7618 | @subsection Macros Affecting All Debugging Formats | |
7619 | ||
7620 | @c prevent bad page break with this line | |
7621 | These macros affect all debugging formats. | |
7622 | ||
7623 | @table @code | |
7624 | @findex DBX_REGISTER_NUMBER | |
7625 | @item DBX_REGISTER_NUMBER (@var{regno}) | |
7626 | A C expression that returns the DBX register number for the compiler | |
4617e3b5 KG |
7627 | register number @var{regno}. In the default macro provided, the value |
7628 | of this expression will be @var{regno} itself. But sometimes there are | |
7629 | some registers that the compiler knows about and DBX does not, or vice | |
7630 | versa. In such cases, some register may need to have one number in the | |
7631 | compiler and another for DBX@. | |
feca2ed3 | 7632 | |
a3a15b4d | 7633 | If two registers have consecutive numbers inside GCC, and they can be |
feca2ed3 JW |
7634 | used as a pair to hold a multiword value, then they @emph{must} have |
7635 | consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}. | |
7636 | Otherwise, debuggers will be unable to access such a pair, because they | |
7637 | expect register pairs to be consecutive in their own numbering scheme. | |
7638 | ||
7639 | If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that | |
7640 | does not preserve register pairs, then what you must do instead is | |
7641 | redefine the actual register numbering scheme. | |
7642 | ||
7643 | @findex DEBUGGER_AUTO_OFFSET | |
7644 | @item DEBUGGER_AUTO_OFFSET (@var{x}) | |
7645 | A C expression that returns the integer offset value for an automatic | |
7646 | variable having address @var{x} (an RTL expression). The default | |
7647 | computation assumes that @var{x} is based on the frame-pointer and | |
7648 | gives the offset from the frame-pointer. This is required for targets | |
7649 | that produce debugging output for DBX or COFF-style debugging output | |
7650 | for SDB and allow the frame-pointer to be eliminated when the | |
630d3d5a | 7651 | @option{-g} options is used. |
feca2ed3 JW |
7652 | |
7653 | @findex DEBUGGER_ARG_OFFSET | |
7654 | @item DEBUGGER_ARG_OFFSET (@var{offset}, @var{x}) | |
7655 | A C expression that returns the integer offset value for an argument | |
7656 | having address @var{x} (an RTL expression). The nominal offset is | |
7657 | @var{offset}. | |
7658 | ||
7659 | @findex PREFERRED_DEBUGGING_TYPE | |
7660 | @item PREFERRED_DEBUGGING_TYPE | |
a3a15b4d | 7661 | A C expression that returns the type of debugging output GCC should |
630d3d5a | 7662 | produce when the user specifies just @option{-g}. Define |
a3a15b4d | 7663 | this if you have arranged for GCC to support more than one format of |
e5e809f4 | 7664 | debugging output. Currently, the allowable values are @code{DBX_DEBUG}, |
5f98259a RK |
7665 | @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG}, |
7666 | @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}. | |
feca2ed3 | 7667 | |
630d3d5a | 7668 | When the user specifies @option{-ggdb}, GCC normally also uses the |
e5e809f4 JL |
7669 | value of this macro to select the debugging output format, but with two |
7670 | exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined and | |
a3a15b4d | 7671 | @code{LINKER_DOES_NOT_WORK_WITH_DWARF2} is not defined, GCC uses the |
e5e809f4 | 7672 | value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is |
a3a15b4d | 7673 | defined, GCC uses @code{DBX_DEBUG}. |
deabc777 | 7674 | |
feca2ed3 | 7675 | The value of this macro only affects the default debugging output; the |
630d3d5a | 7676 | user can always get a specific type of output by using @option{-gstabs}, |
5f98259a RK |
7677 | @option{-gcoff}, @option{-gdwarf-1}, @option{-gdwarf-2}, @option{-gxcoff}, |
7678 | or @option{-gvms}. | |
feca2ed3 JW |
7679 | @end table |
7680 | ||
7681 | @node DBX Options | |
7682 | @subsection Specific Options for DBX Output | |
7683 | ||
7684 | @c prevent bad page break with this line | |
7685 | These are specific options for DBX output. | |
7686 | ||
7687 | @table @code | |
7688 | @findex DBX_DEBUGGING_INFO | |
7689 | @item DBX_DEBUGGING_INFO | |
a3a15b4d | 7690 | Define this macro if GCC should produce debugging output for DBX |
630d3d5a | 7691 | in response to the @option{-g} option. |
feca2ed3 JW |
7692 | |
7693 | @findex XCOFF_DEBUGGING_INFO | |
7694 | @item XCOFF_DEBUGGING_INFO | |
a3a15b4d | 7695 | Define this macro if GCC should produce XCOFF format debugging output |
630d3d5a | 7696 | in response to the @option{-g} option. This is a variant of DBX format. |
feca2ed3 JW |
7697 | |
7698 | @findex DEFAULT_GDB_EXTENSIONS | |
7699 | @item DEFAULT_GDB_EXTENSIONS | |
a3a15b4d | 7700 | Define this macro to control whether GCC should by default generate |
feca2ed3 JW |
7701 | GDB's extended version of DBX debugging information (assuming DBX-format |
7702 | debugging information is enabled at all). If you don't define the | |
7703 | macro, the default is 1: always generate the extended information | |
7704 | if there is any occasion to. | |
7705 | ||
7706 | @findex DEBUG_SYMS_TEXT | |
7707 | @item DEBUG_SYMS_TEXT | |
7708 | Define this macro if all @code{.stabs} commands should be output while | |
7709 | in the text section. | |
7710 | ||
7711 | @findex ASM_STABS_OP | |
7712 | @item ASM_STABS_OP | |
047c1c92 HPN |
7713 | A C string constant, including spacing, naming the assembler pseudo op to |
7714 | use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol. | |
7715 | If you don't define this macro, @code{"\t.stabs\t"} is used. This macro | |
7716 | applies only to DBX debugging information format. | |
feca2ed3 JW |
7717 | |
7718 | @findex ASM_STABD_OP | |
7719 | @item ASM_STABD_OP | |
047c1c92 HPN |
7720 | A C string constant, including spacing, naming the assembler pseudo op to |
7721 | use instead of @code{"\t.stabd\t"} to define a debugging symbol whose | |
7722 | value is the current location. If you don't define this macro, | |
7723 | @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging | |
7724 | information format. | |
feca2ed3 JW |
7725 | |
7726 | @findex ASM_STABN_OP | |
7727 | @item ASM_STABN_OP | |
047c1c92 HPN |
7728 | A C string constant, including spacing, naming the assembler pseudo op to |
7729 | use instead of @code{"\t.stabn\t"} to define a debugging symbol with no | |
7730 | name. If you don't define this macro, @code{"\t.stabn\t"} is used. This | |
7731 | macro applies only to DBX debugging information format. | |
feca2ed3 JW |
7732 | |
7733 | @findex DBX_NO_XREFS | |
7734 | @item DBX_NO_XREFS | |
7735 | Define this macro if DBX on your system does not support the construct | |
7736 | @samp{xs@var{tagname}}. On some systems, this construct is used to | |
7737 | describe a forward reference to a structure named @var{tagname}. | |
7738 | On other systems, this construct is not supported at all. | |
7739 | ||
7740 | @findex DBX_CONTIN_LENGTH | |
7741 | @item DBX_CONTIN_LENGTH | |
7742 | A symbol name in DBX-format debugging information is normally | |
7743 | continued (split into two separate @code{.stabs} directives) when it | |
7744 | exceeds a certain length (by default, 80 characters). On some | |
7745 | operating systems, DBX requires this splitting; on others, splitting | |
7746 | must not be done. You can inhibit splitting by defining this macro | |
7747 | with the value zero. You can override the default splitting-length by | |
7748 | defining this macro as an expression for the length you desire. | |
7749 | ||
7750 | @findex DBX_CONTIN_CHAR | |
7751 | @item DBX_CONTIN_CHAR | |
7752 | Normally continuation is indicated by adding a @samp{\} character to | |
7753 | the end of a @code{.stabs} string when a continuation follows. To use | |
7754 | a different character instead, define this macro as a character | |
7755 | constant for the character you want to use. Do not define this macro | |
7756 | if backslash is correct for your system. | |
7757 | ||
7758 | @findex DBX_STATIC_STAB_DATA_SECTION | |
7759 | @item DBX_STATIC_STAB_DATA_SECTION | |
7760 | Define this macro if it is necessary to go to the data section before | |
7761 | outputting the @samp{.stabs} pseudo-op for a non-global static | |
7762 | variable. | |
7763 | ||
7764 | @findex DBX_TYPE_DECL_STABS_CODE | |
7765 | @item DBX_TYPE_DECL_STABS_CODE | |
7766 | The value to use in the ``code'' field of the @code{.stabs} directive | |
7767 | for a typedef. The default is @code{N_LSYM}. | |
7768 | ||
7769 | @findex DBX_STATIC_CONST_VAR_CODE | |
7770 | @item DBX_STATIC_CONST_VAR_CODE | |
7771 | The value to use in the ``code'' field of the @code{.stabs} directive | |
7772 | for a static variable located in the text section. DBX format does not | |
7773 | provide any ``right'' way to do this. The default is @code{N_FUN}. | |
7774 | ||
7775 | @findex DBX_REGPARM_STABS_CODE | |
7776 | @item DBX_REGPARM_STABS_CODE | |
7777 | The value to use in the ``code'' field of the @code{.stabs} directive | |
7778 | for a parameter passed in registers. DBX format does not provide any | |
7779 | ``right'' way to do this. The default is @code{N_RSYM}. | |
7780 | ||
7781 | @findex DBX_REGPARM_STABS_LETTER | |
7782 | @item DBX_REGPARM_STABS_LETTER | |
7783 | The letter to use in DBX symbol data to identify a symbol as a parameter | |
7784 | passed in registers. DBX format does not customarily provide any way to | |
7785 | do this. The default is @code{'P'}. | |
7786 | ||
7787 | @findex DBX_MEMPARM_STABS_LETTER | |
7788 | @item DBX_MEMPARM_STABS_LETTER | |
7789 | The letter to use in DBX symbol data to identify a symbol as a stack | |
7790 | parameter. The default is @code{'p'}. | |
7791 | ||
7792 | @findex DBX_FUNCTION_FIRST | |
7793 | @item DBX_FUNCTION_FIRST | |
7794 | Define this macro if the DBX information for a function and its | |
7795 | arguments should precede the assembler code for the function. Normally, | |
7796 | in DBX format, the debugging information entirely follows the assembler | |
7797 | code. | |
7798 | ||
7799 | @findex DBX_LBRAC_FIRST | |
7800 | @item DBX_LBRAC_FIRST | |
7801 | Define this macro if the @code{N_LBRAC} symbol for a block should | |
7802 | precede the debugging information for variables and functions defined in | |
7803 | that block. Normally, in DBX format, the @code{N_LBRAC} symbol comes | |
7804 | first. | |
7805 | ||
7806 | @findex DBX_BLOCKS_FUNCTION_RELATIVE | |
7807 | @item DBX_BLOCKS_FUNCTION_RELATIVE | |
7808 | Define this macro if the value of a symbol describing the scope of a | |
7809 | block (@code{N_LBRAC} or @code{N_RBRAC}) should be relative to the start | |
f0523f02 | 7810 | of the enclosing function. Normally, GCC uses an absolute address. |
feca2ed3 JW |
7811 | |
7812 | @findex DBX_USE_BINCL | |
7813 | @item DBX_USE_BINCL | |
f0523f02 | 7814 | Define this macro if GCC should generate @code{N_BINCL} and |
feca2ed3 | 7815 | @code{N_EINCL} stabs for included header files, as on Sun systems. This |
f0523f02 JM |
7816 | macro also directs GCC to output a type number as a pair of a file |
7817 | number and a type number within the file. Normally, GCC does not | |
feca2ed3 JW |
7818 | generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single |
7819 | number for a type number. | |
7820 | @end table | |
7821 | ||
7822 | @node DBX Hooks | |
7823 | @subsection Open-Ended Hooks for DBX Format | |
7824 | ||
7825 | @c prevent bad page break with this line | |
7826 | These are hooks for DBX format. | |
7827 | ||
7828 | @table @code | |
7829 | @findex DBX_OUTPUT_LBRAC | |
7830 | @item DBX_OUTPUT_LBRAC (@var{stream}, @var{name}) | |
7831 | Define this macro to say how to output to @var{stream} the debugging | |
7832 | information for the start of a scope level for variable names. The | |
7833 | argument @var{name} is the name of an assembler symbol (for use with | |
7834 | @code{assemble_name}) whose value is the address where the scope begins. | |
7835 | ||
7836 | @findex DBX_OUTPUT_RBRAC | |
7837 | @item DBX_OUTPUT_RBRAC (@var{stream}, @var{name}) | |
7838 | Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level. | |
7839 | ||
374b0b7d AM |
7840 | @findex DBX_OUTPUT_NFUN |
7841 | @item DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl}) | |
7842 | Define this macro if the target machine requires special handling to | |
7843 | output an @code{N_FUN} entry for the function @var{decl}. | |
7844 | ||
feca2ed3 JW |
7845 | @findex DBX_OUTPUT_ENUM |
7846 | @item DBX_OUTPUT_ENUM (@var{stream}, @var{type}) | |
7847 | Define this macro if the target machine requires special handling to | |
7848 | output an enumeration type. The definition should be a C statement | |
7849 | (sans semicolon) to output the appropriate information to @var{stream} | |
7850 | for the type @var{type}. | |
7851 | ||
7852 | @findex DBX_OUTPUT_FUNCTION_END | |
7853 | @item DBX_OUTPUT_FUNCTION_END (@var{stream}, @var{function}) | |
7854 | Define this macro if the target machine requires special output at the | |
7855 | end of the debugging information for a function. The definition should | |
7856 | be a C statement (sans semicolon) to output the appropriate information | |
7857 | to @var{stream}. @var{function} is the @code{FUNCTION_DECL} node for | |
7858 | the function. | |
7859 | ||
7860 | @findex DBX_OUTPUT_STANDARD_TYPES | |
7861 | @item DBX_OUTPUT_STANDARD_TYPES (@var{syms}) | |
7862 | Define this macro if you need to control the order of output of the | |
7863 | standard data types at the beginning of compilation. The argument | |
7864 | @var{syms} is a @code{tree} which is a chain of all the predefined | |
7865 | global symbols, including names of data types. | |
7866 | ||
7867 | Normally, DBX output starts with definitions of the types for integers | |
7868 | and characters, followed by all the other predefined types of the | |
7869 | particular language in no particular order. | |
7870 | ||
7871 | On some machines, it is necessary to output different particular types | |
7872 | first. To do this, define @code{DBX_OUTPUT_STANDARD_TYPES} to output | |
7873 | those symbols in the necessary order. Any predefined types that you | |
7874 | don't explicitly output will be output afterward in no particular order. | |
7875 | ||
161d7b59 | 7876 | Be careful not to define this macro so that it works only for C@. There |
feca2ed3 JW |
7877 | are no global variables to access most of the built-in types, because |
7878 | another language may have another set of types. The way to output a | |
7879 | particular type is to look through @var{syms} to see if you can find it. | |
7880 | Here is an example: | |
7881 | ||
7882 | @smallexample | |
7883 | @{ | |
7884 | tree decl; | |
7885 | for (decl = syms; decl; decl = TREE_CHAIN (decl)) | |
7886 | if (!strcmp (IDENTIFIER_POINTER (DECL_NAME (decl)), | |
7887 | "long int")) | |
7888 | dbxout_symbol (decl); | |
7889 | @dots{} | |
7890 | @} | |
7891 | @end smallexample | |
7892 | ||
7893 | @noindent | |
7894 | This does nothing if the expected type does not exist. | |
7895 | ||
7896 | See the function @code{init_decl_processing} in @file{c-decl.c} to find | |
7897 | the names to use for all the built-in C types. | |
7898 | ||
7899 | Here is another way of finding a particular type: | |
7900 | ||
7901 | @c this is still overfull. --mew 10feb93 | |
7902 | @smallexample | |
7903 | @{ | |
7904 | tree decl; | |
7905 | for (decl = syms; decl; decl = TREE_CHAIN (decl)) | |
7906 | if (TREE_CODE (decl) == TYPE_DECL | |
7907 | && (TREE_CODE (TREE_TYPE (decl)) | |
7908 | == INTEGER_CST) | |
7909 | && TYPE_PRECISION (TREE_TYPE (decl)) == 16 | |
7910 | && TYPE_UNSIGNED (TREE_TYPE (decl))) | |
7911 | @group | |
7912 | /* @r{This must be @code{unsigned short}.} */ | |
7913 | dbxout_symbol (decl); | |
7914 | @dots{} | |
7915 | @} | |
7916 | @end group | |
7917 | @end smallexample | |
7918 | ||
7919 | @findex NO_DBX_FUNCTION_END | |
7920 | @item NO_DBX_FUNCTION_END | |
7921 | Some stabs encapsulation formats (in particular ECOFF), cannot handle the | |
c771326b | 7922 | @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct. |
feca2ed3 JW |
7923 | On those machines, define this macro to turn this feature off without |
7924 | disturbing the rest of the gdb extensions. | |
7925 | ||
7926 | @end table | |
7927 | ||
7928 | @node File Names and DBX | |
7929 | @subsection File Names in DBX Format | |
7930 | ||
7931 | @c prevent bad page break with this line | |
7932 | This describes file names in DBX format. | |
7933 | ||
7934 | @table @code | |
7935 | @findex DBX_WORKING_DIRECTORY | |
7936 | @item DBX_WORKING_DIRECTORY | |
7937 | Define this if DBX wants to have the current directory recorded in each | |
7938 | object file. | |
7939 | ||
7940 | Note that the working directory is always recorded if GDB extensions are | |
7941 | enabled. | |
7942 | ||
7943 | @findex DBX_OUTPUT_MAIN_SOURCE_FILENAME | |
7944 | @item DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name}) | |
7945 | A C statement to output DBX debugging information to the stdio stream | |
7946 | @var{stream} which indicates that file @var{name} is the main source | |
7947 | file---the file specified as the input file for compilation. | |
7948 | This macro is called only once, at the beginning of compilation. | |
7949 | ||
7950 | This macro need not be defined if the standard form of output | |
7951 | for DBX debugging information is appropriate. | |
7952 | ||
7953 | @findex DBX_OUTPUT_MAIN_SOURCE_DIRECTORY | |
7954 | @item DBX_OUTPUT_MAIN_SOURCE_DIRECTORY (@var{stream}, @var{name}) | |
7955 | A C statement to output DBX debugging information to the stdio stream | |
7956 | @var{stream} which indicates that the current directory during | |
7957 | compilation is named @var{name}. | |
7958 | ||
7959 | This macro need not be defined if the standard form of output | |
7960 | for DBX debugging information is appropriate. | |
7961 | ||
7962 | @findex DBX_OUTPUT_MAIN_SOURCE_FILE_END | |
7963 | @item DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name}) | |
7964 | A C statement to output DBX debugging information at the end of | |
7965 | compilation of the main source file @var{name}. | |
7966 | ||
7967 | If you don't define this macro, nothing special is output at the end | |
7968 | of compilation, which is correct for most machines. | |
7969 | ||
7970 | @findex DBX_OUTPUT_SOURCE_FILENAME | |
7971 | @item DBX_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name}) | |
7972 | A C statement to output DBX debugging information to the stdio stream | |
7973 | @var{stream} which indicates that file @var{name} is the current source | |
7974 | file. This output is generated each time input shifts to a different | |
7975 | source file as a result of @samp{#include}, the end of an included file, | |
7976 | or a @samp{#line} command. | |
7977 | ||
7978 | This macro need not be defined if the standard form of output | |
7979 | for DBX debugging information is appropriate. | |
7980 | @end table | |
7981 | ||
7982 | @need 2000 | |
7983 | @node SDB and DWARF | |
7984 | @subsection Macros for SDB and DWARF Output | |
7985 | ||
7986 | @c prevent bad page break with this line | |
7987 | Here are macros for SDB and DWARF output. | |
7988 | ||
7989 | @table @code | |
7990 | @findex SDB_DEBUGGING_INFO | |
7991 | @item SDB_DEBUGGING_INFO | |
a3a15b4d | 7992 | Define this macro if GCC should produce COFF-style debugging output |
630d3d5a | 7993 | for SDB in response to the @option{-g} option. |
feca2ed3 JW |
7994 | |
7995 | @findex DWARF_DEBUGGING_INFO | |
7996 | @item DWARF_DEBUGGING_INFO | |
a3a15b4d | 7997 | Define this macro if GCC should produce dwarf format debugging output |
630d3d5a | 7998 | in response to the @option{-g} option. |
feca2ed3 | 7999 | |
f3ff3f4a JM |
8000 | @findex DWARF2_DEBUGGING_INFO |
8001 | @item DWARF2_DEBUGGING_INFO | |
a3a15b4d | 8002 | Define this macro if GCC should produce dwarf version 2 format |
630d3d5a | 8003 | debugging output in response to the @option{-g} option. |
f3ff3f4a | 8004 | |
861bb6c1 JL |
8005 | To support optional call frame debugging information, you must also |
8006 | define @code{INCOMING_RETURN_ADDR_RTX} and either set | |
8007 | @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the | |
8008 | prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save} | |
08c148a8 | 8009 | as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't. |
861bb6c1 | 8010 | |
9ec36da5 JL |
8011 | @findex DWARF2_FRAME_INFO |
8012 | @item DWARF2_FRAME_INFO | |
a3a15b4d | 8013 | Define this macro to a nonzero value if GCC should always output |
9ec36da5 | 8014 | Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO} |
a3a15b4d | 8015 | (@pxref{Exception Region Output} is nonzero, GCC will output this |
9ec36da5 JL |
8016 | information not matter how you define @code{DWARF2_FRAME_INFO}. |
8017 | ||
deabc777 PE |
8018 | @findex LINKER_DOES_NOT_WORK_WITH_DWARF2 |
8019 | @item LINKER_DOES_NOT_WORK_WITH_DWARF2 | |
e5e809f4 | 8020 | Define this macro if the linker does not work with Dwarf version 2. |
630d3d5a | 8021 | Normally, if the user specifies only @option{-ggdb} GCC will use Dwarf |
e5e809f4 JL |
8022 | version 2 if available; this macro disables this. See the description |
8023 | of the @code{PREFERRED_DEBUGGING_TYPE} macro for more details. | |
deabc777 | 8024 | |
b366352b MM |
8025 | @findex DWARF2_GENERATE_TEXT_SECTION_LABEL |
8026 | @item DWARF2_GENERATE_TEXT_SECTION_LABEL | |
8027 | By default, the Dwarf 2 debugging information generator will generate a | |
8028 | label to mark the beginning of the text section. If it is better simply | |
8029 | to use the name of the text section itself, rather than an explicit label, | |
8030 | to indicate the beginning of the text section, define this macro to zero. | |
8031 | ||
b2244e22 JW |
8032 | @findex DWARF2_ASM_LINE_DEBUG_INFO |
8033 | @item DWARF2_ASM_LINE_DEBUG_INFO | |
8034 | Define this macro to be a nonzero value if the assembler can generate Dwarf 2 | |
8035 | line debug info sections. This will result in much more compact line number | |
8036 | tables, and hence is desirable if it works. | |
8037 | ||
7606e68f SS |
8038 | @findex ASM_OUTPUT_DWARF_DELTA |
8039 | @item ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2}) | |
8040 | A C statement to issue assembly directives that create a difference | |
8041 | between the two given labels, using an integer of the given size. | |
8042 | ||
8043 | @findex ASM_OUTPUT_DWARF_OFFSET | |
8044 | @item ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}) | |
8045 | A C statement to issue assembly directives that create a | |
8046 | section-relative reference to the given label, using an integer of the | |
8047 | given size. | |
8048 | ||
8049 | @findex ASM_OUTPUT_DWARF_PCREL | |
8050 | @item ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label}) | |
8051 | A C statement to issue assembly directives that create a self-relative | |
8052 | reference to the given label, using an integer of the given size. | |
8053 | ||
feca2ed3 JW |
8054 | @findex PUT_SDB_@dots{} |
8055 | @item PUT_SDB_@dots{} | |
8056 | Define these macros to override the assembler syntax for the special | |
8057 | SDB assembler directives. See @file{sdbout.c} for a list of these | |
8058 | macros and their arguments. If the standard syntax is used, you need | |
8059 | not define them yourself. | |
8060 | ||
8061 | @findex SDB_DELIM | |
8062 | @item SDB_DELIM | |
8063 | Some assemblers do not support a semicolon as a delimiter, even between | |
8064 | SDB assembler directives. In that case, define this macro to be the | |
8065 | delimiter to use (usually @samp{\n}). It is not necessary to define | |
8066 | a new set of @code{PUT_SDB_@var{op}} macros if this is the only change | |
8067 | required. | |
8068 | ||
8069 | @findex SDB_GENERATE_FAKE | |
8070 | @item SDB_GENERATE_FAKE | |
8071 | Define this macro to override the usual method of constructing a dummy | |
8072 | name for anonymous structure and union types. See @file{sdbout.c} for | |
8073 | more information. | |
8074 | ||
8075 | @findex SDB_ALLOW_UNKNOWN_REFERENCES | |
8076 | @item SDB_ALLOW_UNKNOWN_REFERENCES | |
8077 | Define this macro to allow references to unknown structure, | |
8078 | union, or enumeration tags to be emitted. Standard COFF does not | |
8079 | allow handling of unknown references, MIPS ECOFF has support for | |
8080 | it. | |
8081 | ||
8082 | @findex SDB_ALLOW_FORWARD_REFERENCES | |
8083 | @item SDB_ALLOW_FORWARD_REFERENCES | |
8084 | Define this macro to allow references to structure, union, or | |
8085 | enumeration tags that have not yet been seen to be handled. Some | |
8086 | assemblers choke if forward tags are used, while some require it. | |
8087 | @end table | |
8088 | ||
5f98259a RK |
8089 | @need 2000 |
8090 | @node VMS Debug | |
8091 | @subsection Macros for VMS Debug Format | |
8092 | ||
8093 | @c prevent bad page break with this line | |
8094 | Here are macros for VMS debug format. | |
8095 | ||
8096 | @table @code | |
8097 | @findex VMS_DEBUGGING_INFO | |
8098 | @item VMS_DEBUGGING_INFO | |
8099 | Define this macro if GCC should produce debugging output for VMS | |
8100 | in response to the @option{-g} option. The default behavior for VMS | |
8101 | is to generate minimal debug info for a traceback in the absence of | |
8102 | @option{-g} unless explicitly overridden with @option{-g0}. This | |
8103 | behavior is controlled by @code{OPTIMIZATION_OPTIONS} and | |
8104 | @code{OVERRIDE_OPTIONS}. | |
8105 | @end table | |
8106 | ||
b216cd4a | 8107 | @node Floating Point |
feca2ed3 JW |
8108 | @section Cross Compilation and Floating Point |
8109 | @cindex cross compilation and floating point | |
8110 | @cindex floating point and cross compilation | |
8111 | ||
b216cd4a | 8112 | While all modern machines use twos-complement representation for integers, |
feca2ed3 JW |
8113 | there are a variety of representations for floating point numbers. This |
8114 | means that in a cross-compiler the representation of floating point numbers | |
8115 | in the compiled program may be different from that used in the machine | |
8116 | doing the compilation. | |
8117 | ||
feca2ed3 | 8118 | Because different representation systems may offer different amounts of |
b216cd4a ZW |
8119 | range and precision, all floating point constants must be represented in |
8120 | the target machine's format. Therefore, the cross compiler cannot | |
8121 | safely use the host machine's floating point arithmetic; it must emulate | |
8122 | the target's arithmetic. To ensure consistency, GCC always uses | |
8123 | emulation to work with floating point values, even when the host and | |
8124 | target floating point formats are identical. | |
8125 | ||
8126 | The following macros are provided by @file{real.h} for the compiler to | |
8127 | use. All parts of the compiler which generate or optimize | |
ba31d94e ZW |
8128 | floating-point calculations must use these macros. They may evaluate |
8129 | their operands more than once, so operands must not have side effects. | |
feca2ed3 | 8130 | |
b216cd4a ZW |
8131 | @defmac REAL_VALUE_TYPE |
8132 | The C data type to be used to hold a floating point value in the target | |
8133 | machine's format. Typically this is a @code{struct} containing an | |
8134 | array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque | |
8135 | quantity. | |
8136 | @end defmac | |
8137 | ||
8138 | @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) | |
8139 | Compares for equality the two values, @var{x} and @var{y}. If the target | |
8140 | floating point format supports negative zeroes and/or NaNs, | |
8141 | @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and | |
8142 | @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false. | |
8143 | @end deftypefn | |
8144 | ||
8145 | @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) | |
8146 | Tests whether @var{x} is less than @var{y}. | |
8147 | @end deftypefn | |
8148 | ||
b216cd4a ZW |
8149 | @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x}) |
8150 | Truncates @var{x} to a signed integer, rounding toward zero. | |
8151 | @end deftypefn | |
8152 | ||
8153 | @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x}) | |
8154 | Truncates @var{x} to an unsigned integer, rounding toward zero. If | |
8155 | @var{x} is negative, returns zero. | |
8156 | @end deftypefn | |
8157 | ||
b216cd4a ZW |
8158 | @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode}) |
8159 | Converts @var{string} into a floating point number in the target machine's | |
8160 | representation for mode @var{mode}. This routine can handle both | |
8161 | decimal and hexadecimal floating point constants, using the syntax | |
8162 | defined by the C language for both. | |
8163 | @end deftypefn | |
feca2ed3 | 8164 | |
15e5ad76 | 8165 | @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x}) |
ce3649d2 | 8166 | Returns 1 if @var{x} is negative (including negative zero), 0 otherwise. |
15e5ad76 ZW |
8167 | @end deftypefn |
8168 | ||
b216cd4a ZW |
8169 | @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x}) |
8170 | Determines whether @var{x} represents infinity (positive or negative). | |
8171 | @end deftypefn | |
8172 | ||
8173 | @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x}) | |
8174 | Determines whether @var{x} represents a ``NaN'' (not-a-number). | |
8175 | @end deftypefn | |
8176 | ||
8177 | @deftypefn Macro void REAL_ARITHMETIC (REAL_VALUE_TYPE @var{output}, enum tree_code @var{code}, REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y}) | |
8178 | Calculates an arithmetic operation on the two floating point values | |
8179 | @var{x} and @var{y}, storing the result in @var{output} (which must be a | |
8180 | variable). | |
8181 | ||
8182 | The operation to be performed is specified by @var{code}. Only the | |
8183 | following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR}, | |
8184 | @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}. | |
8185 | ||
8186 | If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the | |
8187 | target's floating point format cannot represent infinity, it will call | |
8188 | @code{abort}. Callers should check for this situation first, using | |
8189 | @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}. | |
8190 | @end deftypefn | |
8191 | ||
8192 | @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x}) | |
8193 | Returns the negative of the floating point value @var{x}. | |
8194 | @end deftypefn | |
8195 | ||
15e5ad76 ZW |
8196 | @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x}) |
8197 | Returns the absolute value of @var{x}. | |
8198 | @end deftypefn | |
8199 | ||
b216cd4a ZW |
8200 | @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x}) |
8201 | Truncates the floating point value @var{x} to fit in @var{mode}. The | |
8202 | return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an | |
8203 | appropriate bit pattern to be output asa floating constant whose | |
8204 | precision accords with mode @var{mode}. | |
8205 | @end deftypefn | |
8206 | ||
8207 | @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x}) | |
8208 | Converts a floating point value @var{x} into a double-precision integer | |
8209 | which is then stored into @var{low} and @var{high}. If the value is not | |
8210 | integral, it is truncated. | |
8211 | @end deftypefn | |
8212 | ||
8213 | @deftypefn Macro void REAL_VALUE_FROM_INT (REAL_VALUE_TYPE @var{x}, HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, enum machine_mode @var{mode}) | |
feca2ed3 | 8214 | @findex REAL_VALUE_FROM_INT |
b216cd4a ZW |
8215 | Converts a double-precision integer found in @var{low} and @var{high}, |
8216 | into a floating point value which is then stored into @var{x}. The | |
8217 | value is truncated to fit in mode @var{mode}. | |
8218 | @end deftypefn | |
feca2ed3 | 8219 | |
9f09b1f2 R |
8220 | @node Mode Switching |
8221 | @section Mode Switching Instructions | |
8222 | @cindex mode switching | |
8223 | The following macros control mode switching optimizations: | |
8224 | ||
8225 | @table @code | |
8226 | @findex OPTIMIZE_MODE_SWITCHING | |
8227 | @item OPTIMIZE_MODE_SWITCHING (@var{entity}) | |
8228 | Define this macro if the port needs extra instructions inserted for mode | |
8229 | switching in an optimizing compilation. | |
8230 | ||
8231 | For an example, the SH4 can perform both single and double precision | |
8232 | floating point operations, but to perform a single precision operation, | |
8233 | the FPSCR PR bit has to be cleared, while for a double precision | |
8234 | operation, this bit has to be set. Changing the PR bit requires a general | |
8235 | purpose register as a scratch register, hence these FPSCR sets have to | |
e979f9e8 | 8236 | be inserted before reload, i.e.@: you can't put this into instruction emitting |
aee96fe9 | 8237 | or @code{MACHINE_DEPENDENT_REORG}. |
9f09b1f2 R |
8238 | |
8239 | You can have multiple entities that are mode-switched, and select at run time | |
8240 | which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should | |
14976c58 | 8241 | return nonzero for any @var{entity} that needs mode-switching. |
9f09b1f2 R |
8242 | If you define this macro, you also have to define |
8243 | @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED}, | |
8244 | @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}. | |
1270c255 | 8245 | @code{NORMAL_MODE} is optional. |
9f09b1f2 R |
8246 | |
8247 | @findex NUM_MODES_FOR_MODE_SWITCHING | |
8248 | @item NUM_MODES_FOR_MODE_SWITCHING | |
8249 | If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as | |
8250 | initializer for an array of integers. Each initializer element | |
8251 | N refers to an entity that needs mode switching, and specifies the number | |
8252 | of different modes that might need to be set for this entity. | |
8253 | The position of the initializer in the initializer - starting counting at | |
8254 | zero - determines the integer that is used to refer to the mode-switched | |
8255 | entity in question. | |
8256 | In macros that take mode arguments / yield a mode result, modes are | |
630d3d5a | 8257 | represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode |
9f09b1f2 R |
8258 | switch is needed / supplied. |
8259 | ||
9f09b1f2 R |
8260 | @findex MODE_NEEDED |
8261 | @item MODE_NEEDED (@var{entity}, @var{insn}) | |
8262 | @var{entity} is an integer specifying a mode-switched entity. If | |
8263 | @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to | |
8264 | return an integer value not larger than the corresponding element in | |
aee96fe9 JM |
8265 | @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must |
8266 | be switched into prior to the execution of @var{insn}. | |
9f09b1f2 | 8267 | |
02f52e19 | 8268 | @findex NORMAL_MODE |
1270c255 | 8269 | @item NORMAL_MODE (@var{entity}) |
9f09b1f2 R |
8270 | If this macro is defined, it is evaluated for every @var{entity} that needs |
8271 | mode switching. It should evaluate to an integer, which is a mode that | |
1270c255 | 8272 | @var{entity} is assumed to be switched to at function entry and exit. |
9f09b1f2 R |
8273 | |
8274 | @findex MODE_PRIORITY_TO_MODE | |
8275 | @item MODE_PRIORITY_TO_MODE (@var{entity}, @var{n}) | |
aee96fe9 JM |
8276 | This macro specifies the order in which modes for @var{entity} are processed. |
8277 | 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the | |
9f09b1f2 | 8278 | lowest. The value of the macro should be an integer designating a mode |
aee96fe9 | 8279 | for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode} |
630d3d5a | 8280 | (@var{entity}, @var{n}) shall be a bijection in 0 @dots{} |
aee96fe9 | 8281 | @code{num_modes_for_mode_switching[@var{entity}] - 1}. |
9f09b1f2 R |
8282 | |
8283 | @findex EMIT_MODE_SET | |
8284 | @item EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live}) | |
8285 | Generate one or more insns to set @var{entity} to @var{mode}. | |
8286 | @var{hard_reg_live} is the set of hard registers live at the point where | |
8287 | the insn(s) are to be inserted. | |
8288 | @end table | |
8289 | ||
91d231cb JM |
8290 | @node Target Attributes |
8291 | @section Defining target-specific uses of @code{__attribute__} | |
8292 | @cindex target attributes | |
8293 | @cindex machine attributes | |
8294 | @cindex attributes, target-specific | |
8295 | ||
8296 | Target-specific attributes may be defined for functions, data and types. | |
8297 | These are described using the following target hooks; they also need to | |
8298 | be documented in @file{extend.texi}. | |
8299 | ||
8300 | @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE | |
8301 | If defined, this target hook points to an array of @samp{struct | |
8302 | attribute_spec} (defined in @file{tree.h}) specifying the machine | |
8303 | specific attributes for this target and some of the restrictions on the | |
8304 | entities to which these attributes are applied and the arguments they | |
8305 | take. | |
8306 | @end deftypevr | |
8307 | ||
8308 | @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2}) | |
8309 | If defined, this target hook is a function which returns zero if the attributes on | |
8310 | @var{type1} and @var{type2} are incompatible, one if they are compatible, | |
8311 | and two if they are nearly compatible (which causes a warning to be | |
8312 | generated). If this is not defined, machine-specific attributes are | |
8313 | supposed always to be compatible. | |
8314 | @end deftypefn | |
8315 | ||
8316 | @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type}) | |
8317 | If defined, this target hook is a function which assigns default attributes to | |
8318 | newly defined @var{type}. | |
8319 | @end deftypefn | |
8320 | ||
8321 | @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2}) | |
8322 | Define this target hook if the merging of type attributes needs special | |
8323 | handling. If defined, the result is a list of the combined | |
8324 | @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed | |
8325 | that @code{comptypes} has already been called and returned 1. This | |
8326 | function may call @code{merge_attributes} to handle machine-independent | |
8327 | merging. | |
8328 | @end deftypefn | |
8329 | ||
8330 | @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl}) | |
8331 | Define this target hook if the merging of decl attributes needs special | |
8332 | handling. If defined, the result is a list of the combined | |
8333 | @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}. | |
8334 | @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of | |
8335 | when this is needed are when one attribute overrides another, or when an | |
8336 | attribute is nullified by a subsequent definition. This function may | |
8337 | call @code{merge_attributes} to handle machine-independent merging. | |
8338 | ||
8339 | @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES | |
8340 | If the only target-specific handling you require is @samp{dllimport} for | |
8341 | Windows targets, you should define the macro | |
8342 | @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. This links in a function | |
8343 | called @code{merge_dllimport_decl_attributes} which can then be defined | |
8344 | as the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. This is done | |
8345 | in @file{i386/cygwin.h} and @file{i386/i386.c}, for example. | |
8346 | @end deftypefn | |
8347 | ||
8348 | @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr}) | |
8349 | Define this target hook if you want to be able to add attributes to a decl | |
8350 | when it is being created. This is normally useful for back ends which | |
8351 | wish to implement a pragma by using the attributes which correspond to | |
8352 | the pragma's effect. The @var{node} argument is the decl which is being | |
8353 | created. The @var{attr_ptr} argument is a pointer to the attribute list | |
8354 | for this decl. The list itself should not be modified, since it may be | |
8355 | shared with other decls, but attributes may be chained on the head of | |
8356 | the list and @code{*@var{attr_ptr}} modified to point to the new | |
8357 | attributes, or a copy of the list may be made if further changes are | |
8358 | needed. | |
8359 | @end deftypefn | |
8360 | ||
8361 | @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl}) | |
8362 | @cindex inlining | |
8363 | This target hook returns @code{true} if it is ok to inline @var{fndecl} | |
8364 | into the current function, despite its having target-specific | |
8365 | attributes, @code{false} otherwise. By default, if a function has a | |
8366 | target specific attribute attached to it, it will not be inlined. | |
8367 | @end deftypefn | |
8368 | ||
d604bca3 MH |
8369 | @node MIPS Coprocessors |
8370 | @section Defining coprocessor specifics for MIPS targets. | |
8371 | @cindex MIPS coprocessor-definition macros | |
8372 | ||
8373 | The MIPS specification allows MIPS implementations to have as many as 4 | |
8374 | coprocessors, each with as many as 32 private registers. gcc supports | |
8375 | accessing these registers and transferring values between the registers | |
8376 | and memory using asm-ized variables. For example: | |
8377 | ||
8378 | @smallexample | |
8379 | register unsigned int cp0count asm ("c0r1"); | |
8380 | unsigned int d; | |
8381 | ||
8382 | d = cp0count + 3; | |
8383 | @end smallexample | |
8384 | ||
8385 | (``c0r1'' is the default name of register 1 in coprocessor 0; alternate | |
8386 | names may be added as described below, or the default names may be | |
8387 | overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.) | |
8388 | ||
8389 | Coprocessor registers are assumed to be epilogue-used; sets to them will | |
8390 | be preserved even if it does not appear that the register is used again | |
8391 | later in the function. | |
8392 | ||
8393 | Another note: according to the MIPS spec, coprocessor 1 (if present) is | |
8394 | the FPU. One accesses COP1 registers through standard mips | |
8395 | floating-point support; they are not included in this mechanism. | |
8396 | ||
8397 | There is one macro used in defining the MIPS coprocessor interface which | |
8398 | you may want to override in subtargets; it is described below. | |
8399 | ||
8400 | @table @code | |
8401 | ||
8402 | @item ALL_COP_ADDITIONAL_REGISTER_NAMES | |
8403 | @findex ALL_COP_ADDITIONAL_REGISTER_NAMES | |
8404 | A comma-separated list (with leading comma) of pairs describing the | |
8405 | alternate names of coprocessor registers. The format of each entry should be | |
8406 | @smallexample | |
8407 | @{ @var{alternatename}, @var{register_number}@} | |
8408 | @end smallexample | |
8409 | Default: empty. | |
8410 | ||
8411 | @end table | |
8412 | ||
feca2ed3 JW |
8413 | @node Misc |
8414 | @section Miscellaneous Parameters | |
8415 | @cindex parameters, miscellaneous | |
8416 | ||
8417 | @c prevent bad page break with this line | |
8418 | Here are several miscellaneous parameters. | |
8419 | ||
8420 | @table @code | |
8421 | @item PREDICATE_CODES | |
8422 | @findex PREDICATE_CODES | |
8423 | Define this if you have defined special-purpose predicates in the file | |
8424 | @file{@var{machine}.c}. This macro is called within an initializer of an | |
8425 | array of structures. The first field in the structure is the name of a | |
8426 | predicate and the second field is an array of rtl codes. For each | |
8427 | predicate, list all rtl codes that can be in expressions matched by the | |
8428 | predicate. The list should have a trailing comma. Here is an example | |
8429 | of two entries in the list for a typical RISC machine: | |
8430 | ||
8431 | @smallexample | |
8432 | #define PREDICATE_CODES \ | |
8433 | @{"gen_reg_rtx_operand", @{SUBREG, REG@}@}, \ | |
8434 | @{"reg_or_short_cint_operand", @{SUBREG, REG, CONST_INT@}@}, | |
8435 | @end smallexample | |
8436 | ||
8437 | Defining this macro does not affect the generated code (however, | |
8438 | incorrect definitions that omit an rtl code that may be matched by the | |
8439 | predicate can cause the compiler to malfunction). Instead, it allows | |
8440 | the table built by @file{genrecog} to be more compact and efficient, | |
8441 | thus speeding up the compiler. The most important predicates to include | |
556e0f21 | 8442 | in the list specified by this macro are those used in the most insn |
feca2ed3 JW |
8443 | patterns. |
8444 | ||
aee96fe9 | 8445 | For each predicate function named in @code{PREDICATE_CODES}, a |
975d393a AO |
8446 | declaration will be generated in @file{insn-codes.h}. |
8447 | ||
8fe0ca0c RH |
8448 | @item SPECIAL_MODE_PREDICATES |
8449 | @findex SPECIAL_MODE_PREDICATES | |
8450 | Define this if you have special predicates that know special things | |
02f52e19 | 8451 | about modes. Genrecog will warn about certain forms of |
8fe0ca0c | 8452 | @code{match_operand} without a mode; if the operand predicate is |
02f52e19 | 8453 | listed in @code{SPECIAL_MODE_PREDICATES}, the warning will be |
8fe0ca0c RH |
8454 | suppressed. |
8455 | ||
8456 | Here is an example from the IA-32 port (@code{ext_register_operand} | |
8457 | specially checks for @code{HImode} or @code{SImode} in preparation | |
8458 | for a byte extraction from @code{%ah} etc.). | |
8459 | ||
8460 | @smallexample | |
8461 | #define SPECIAL_MODE_PREDICATES \ | |
8462 | "ext_register_operand", | |
8463 | @end smallexample | |
8464 | ||
feca2ed3 JW |
8465 | @findex CASE_VECTOR_MODE |
8466 | @item CASE_VECTOR_MODE | |
8467 | An alias for a machine mode name. This is the machine mode that | |
8468 | elements of a jump-table should have. | |
8469 | ||
33f7f353 JR |
8470 | @findex CASE_VECTOR_SHORTEN_MODE |
8471 | @item CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body}) | |
8472 | Optional: return the preferred mode for an @code{addr_diff_vec} | |
8473 | when the minimum and maximum offset are known. If you define this, | |
8474 | it enables extra code in branch shortening to deal with @code{addr_diff_vec}. | |
4226378a | 8475 | To make this work, you also have to define @code{INSN_ALIGN} and |
33f7f353 | 8476 | make the alignment for @code{addr_diff_vec} explicit. |
391aaa6b | 8477 | The @var{body} argument is provided so that the offset_unsigned and scale |
33f7f353 JR |
8478 | flags can be updated. |
8479 | ||
feca2ed3 JW |
8480 | @findex CASE_VECTOR_PC_RELATIVE |
8481 | @item CASE_VECTOR_PC_RELATIVE | |
18543a22 ILT |
8482 | Define this macro to be a C expression to indicate when jump-tables |
8483 | should contain relative addresses. If jump-tables never contain | |
8484 | relative addresses, then you need not define this macro. | |
feca2ed3 JW |
8485 | |
8486 | @findex CASE_DROPS_THROUGH | |
8487 | @item CASE_DROPS_THROUGH | |
8488 | Define this if control falls through a @code{case} insn when the index | |
8489 | value is out of range. This means the specified default-label is | |
8490 | actually ignored by the @code{case} insn proper. | |
8491 | ||
8492 | @findex CASE_VALUES_THRESHOLD | |
8493 | @item CASE_VALUES_THRESHOLD | |
8494 | Define this to be the smallest number of different values for which it | |
8495 | is best to use a jump-table instead of a tree of conditional branches. | |
8496 | The default is four for machines with a @code{casesi} instruction and | |
8497 | five otherwise. This is best for most machines. | |
8498 | ||
8499 | @findex WORD_REGISTER_OPERATIONS | |
8500 | @item WORD_REGISTER_OPERATIONS | |
8501 | Define this macro if operations between registers with integral mode | |
8502 | smaller than a word are always performed on the entire register. | |
8503 | Most RISC machines have this property and most CISC machines do not. | |
8504 | ||
8505 | @findex LOAD_EXTEND_OP | |
8506 | @item LOAD_EXTEND_OP (@var{mode}) | |
8507 | Define this macro to be a C expression indicating when insns that read | |
8508 | memory in @var{mode}, an integral mode narrower than a word, set the | |
8509 | bits outside of @var{mode} to be either the sign-extension or the | |
8510 | zero-extension of the data read. Return @code{SIGN_EXTEND} for values | |
8511 | of @var{mode} for which the | |
8512 | insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and | |
8513 | @code{NIL} for other modes. | |
8514 | ||
8515 | This macro is not called with @var{mode} non-integral or with a width | |
8516 | greater than or equal to @code{BITS_PER_WORD}, so you may return any | |
8517 | value in this case. Do not define this macro if it would always return | |
8518 | @code{NIL}. On machines where this macro is defined, you will normally | |
8519 | define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}. | |
8520 | ||
77643ab8 MM |
8521 | @findex SHORT_IMMEDIATES_SIGN_EXTEND |
8522 | @item SHORT_IMMEDIATES_SIGN_EXTEND | |
8523 | Define this macro if loading short immediate values into registers sign | |
8524 | extends. | |
8525 | ||
feca2ed3 JW |
8526 | @findex FIXUNS_TRUNC_LIKE_FIX_TRUNC |
8527 | @item FIXUNS_TRUNC_LIKE_FIX_TRUNC | |
8528 | Define this macro if the same instructions that convert a floating | |
8529 | point number to a signed fixed point number also convert validly to an | |
8530 | unsigned one. | |
8531 | ||
feca2ed3 JW |
8532 | @findex MOVE_MAX |
8533 | @item MOVE_MAX | |
8534 | The maximum number of bytes that a single instruction can move quickly | |
8535 | between memory and registers or between two memory locations. | |
8536 | ||
8537 | @findex MAX_MOVE_MAX | |
8538 | @item MAX_MOVE_MAX | |
8539 | The maximum number of bytes that a single instruction can move quickly | |
8540 | between memory and registers or between two memory locations. If this | |
8541 | is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the | |
8542 | constant value that is the largest value that @code{MOVE_MAX} can have | |
8543 | at run-time. | |
8544 | ||
8545 | @findex SHIFT_COUNT_TRUNCATED | |
8546 | @item SHIFT_COUNT_TRUNCATED | |
8547 | A C expression that is nonzero if on this machine the number of bits | |
8548 | actually used for the count of a shift operation is equal to the number | |
8549 | of bits needed to represent the size of the object being shifted. When | |
df2a54e9 | 8550 | this macro is nonzero, the compiler will assume that it is safe to omit |
feca2ed3 JW |
8551 | a sign-extend, zero-extend, and certain bitwise `and' instructions that |
8552 | truncates the count of a shift operation. On machines that have | |
c771326b | 8553 | instructions that act on bit-fields at variable positions, which may |
feca2ed3 JW |
8554 | include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED} |
8555 | also enables deletion of truncations of the values that serve as | |
c771326b | 8556 | arguments to bit-field instructions. |
feca2ed3 JW |
8557 | |
8558 | If both types of instructions truncate the count (for shifts) and | |
c771326b | 8559 | position (for bit-field operations), or if no variable-position bit-field |
feca2ed3 JW |
8560 | instructions exist, you should define this macro. |
8561 | ||
8562 | However, on some machines, such as the 80386 and the 680x0, truncation | |
8563 | only applies to shift operations and not the (real or pretended) | |
c771326b | 8564 | bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on |
feca2ed3 JW |
8565 | such machines. Instead, add patterns to the @file{md} file that include |
8566 | the implied truncation of the shift instructions. | |
8567 | ||
8568 | You need not define this macro if it would always have the value of zero. | |
8569 | ||
8570 | @findex TRULY_NOOP_TRUNCATION | |
8571 | @item TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec}) | |
8572 | A C expression which is nonzero if on this machine it is safe to | |
8573 | ``convert'' an integer of @var{inprec} bits to one of @var{outprec} | |
8574 | bits (where @var{outprec} is smaller than @var{inprec}) by merely | |
8575 | operating on it as if it had only @var{outprec} bits. | |
8576 | ||
8577 | On many machines, this expression can be 1. | |
8578 | ||
8579 | @c rearranged this, removed the phrase "it is reported that". this was | |
8580 | @c to fix an overfull hbox. --mew 10feb93 | |
8581 | When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for | |
8582 | modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result. | |
8583 | If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in | |
8584 | such cases may improve things. | |
8585 | ||
8586 | @findex STORE_FLAG_VALUE | |
8587 | @item STORE_FLAG_VALUE | |
8588 | A C expression describing the value returned by a comparison operator | |
8589 | with an integral mode and stored by a store-flag instruction | |
8590 | (@samp{s@var{cond}}) when the condition is true. This description must | |
8591 | apply to @emph{all} the @samp{s@var{cond}} patterns and all the | |
8592 | comparison operators whose results have a @code{MODE_INT} mode. | |
8593 | ||
630d3d5a JM |
8594 | A value of 1 or @minus{}1 means that the instruction implementing the |
8595 | comparison operator returns exactly 1 or @minus{}1 when the comparison is true | |
feca2ed3 JW |
8596 | and 0 when the comparison is false. Otherwise, the value indicates |
8597 | which bits of the result are guaranteed to be 1 when the comparison is | |
8598 | true. This value is interpreted in the mode of the comparison | |
8599 | operation, which is given by the mode of the first operand in the | |
8600 | @samp{s@var{cond}} pattern. Either the low bit or the sign bit of | |
8601 | @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by | |
8602 | the compiler. | |
8603 | ||
630d3d5a | 8604 | If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will |
feca2ed3 JW |
8605 | generate code that depends only on the specified bits. It can also |
8606 | replace comparison operators with equivalent operations if they cause | |
8607 | the required bits to be set, even if the remaining bits are undefined. | |
8608 | For example, on a machine whose comparison operators return an | |
8609 | @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as | |
8610 | @samp{0x80000000}, saying that just the sign bit is relevant, the | |
8611 | expression | |
8612 | ||
8613 | @smallexample | |
8614 | (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0)) | |
8615 | @end smallexample | |
8616 | ||
8617 | @noindent | |
8618 | can be converted to | |
8619 | ||
8620 | @smallexample | |
8621 | (ashift:SI @var{x} (const_int @var{n})) | |
8622 | @end smallexample | |
8623 | ||
8624 | @noindent | |
8625 | where @var{n} is the appropriate shift count to move the bit being | |
8626 | tested into the sign bit. | |
8627 | ||
8628 | There is no way to describe a machine that always sets the low-order bit | |
8629 | for a true value, but does not guarantee the value of any other bits, | |
8630 | but we do not know of any machine that has such an instruction. If you | |
a3a15b4d | 8631 | are trying to port GCC to such a machine, include an instruction to |
feca2ed3 | 8632 | perform a logical-and of the result with 1 in the pattern for the |
b11cc610 | 8633 | comparison operators and let us know at @email{gcc@@gcc.gnu.org}. |
feca2ed3 JW |
8634 | |
8635 | Often, a machine will have multiple instructions that obtain a value | |
8636 | from a comparison (or the condition codes). Here are rules to guide the | |
8637 | choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions | |
8638 | to be used: | |
8639 | ||
8640 | @itemize @bullet | |
8641 | @item | |
8642 | Use the shortest sequence that yields a valid definition for | |
8643 | @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to | |
8644 | ``normalize'' the value (convert it to, e.g., 1 or 0) than for the | |
8645 | comparison operators to do so because there may be opportunities to | |
8646 | combine the normalization with other operations. | |
8647 | ||
8648 | @item | |
630d3d5a | 8649 | For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being |
feca2ed3 JW |
8650 | slightly preferred on machines with expensive jumps and 1 preferred on |
8651 | other machines. | |
8652 | ||
8653 | @item | |
8654 | As a second choice, choose a value of @samp{0x80000001} if instructions | |
8655 | exist that set both the sign and low-order bits but do not define the | |
8656 | others. | |
8657 | ||
8658 | @item | |
8659 | Otherwise, use a value of @samp{0x80000000}. | |
8660 | @end itemize | |
8661 | ||
8662 | Many machines can produce both the value chosen for | |
8663 | @code{STORE_FLAG_VALUE} and its negation in the same number of | |
8664 | instructions. On those machines, you should also define a pattern for | |
8665 | those cases, e.g., one matching | |
8666 | ||
8667 | @smallexample | |
8668 | (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C}))) | |
8669 | @end smallexample | |
8670 | ||
8671 | Some machines can also perform @code{and} or @code{plus} operations on | |
8672 | condition code values with less instructions than the corresponding | |
8673 | @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those | |
8674 | machines, define the appropriate patterns. Use the names @code{incscc} | |
8675 | and @code{decscc}, respectively, for the patterns which perform | |
8676 | @code{plus} or @code{minus} operations on condition code values. See | |
8677 | @file{rs6000.md} for some examples. The GNU Superoptizer can be used to | |
8678 | find such instruction sequences on other machines. | |
8679 | ||
8680 | You need not define @code{STORE_FLAG_VALUE} if the machine has no store-flag | |
8681 | instructions. | |
8682 | ||
8683 | @findex FLOAT_STORE_FLAG_VALUE | |
12530dbe | 8684 | @item FLOAT_STORE_FLAG_VALUE (@var{mode}) |
df2a54e9 | 8685 | A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is |
feca2ed3 JW |
8686 | returned when comparison operators with floating-point results are true. |
8687 | Define this macro on machine that have comparison operations that return | |
8688 | floating-point values. If there are no such operations, do not define | |
8689 | this macro. | |
8690 | ||
8691 | @findex Pmode | |
8692 | @item Pmode | |
8693 | An alias for the machine mode for pointers. On most machines, define | |
8694 | this to be the integer mode corresponding to the width of a hardware | |
8695 | pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines. | |
8696 | On some machines you must define this to be one of the partial integer | |
8697 | modes, such as @code{PSImode}. | |
8698 | ||
8699 | The width of @code{Pmode} must be at least as large as the value of | |
8700 | @code{POINTER_SIZE}. If it is not equal, you must define the macro | |
8701 | @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended | |
8702 | to @code{Pmode}. | |
8703 | ||
8704 | @findex FUNCTION_MODE | |
8705 | @item FUNCTION_MODE | |
8706 | An alias for the machine mode used for memory references to functions | |
8707 | being called, in @code{call} RTL expressions. On most machines this | |
8708 | should be @code{QImode}. | |
8709 | ||
8710 | @findex INTEGRATE_THRESHOLD | |
8711 | @item INTEGRATE_THRESHOLD (@var{decl}) | |
8712 | A C expression for the maximum number of instructions above which the | |
8713 | function @var{decl} should not be inlined. @var{decl} is a | |
8714 | @code{FUNCTION_DECL} node. | |
8715 | ||
8716 | The default definition of this macro is 64 plus 8 times the number of | |
8717 | arguments that the function accepts. Some people think a larger | |
8718 | threshold should be used on RISC machines. | |
8719 | ||
ee773fcc NB |
8720 | @findex STDC_0_IN_SYSTEM_HEADERS |
8721 | @item STDC_0_IN_SYSTEM_HEADERS | |
8722 | In normal operation, the preprocessor expands @code{__STDC__} to the | |
8723 | constant 1, to signify that GCC conforms to ISO Standard C@. On some | |
8724 | hosts, like Solaris, the system compiler uses a different convention, | |
8725 | where @code{__STDC__} is normally 0, but is 1 if the user specifies | |
8726 | strict conformance to the C Standard. | |
8727 | ||
8728 | Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host | |
8729 | convention when processing system header files, but when processing user | |
8730 | files @code{__STDC__} will always expand to 1. | |
8731 | ||
feca2ed3 JW |
8732 | @findex NO_IMPLICIT_EXTERN_C |
8733 | @item NO_IMPLICIT_EXTERN_C | |
161d7b59 | 8734 | Define this macro if the system header files support C++ as well as C@. |
feca2ed3 JW |
8735 | This macro inhibits the usual method of using system header files in |
8736 | C++, which is to pretend that the file's contents are enclosed in | |
8737 | @samp{extern "C" @{@dots{}@}}. | |
8738 | ||
8739 | @findex HANDLE_PRAGMA | |
8b97c5f8 ZW |
8740 | @item HANDLE_PRAGMA (@var{getc}, @var{ungetc}, @var{name}) |
8741 | This macro is no longer supported. You must use | |
8742 | @code{REGISTER_TARGET_PRAGMAS} instead. | |
8743 | ||
8744 | @findex REGISTER_TARGET_PRAGMAS | |
feca2ed3 JW |
8745 | @findex #pragma |
8746 | @findex pragma | |
8b97c5f8 ZW |
8747 | @item REGISTER_TARGET_PRAGMAS (@var{pfile}) |
8748 | Define this macro if you want to implement any target-specific pragmas. | |
a5da89c6 NB |
8749 | If defined, it is a C expression which makes a series of calls to |
8750 | @code{cpp_register_pragma} for each pragma, with @var{pfile} passed as | |
8751 | the first argument to to these functions. The macro may also do any | |
8752 | setup required for the pragmas. | |
8b97c5f8 ZW |
8753 | |
8754 | The primary reason to define this macro is to provide compatibility with | |
8755 | other compilers for the same target. In general, we discourage | |
161d7b59 | 8756 | definition of target-specific pragmas for GCC@. |
feca2ed3 | 8757 | |
c237e94a | 8758 | If the pragma can be implemented by attributes then you should consider |
91d231cb | 8759 | defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well. |
f09db6e0 | 8760 | |
8b97c5f8 ZW |
8761 | Preprocessor macros that appear on pragma lines are not expanded. All |
8762 | @samp{#pragma} directives that do not match any registered pragma are | |
630d3d5a | 8763 | silently ignored, unless the user specifies @option{-Wunknown-pragmas}. |
8b97c5f8 ZW |
8764 | |
8765 | @deftypefun void cpp_register_pragma (cpp_reader *@var{pfile}, const char *@var{space}, const char *@var{name}, void (*@var{callback}) (cpp_reader *)) | |
8766 | ||
8767 | Each call to @code{cpp_register_pragma} establishes one pragma. The | |
8768 | @var{callback} routine will be called when the preprocessor encounters a | |
8769 | pragma of the form | |
8770 | ||
8771 | @smallexample | |
8772 | #pragma [@var{space}] @var{name} @dots{} | |
8773 | @end smallexample | |
8774 | ||
a5da89c6 NB |
8775 | @var{space} is the case-sensitive namespace of the pragma, or |
8776 | @code{NULL} to put the pragma in the global namespace. The callback | |
8777 | routine receives @var{pfile} as its first argument, which can be passed | |
51fabca5 NB |
8778 | on to cpplib's functions if necessary. You can lex tokens after the |
8779 | @var{name} by calling @code{c_lex}. Tokens that are not read by the | |
8780 | callback will be silently ignored. The end of the line is indicated by | |
8781 | a token of type @code{CPP_EOF}. | |
8b97c5f8 ZW |
8782 | |
8783 | For an example use of this routine, see @file{c4x.h} and the callback | |
51fabca5 | 8784 | routines defined in @file{c4x-c.c}. |
aac69a49 NC |
8785 | |
8786 | Note that the use of @code{c_lex} is specific to the C and C++ | |
8787 | compilers. It will not work in the Java or Fortran compilers, or any | |
8788 | other language compilers for that matter. Thus if @code{c_lex} is going | |
8789 | to be called from target-specific code, it must only be done so when | |
c771326b | 8790 | building the C and C++ compilers. This can be done by defining the |
aac69a49 | 8791 | variables @code{c_target_objs} and @code{cxx_target_objs} in the |
aee96fe9 | 8792 | target entry in the @file{config.gcc} file. These variables should name |
aac69a49 NC |
8793 | the target-specific, language-specific object file which contains the |
8794 | code that uses @code{c_lex}. Note it will also be necessary to add a | |
8795 | rule to the makefile fragment pointed to by @code{tmake_file} that shows | |
8796 | how to build this object file. | |
8b97c5f8 ZW |
8797 | @end deftypefun |
8798 | ||
e2af664c NC |
8799 | @findex HANDLE_SYSV_PRAGMA |
8800 | @findex #pragma | |
8801 | @findex pragma | |
8802 | @item HANDLE_SYSV_PRAGMA | |
8803 | Define this macro (to a value of 1) if you want the System V style | |
8804 | pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name> | |
8805 | [=<value>]} to be supported by gcc. | |
8806 | ||
8807 | The pack pragma specifies the maximum alignment (in bytes) of fields | |
8808 | within a structure, in much the same way as the @samp{__aligned__} and | |
8809 | @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets | |
c21cd8b1 | 8810 | the behavior to the default. |
e2af664c | 8811 | |
e4850f36 DR |
8812 | A subtlety for Microsoft Visual C/C++ style bit-field packing |
8813 | (e.g. -mms-bitfields) for targets that support it: | |
8814 | When a bit-field is inserted into a packed record, the whole size | |
8815 | of the underlying type is used by one or more same-size adjacent | |
8816 | bit-fields (that is, if its long:3, 32 bits is used in the record, | |
8817 | and any additional adjacent long bit-fields are packed into the same | |
8818 | chunk of 32 bits. However, if the size changes, a new field of that | |
8819 | size is allocated). | |
8820 | ||
8821 | If both MS bit-fields and @samp{__attribute__((packed))} are used, | |
8822 | the latter will take precedence. If @samp{__attribute__((packed))} is | |
8823 | used on a single field when MS bit-fields are in use, it will take | |
8824 | precedence for that field, but the alignment of the rest of the structure | |
8825 | may affect its placement. | |
8826 | ||
e2af664c NC |
8827 | The weak pragma only works if @code{SUPPORTS_WEAK} and |
8828 | @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation | |
8829 | of specifically named weak labels, optionally with a value. | |
8830 | ||
8831 | @findex HANDLE_PRAGMA_PACK_PUSH_POP | |
8832 | @findex #pragma | |
8833 | @findex pragma | |
8834 | @item HANDLE_PRAGMA_PACK_PUSH_POP | |
8835 | Define this macro (to a value of 1) if you want to support the Win32 | |
aee96fe9 JM |
8836 | style pragmas @samp{#pragma pack(push,@var{n})} and @samp{#pragma |
8837 | pack(pop)}. The @samp{pack(push,@var{n})} pragma specifies the maximum alignment | |
e2af664c NC |
8838 | (in bytes) of fields within a structure, in much the same way as the |
8839 | @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A | |
c21cd8b1 | 8840 | pack value of zero resets the behavior to the default. Successive |
e2af664c NC |
8841 | invocations of this pragma cause the previous values to be stacked, so |
8842 | that invocations of @samp{#pragma pack(pop)} will return to the previous | |
8843 | value. | |
feca2ed3 JW |
8844 | |
8845 | @findex DOLLARS_IN_IDENTIFIERS | |
8846 | @item DOLLARS_IN_IDENTIFIERS | |
8847 | Define this macro to control use of the character @samp{$} in identifier | |
37d13a29 | 8848 | names. 0 means @samp{$} is not allowed by default; 1 means it is allowed. |
feca2ed3 | 8849 | 1 is the default; there is no need to define this macro in that case. |
37d13a29 | 8850 | This macro controls the compiler proper; it does not affect the preprocessor. |
feca2ed3 JW |
8851 | |
8852 | @findex NO_DOLLAR_IN_LABEL | |
8853 | @item NO_DOLLAR_IN_LABEL | |
8854 | Define this macro if the assembler does not accept the character | |
8855 | @samp{$} in label names. By default constructors and destructors in | |
8856 | G++ have @samp{$} in the identifiers. If this macro is defined, | |
8857 | @samp{.} is used instead. | |
8858 | ||
8859 | @findex NO_DOT_IN_LABEL | |
8860 | @item NO_DOT_IN_LABEL | |
8861 | Define this macro if the assembler does not accept the character | |
8862 | @samp{.} in label names. By default constructors and destructors in G++ | |
8863 | have names that use @samp{.}. If this macro is defined, these names | |
8864 | are rewritten to avoid @samp{.}. | |
8865 | ||
8866 | @findex DEFAULT_MAIN_RETURN | |
8867 | @item DEFAULT_MAIN_RETURN | |
8868 | Define this macro if the target system expects every program's @code{main} | |
8869 | function to return a standard ``success'' value by default (if no other | |
8870 | value is explicitly returned). | |
8871 | ||
8872 | The definition should be a C statement (sans semicolon) to generate the | |
8873 | appropriate rtl instructions. It is used only when compiling the end of | |
8874 | @code{main}. | |
8875 | ||
c063dc98 JM |
8876 | @item NEED_ATEXIT |
8877 | @findex NEED_ATEXIT | |
8878 | Define this if the target system lacks the function @code{atexit} | |
5490d604 | 8879 | from the ISO C standard. If this macro is defined, a default definition |
c063dc98 JM |
8880 | will be provided to support C++. If @code{ON_EXIT} is not defined, |
8881 | a default @code{exit} function will also be provided. | |
8882 | ||
8883 | @item ON_EXIT | |
8884 | @findex ON_EXIT | |
8885 | Define this macro if the target has another way to implement atexit | |
8886 | functionality without replacing @code{exit}. For instance, SunOS 4 has | |
8887 | a similar @code{on_exit} library function. | |
8888 | ||
8889 | The definition should be a functional macro which can be used just like | |
8890 | the @code{atexit} function. | |
feca2ed3 JW |
8891 | |
8892 | @item EXIT_BODY | |
8893 | @findex EXIT_BODY | |
8894 | Define this if your @code{exit} function needs to do something | |
8895 | besides calling an external function @code{_cleanup} before | |
8896 | terminating with @code{_exit}. The @code{EXIT_BODY} macro is | |
9e9b9afe JM |
8897 | only needed if @code{NEED_ATEXIT} is defined and @code{ON_EXIT} is not |
8898 | defined. | |
feca2ed3 JW |
8899 | |
8900 | @findex INSN_SETS_ARE_DELAYED | |
8901 | @item INSN_SETS_ARE_DELAYED (@var{insn}) | |
8902 | Define this macro as a C expression that is nonzero if it is safe for the | |
8903 | delay slot scheduler to place instructions in the delay slot of @var{insn}, | |
8904 | even if they appear to use a resource set or clobbered in @var{insn}. | |
a3a15b4d | 8905 | @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that |
feca2ed3 JW |
8906 | every @code{call_insn} has this behavior. On machines where some @code{insn} |
8907 | or @code{jump_insn} is really a function call and hence has this behavior, | |
8908 | you should define this macro. | |
8909 | ||
8910 | You need not define this macro if it would always return zero. | |
8911 | ||
8912 | @findex INSN_REFERENCES_ARE_DELAYED | |
8913 | @item INSN_REFERENCES_ARE_DELAYED (@var{insn}) | |
8914 | Define this macro as a C expression that is nonzero if it is safe for the | |
8915 | delay slot scheduler to place instructions in the delay slot of @var{insn}, | |
8916 | even if they appear to set or clobber a resource referenced in @var{insn}. | |
8917 | @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where | |
8918 | some @code{insn} or @code{jump_insn} is really a function call and its operands | |
8919 | are registers whose use is actually in the subroutine it calls, you should | |
8920 | define this macro. Doing so allows the delay slot scheduler to move | |
8921 | instructions which copy arguments into the argument registers into the delay | |
8922 | slot of @var{insn}. | |
8923 | ||
8924 | You need not define this macro if it would always return zero. | |
8925 | ||
8926 | @findex MACHINE_DEPENDENT_REORG | |
8927 | @item MACHINE_DEPENDENT_REORG (@var{insn}) | |
8928 | In rare cases, correct code generation requires extra machine | |
8929 | dependent processing between the second jump optimization pass and | |
8930 | delayed branch scheduling. On those machines, define this macro as a C | |
8931 | statement to act on the code starting at @var{insn}. | |
8932 | ||
861bb6c1 JL |
8933 | @findex MULTIPLE_SYMBOL_SPACES |
8934 | @item MULTIPLE_SYMBOL_SPACES | |
8935 | Define this macro if in some cases global symbols from one translation | |
8936 | unit may not be bound to undefined symbols in another translation unit | |
8937 | without user intervention. For instance, under Microsoft Windows | |
8938 | symbols must be explicitly imported from shared libraries (DLLs). | |
8939 | ||
57bcb97a | 8940 | @findex MD_ASM_CLOBBERS |
aee96fe9 JM |
8941 | @item MD_ASM_CLOBBERS (@var{clobbers}) |
8942 | A C statement that adds to @var{clobbers} @code{STRING_CST} trees for | |
57bcb97a RH |
8943 | any hard regs the port wishes to automatically clobber for all asms. |
8944 | ||
dbecbbe4 JL |
8945 | @findex MAX_INTEGER_COMPUTATION_MODE |
8946 | @item MAX_INTEGER_COMPUTATION_MODE | |
8947 | Define this to the largest integer machine mode which can be used for | |
8948 | operations other than load, store and copy operations. | |
8949 | ||
8950 | You need only define this macro if the target holds values larger than | |
8951 | @code{word_mode} in general purpose registers. Most targets should not define | |
8952 | this macro. | |
f89223a9 | 8953 | |
71d718e0 JM |
8954 | @findex MATH_LIBRARY |
8955 | @item MATH_LIBRARY | |
8956 | Define this macro as a C string constant for the linker argument to link | |
8957 | in the system math library, or @samp{""} if the target does not have a | |
8958 | separate math library. | |
8959 | ||
8960 | You need only define this macro if the default of @samp{"-lm"} is wrong. | |
512b62fb JM |
8961 | |
8962 | @findex LIBRARY_PATH_ENV | |
8963 | @item LIBRARY_PATH_ENV | |
8964 | Define this macro as a C string constant for the environment variable that | |
8965 | specifies where the linker should look for libraries. | |
8966 | ||
8967 | You need only define this macro if the default of @samp{"LIBRARY_PATH"} | |
8968 | is wrong. | |
e09d24ff R |
8969 | |
8970 | @findex TARGET_HAS_F_SETLKW | |
8971 | @item TARGET_HAS_F_SETLKW | |
161d7b59 JM |
8972 | Define this macro if the target supports file locking with fcntl / F_SETLKW@. |
8973 | Note that this functionality is part of POSIX@. | |
e09d24ff R |
8974 | Defining @code{TARGET_HAS_F_SETLKW} will enable the test coverage code |
8975 | to use file locking when exiting a program, which avoids race conditions | |
8976 | if the program has forked. | |
0c99ec5c RH |
8977 | |
8978 | @findex MAX_CONDITIONAL_EXECUTE | |
8979 | @item MAX_CONDITIONAL_EXECUTE | |
8980 | ||
8981 | A C expression for the maximum number of instructions to execute via | |
8982 | conditional execution instructions instead of a branch. A value of | |
8983 | @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and | |
8984 | 1 if it does use cc0. | |
90280148 MM |
8985 | |
8986 | @findex IFCVT_MODIFY_TESTS | |
c05ffc49 BS |
8987 | @item IFCVT_MODIFY_TESTS(@var{ce_info}, @var{true_expr}, @var{false_expr}) |
8988 | Used if the target needs to perform machine-dependent modifications on the | |
8989 | conditionals used for turning basic blocks into conditionally executed code. | |
8990 | @var{ce_info} points to a data structure, @code{struct ce_if_block}, which | |
8991 | contains information about the currently processed blocks. @var{true_expr} | |
8992 | and @var{false_expr} are the tests that are used for converting the | |
8993 | then-block and the else-block, respectively. Set either @var{true_expr} or | |
8994 | @var{false_expr} to a null pointer if the tests cannot be converted. | |
8995 | ||
8996 | @findex IFCVT_MODIFY_MULTIPLE_TESTS | |
8997 | @item IFCVT_MODIFY_MULTIPLE_TESTS(@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr}) | |
8998 | Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated | |
8999 | if-statements into conditions combined by @code{and} and @code{or} operations. | |
9000 | @var{bb} contains the basic block that contains the test that is currently | |
9001 | being processed and about to be turned into a condition. | |
90280148 MM |
9002 | |
9003 | @findex IFCVT_MODIFY_INSN | |
c05ffc49 BS |
9004 | @item IFCVT_MODIFY_INSN(@var{ce_info}, @var{pattern}, @var{insn}) |
9005 | A C expression to modify the @var{PATTERN} of an @var{INSN} that is to | |
9006 | be converted to conditional execution format. @var{ce_info} points to | |
9007 | a data structure, @code{struct ce_if_block}, which contains information | |
9008 | about the currently processed blocks. | |
90280148 MM |
9009 | |
9010 | @findex IFCVT_MODIFY_FINAL | |
c05ffc49 | 9011 | @item IFCVT_MODIFY_FINAL(@var{ce_info}) |
90280148 | 9012 | A C expression to perform any final machine dependent modifications in |
c05ffc49 BS |
9013 | converting code to conditional execution. The involved basic blocks |
9014 | can be found in the @code{struct ce_if_block} structure that is pointed | |
9015 | to by @var{ce_info}. | |
90280148 MM |
9016 | |
9017 | @findex IFCVT_MODIFY_CANCEL | |
c05ffc49 | 9018 | @item IFCVT_MODIFY_CANCEL(@var{ce_info}) |
90280148 | 9019 | A C expression to cancel any machine dependent modifications in |
c05ffc49 BS |
9020 | converting code to conditional execution. The involved basic blocks |
9021 | can be found in the @code{struct ce_if_block} structure that is pointed | |
9022 | to by @var{ce_info}. | |
9023 | ||
9024 | @findex IFCVT_INIT_EXTRA_FIELDS | |
9025 | @item IFCVT_INIT_EXTRA_FIELDS(@var{ce_info}) | |
9026 | A C expression to initialize any extra fields in a @code{struct ce_if_block} | |
9027 | structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro. | |
9028 | ||
9029 | @findex IFCVT_EXTRA_FIELDS | |
9030 | @item IFCVT_EXTRA_FIELDS | |
9031 | If defined, it should expand to a set of field declarations that will be | |
9032 | added to the @code{struct ce_if_block} structure. These should be intialized | |
9033 | by the @code{IFCVT_INIT_EXTRA_FIELDS} macro. | |
9034 | ||
c237e94a | 9035 | @end table |
4a1d48f6 | 9036 | |
f6155fda SS |
9037 | @deftypefn {Target Hook} void TARGET_INIT_BUILTINS () |
9038 | Define this hook if you have any machine-specific built-in functions | |
9039 | that need to be defined. It should be a function that performs the | |
4a1d48f6 BS |
9040 | necessary setup. |
9041 | ||
c771326b | 9042 | Machine specific built-in functions can be useful to expand special machine |
4a1d48f6 BS |
9043 | instructions that would otherwise not normally be generated because |
9044 | they have no equivalent in the source language (for example, SIMD vector | |
9045 | instructions or prefetch instructions). | |
9046 | ||
c771326b JM |
9047 | To create a built-in function, call the function @code{builtin_function} |
9048 | which is defined by the language front end. You can use any type nodes set | |
4a1d48f6 | 9049 | up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2}; |
c237e94a | 9050 | only language front ends that use those two functions will call |
f6155fda | 9051 | @samp{TARGET_INIT_BUILTINS}. |
acdcefcc | 9052 | @end deftypefn |
4a1d48f6 | 9053 | |
f6155fda | 9054 | @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore}) |
4a1d48f6 | 9055 | |
c771326b | 9056 | Expand a call to a machine specific built-in function that was set up by |
f6155fda SS |
9057 | @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the |
9058 | function call; the result should go to @var{target} if that is | |
9059 | convenient, and have mode @var{mode} if that is convenient. | |
9060 | @var{subtarget} may be used as the target for computing one of | |
9061 | @var{exp}'s operands. @var{ignore} is nonzero if the value is to be | |
9062 | ignored. This function should return the result of the call to the | |
9063 | built-in function. | |
acdcefcc | 9064 | @end deftypefn |
4a1d48f6 | 9065 | |
c237e94a | 9066 | @table @code |
6e7b03e1 AH |
9067 | @findex MD_CAN_REDIRECT_BRANCH |
9068 | @item MD_CAN_REDIRECT_BRANCH(@var{branch1}, @var{branch2}) | |
9069 | ||
4fe9b91c | 9070 | Take a branch insn in @var{branch1} and another in @var{branch2}. |
6e7b03e1 AH |
9071 | Return true if redirecting @var{branch1} to the destination of |
9072 | @var{branch2} is possible. | |
9073 | ||
9074 | On some targets, branches may have a limited range. Optimizing the | |
9075 | filling of delay slots can result in branches being redirected, and this | |
9076 | may in turn cause a branch offset to overflow. | |
9077 | ||
385b6e2d R |
9078 | @findex ALLOCATE_INITIAL_VALUE |
9079 | @item ALLOCATE_INITIAL_VALUE(@var{hard_reg}) | |
9080 | ||
9081 | When the initial value of a hard register has been copied in a pseudo | |
4fe9b91c | 9082 | register, it is often not necessary to actually allocate another register |
385b6e2d R |
9083 | to this pseudo register, because the original hard register or a stack slot |
9084 | it has been saved into can be used. @code{ALLOCATE_INITIAL_VALUE}, if | |
9085 | defined, is called at the start of register allocation once for each | |
9086 | hard register that had its initial value copied by using | |
9087 | @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}. | |
9088 | Possible values are @code{NULL_RTX}, if you don't want | |
9089 | to do any special allocation, a @code{REG} rtx---that would typically be | |
9090 | the hard register itself, if it is known not to be clobbered---or a | |
9091 | @code{MEM}. | |
9092 | If you are returning a @code{MEM}, this is only a hint for the allocator; | |
9093 | it might decide to use another register anyways. | |
9094 | You may use @code{current_function_leaf_function} in the definition of the | |
9095 | macro, functions that use @code{REG_N_SETS}, to determine if the hard | |
9096 | register in question will not be clobbered. | |
9097 | ||
807633e5 ZW |
9098 | @findex TARGET_OBJECT_SUFFIX |
9099 | @item TARGET_OBJECT_SUFFIX | |
9100 | Define this macro to be a C string representing the suffix for object | |
9101 | files on your target machine. If you do not define this macro, GCC will | |
9102 | use @samp{.o} as the suffix for object files. | |
9103 | ||
9104 | @findex TARGET_EXECUTABLE_SUFFIX | |
9105 | @item TARGET_EXECUTABLE_SUFFIX | |
9106 | Define this macro to be a C string representing the suffix to be | |
9107 | automatically added to executable files on your target machine. If you | |
9108 | do not define this macro, GCC will use the null string as the suffix for | |
9109 | executable files. | |
9110 | ||
9111 | @findex COLLECT_EXPORT_LIST | |
9112 | @item COLLECT_EXPORT_LIST | |
9113 | If defined, @code{collect2} will scan the individual object files | |
9114 | specified on its command line and create an export list for the linker. | |
9115 | Define this macro for systems like AIX, where the linker discards | |
9116 | object files that are not referenced from @code{main} and uses export | |
9117 | lists. | |
9118 | ||
feca2ed3 | 9119 | @end table |
e4ec2cac AO |
9120 | |
9121 | @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void) | |
9122 | This target hook returns @code{true} past the point in which new jump | |
9123 | instructions could be created. On machines that require a register for | |
9124 | every jump such as the SHmedia ISA of SH5, this point would typically be | |
9125 | reload, so this target hook should be defined to a function such as: | |
9126 | ||
9127 | @smallexample | |
9128 | static bool | |
9129 | cannot_modify_jumps_past_reload_p () | |
9130 | @{ | |
9131 | return (reload_completed || reload_in_progress); | |
9132 | @} | |
9133 | @end smallexample | |
9134 | @end deftypefn |