Commit | Line | Data |
---|---|---|
aa820537 AM |
1 | @c Copyright 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005, |
2 | @c 2006, 2007, 2008, 2009 Free Software Foundation, Inc. | |
252b5132 RH |
3 | @c This is part of the GAS manual. |
4 | @c For copying conditions, see the file as.texinfo. | |
5 | ||
6 | @ifset GENERIC | |
7 | @page | |
8 | @node ARM-Dependent | |
9 | @chapter ARM Dependent Features | |
10 | @end ifset | |
11 | ||
12 | @ifclear GENERIC | |
13 | @node Machine Dependencies | |
14 | @chapter ARM Dependent Features | |
15 | @end ifclear | |
16 | ||
17 | @cindex ARM support | |
18 | @cindex Thumb support | |
19 | @menu | |
20 | * ARM Options:: Options | |
21 | * ARM Syntax:: Syntax | |
22 | * ARM Floating Point:: Floating Point | |
23 | * ARM Directives:: ARM Machine Directives | |
24 | * ARM Opcodes:: Opcodes | |
6057a28f | 25 | * ARM Mapping Symbols:: Mapping Symbols |
7da4f750 | 26 | * ARM Unwinding Tutorial:: Unwinding |
252b5132 RH |
27 | @end menu |
28 | ||
29 | @node ARM Options | |
30 | @section Options | |
31 | @cindex ARM options (none) | |
32 | @cindex options for ARM (none) | |
adcf07e6 | 33 | |
252b5132 | 34 | @table @code |
adcf07e6 | 35 | |
03b1477f | 36 | @cindex @code{-mcpu=} command line option, ARM |
92081f48 | 37 | @item -mcpu=@var{processor}[+@var{extension}@dots{}] |
252b5132 RH |
38 | This option specifies the target processor. The assembler will issue an |
39 | error message if an attempt is made to assemble an instruction which | |
03b1477f RE |
40 | will not execute on the target processor. The following processor names are |
41 | recognized: | |
42 | @code{arm1}, | |
43 | @code{arm2}, | |
44 | @code{arm250}, | |
45 | @code{arm3}, | |
46 | @code{arm6}, | |
47 | @code{arm60}, | |
48 | @code{arm600}, | |
49 | @code{arm610}, | |
50 | @code{arm620}, | |
51 | @code{arm7}, | |
52 | @code{arm7m}, | |
53 | @code{arm7d}, | |
54 | @code{arm7dm}, | |
55 | @code{arm7di}, | |
56 | @code{arm7dmi}, | |
57 | @code{arm70}, | |
58 | @code{arm700}, | |
59 | @code{arm700i}, | |
60 | @code{arm710}, | |
61 | @code{arm710t}, | |
62 | @code{arm720}, | |
63 | @code{arm720t}, | |
64 | @code{arm740t}, | |
65 | @code{arm710c}, | |
66 | @code{arm7100}, | |
67 | @code{arm7500}, | |
68 | @code{arm7500fe}, | |
69 | @code{arm7t}, | |
70 | @code{arm7tdmi}, | |
1ff4677c | 71 | @code{arm7tdmi-s}, |
03b1477f RE |
72 | @code{arm8}, |
73 | @code{arm810}, | |
74 | @code{strongarm}, | |
75 | @code{strongarm1}, | |
76 | @code{strongarm110}, | |
77 | @code{strongarm1100}, | |
78 | @code{strongarm1110}, | |
79 | @code{arm9}, | |
80 | @code{arm920}, | |
81 | @code{arm920t}, | |
82 | @code{arm922t}, | |
83 | @code{arm940t}, | |
84 | @code{arm9tdmi}, | |
7fac0536 NC |
85 | @code{fa526} (Faraday FA526 processor), |
86 | @code{fa626} (Faraday FA626 processor), | |
03b1477f | 87 | @code{arm9e}, |
7de9afa2 | 88 | @code{arm926e}, |
1ff4677c | 89 | @code{arm926ej-s}, |
03b1477f RE |
90 | @code{arm946e-r0}, |
91 | @code{arm946e}, | |
db8ac8f9 | 92 | @code{arm946e-s}, |
03b1477f RE |
93 | @code{arm966e-r0}, |
94 | @code{arm966e}, | |
db8ac8f9 PB |
95 | @code{arm966e-s}, |
96 | @code{arm968e-s}, | |
03b1477f | 97 | @code{arm10t}, |
db8ac8f9 | 98 | @code{arm10tdmi}, |
03b1477f RE |
99 | @code{arm10e}, |
100 | @code{arm1020}, | |
101 | @code{arm1020t}, | |
7de9afa2 | 102 | @code{arm1020e}, |
db8ac8f9 | 103 | @code{arm1022e}, |
1ff4677c | 104 | @code{arm1026ej-s}, |
7fac0536 NC |
105 | @code{fa626te} (Faraday FA626TE processor), |
106 | @code{fa726te} (Faraday FA726TE processor), | |
1ff4677c RE |
107 | @code{arm1136j-s}, |
108 | @code{arm1136jf-s}, | |
db8ac8f9 PB |
109 | @code{arm1156t2-s}, |
110 | @code{arm1156t2f-s}, | |
0dd132b6 NC |
111 | @code{arm1176jz-s}, |
112 | @code{arm1176jzf-s}, | |
113 | @code{mpcore}, | |
114 | @code{mpcorenovfp}, | |
b38f9f31 | 115 | @code{cortex-a5}, |
62b3e311 | 116 | @code{cortex-a8}, |
15290f0a | 117 | @code{cortex-a9}, |
62b3e311 | 118 | @code{cortex-r4}, |
307c948d | 119 | @code{cortex-r4f}, |
62b3e311 | 120 | @code{cortex-m3}, |
5b19eaba NC |
121 | @code{cortex-m1}, |
122 | @code{cortex-m0}, | |
03b1477f RE |
123 | @code{ep9312} (ARM920 with Cirrus Maverick coprocessor), |
124 | @code{i80200} (Intel XScale processor) | |
e16bb312 | 125 | @code{iwmmxt} (Intel(r) XScale processor with Wireless MMX(tm) technology coprocessor) |
03b1477f RE |
126 | and |
127 | @code{xscale}. | |
128 | The special name @code{all} may be used to allow the | |
129 | assembler to accept instructions valid for any ARM processor. | |
130 | ||
131 | In addition to the basic instruction set, the assembler can be told to | |
132 | accept various extension mnemonics that extend the processor using the | |
133 | co-processor instruction space. For example, @code{-mcpu=arm920+maverick} | |
134 | is equivalent to specifying @code{-mcpu=ep9312}. The following extensions | |
135 | are currently supported: | |
136 | @code{+maverick} | |
e16bb312 | 137 | @code{+iwmmxt} |
03b1477f RE |
138 | and |
139 | @code{+xscale}. | |
140 | ||
141 | @cindex @code{-march=} command line option, ARM | |
92081f48 | 142 | @item -march=@var{architecture}[+@var{extension}@dots{}] |
252b5132 RH |
143 | This option specifies the target architecture. The assembler will issue |
144 | an error message if an attempt is made to assemble an instruction which | |
03b1477f RE |
145 | will not execute on the target architecture. The following architecture |
146 | names are recognized: | |
147 | @code{armv1}, | |
148 | @code{armv2}, | |
149 | @code{armv2a}, | |
150 | @code{armv2s}, | |
151 | @code{armv3}, | |
152 | @code{armv3m}, | |
153 | @code{armv4}, | |
154 | @code{armv4xm}, | |
155 | @code{armv4t}, | |
156 | @code{armv4txm}, | |
157 | @code{armv5}, | |
158 | @code{armv5t}, | |
159 | @code{armv5txm}, | |
160 | @code{armv5te}, | |
09d92015 | 161 | @code{armv5texp}, |
c5f98204 | 162 | @code{armv6}, |
1ddd7f43 | 163 | @code{armv6j}, |
0dd132b6 NC |
164 | @code{armv6k}, |
165 | @code{armv6z}, | |
166 | @code{armv6zk}, | |
62b3e311 | 167 | @code{armv7}, |
c450d570 PB |
168 | @code{armv7-a}, |
169 | @code{armv7-r}, | |
170 | @code{armv7-m}, | |
e16bb312 | 171 | @code{iwmmxt} |
03b1477f RE |
172 | and |
173 | @code{xscale}. | |
174 | If both @code{-mcpu} and | |
175 | @code{-march} are specified, the assembler will use | |
176 | the setting for @code{-mcpu}. | |
177 | ||
178 | The architecture option can be extended with the same instruction set | |
179 | extension options as the @code{-mcpu} option. | |
180 | ||
181 | @cindex @code{-mfpu=} command line option, ARM | |
182 | @item -mfpu=@var{floating-point-format} | |
183 | ||
184 | This option specifies the floating point format to assemble for. The | |
185 | assembler will issue an error message if an attempt is made to assemble | |
186 | an instruction which will not execute on the target floating point unit. | |
187 | The following format options are recognized: | |
188 | @code{softfpa}, | |
189 | @code{fpe}, | |
bc89618b RE |
190 | @code{fpe2}, |
191 | @code{fpe3}, | |
03b1477f RE |
192 | @code{fpa}, |
193 | @code{fpa10}, | |
194 | @code{fpa11}, | |
195 | @code{arm7500fe}, | |
196 | @code{softvfp}, | |
197 | @code{softvfp+vfp}, | |
198 | @code{vfp}, | |
199 | @code{vfp10}, | |
200 | @code{vfp10-r0}, | |
201 | @code{vfp9}, | |
202 | @code{vfpxd}, | |
62f3b8c8 PB |
203 | @code{vfpv2}, |
204 | @code{vfpv3}, | |
205 | @code{vfpv3-fp16}, | |
206 | @code{vfpv3-d16}, | |
207 | @code{vfpv3-d16-fp16}, | |
208 | @code{vfpv3xd}, | |
209 | @code{vfpv3xd-d16}, | |
210 | @code{vfpv4}, | |
211 | @code{vfpv4-d16}, | |
09d92015 MM |
212 | @code{arm1020t}, |
213 | @code{arm1020e}, | |
b1cc4aeb | 214 | @code{arm1136jf-s}, |
62f3b8c8 PB |
215 | @code{maverick}, |
216 | @code{neon}, | |
03b1477f | 217 | and |
62f3b8c8 | 218 | @code{neon-vfpv4}. |
03b1477f RE |
219 | |
220 | In addition to determining which instructions are assembled, this option | |
221 | also affects the way in which the @code{.double} assembler directive behaves | |
222 | when assembling little-endian code. | |
223 | ||
224 | The default is dependent on the processor selected. For Architecture 5 or | |
225 | later, the default is to assembler for VFP instructions; for earlier | |
226 | architectures the default is to assemble for FPA instructions. | |
adcf07e6 | 227 | |
252b5132 RH |
228 | @cindex @code{-mthumb} command line option, ARM |
229 | @item -mthumb | |
03b1477f RE |
230 | This option specifies that the assembler should start assembling Thumb |
231 | instructions; that is, it should behave as though the file starts with a | |
232 | @code{.code 16} directive. | |
adcf07e6 | 233 | |
252b5132 RH |
234 | @cindex @code{-mthumb-interwork} command line option, ARM |
235 | @item -mthumb-interwork | |
236 | This option specifies that the output generated by the assembler should | |
237 | be marked as supporting interworking. | |
adcf07e6 | 238 | |
52970753 NC |
239 | @cindex @code{-mimplicit-it} command line option, ARM |
240 | @item -mimplicit-it=never | |
241 | @itemx -mimplicit-it=always | |
242 | @itemx -mimplicit-it=arm | |
243 | @itemx -mimplicit-it=thumb | |
244 | The @code{-mimplicit-it} option controls the behavior of the assembler when | |
245 | conditional instructions are not enclosed in IT blocks. | |
246 | There are four possible behaviors. | |
247 | If @code{never} is specified, such constructs cause a warning in ARM | |
248 | code and an error in Thumb-2 code. | |
249 | If @code{always} is specified, such constructs are accepted in both | |
250 | ARM and Thumb-2 code, where the IT instruction is added implicitly. | |
251 | If @code{arm} is specified, such constructs are accepted in ARM code | |
252 | and cause an error in Thumb-2 code. | |
253 | If @code{thumb} is specified, such constructs cause a warning in ARM | |
254 | code and are accepted in Thumb-2 code. If you omit this option, the | |
255 | behavior is equivalent to @code{-mimplicit-it=arm}. | |
e07e6e58 | 256 | |
5a5829dd NS |
257 | @cindex @code{-mapcs-26} command line option, ARM |
258 | @cindex @code{-mapcs-32} command line option, ARM | |
259 | @item -mapcs-26 | |
260 | @itemx -mapcs-32 | |
261 | These options specify that the output generated by the assembler should | |
252b5132 RH |
262 | be marked as supporting the indicated version of the Arm Procedure. |
263 | Calling Standard. | |
adcf07e6 | 264 | |
077b8428 NC |
265 | @cindex @code{-matpcs} command line option, ARM |
266 | @item -matpcs | |
267 | This option specifies that the output generated by the assembler should | |
268 | be marked as supporting the Arm/Thumb Procedure Calling Standard. If | |
269 | enabled this option will cause the assembler to create an empty | |
270 | debugging section in the object file called .arm.atpcs. Debuggers can | |
271 | use this to determine the ABI being used by. | |
272 | ||
adcf07e6 | 273 | @cindex @code{-mapcs-float} command line option, ARM |
252b5132 | 274 | @item -mapcs-float |
1be59579 | 275 | This indicates the floating point variant of the APCS should be |
252b5132 | 276 | used. In this variant floating point arguments are passed in FP |
550262c4 | 277 | registers rather than integer registers. |
adcf07e6 NC |
278 | |
279 | @cindex @code{-mapcs-reentrant} command line option, ARM | |
252b5132 RH |
280 | @item -mapcs-reentrant |
281 | This indicates that the reentrant variant of the APCS should be used. | |
282 | This variant supports position independent code. | |
adcf07e6 | 283 | |
33a392fb PB |
284 | @cindex @code{-mfloat-abi=} command line option, ARM |
285 | @item -mfloat-abi=@var{abi} | |
286 | This option specifies that the output generated by the assembler should be | |
287 | marked as using specified floating point ABI. | |
288 | The following values are recognized: | |
289 | @code{soft}, | |
290 | @code{softfp} | |
291 | and | |
292 | @code{hard}. | |
293 | ||
d507cf36 PB |
294 | @cindex @code{-eabi=} command line option, ARM |
295 | @item -meabi=@var{ver} | |
296 | This option specifies which EABI version the produced object files should | |
297 | conform to. | |
b45619c0 | 298 | The following values are recognized: |
3a4a14e9 PB |
299 | @code{gnu}, |
300 | @code{4} | |
d507cf36 | 301 | and |
3a4a14e9 | 302 | @code{5}. |
d507cf36 | 303 | |
252b5132 RH |
304 | @cindex @code{-EB} command line option, ARM |
305 | @item -EB | |
306 | This option specifies that the output generated by the assembler should | |
307 | be marked as being encoded for a big-endian processor. | |
adcf07e6 | 308 | |
252b5132 RH |
309 | @cindex @code{-EL} command line option, ARM |
310 | @item -EL | |
311 | This option specifies that the output generated by the assembler should | |
312 | be marked as being encoded for a little-endian processor. | |
adcf07e6 | 313 | |
252b5132 RH |
314 | @cindex @code{-k} command line option, ARM |
315 | @cindex PIC code generation for ARM | |
316 | @item -k | |
a349d9dd PB |
317 | This option specifies that the output of the assembler should be marked |
318 | as position-independent code (PIC). | |
adcf07e6 | 319 | |
845b51d6 PB |
320 | @cindex @code{--fix-v4bx} command line option, ARM |
321 | @item --fix-v4bx | |
322 | Allow @code{BX} instructions in ARMv4 code. This is intended for use with | |
323 | the linker option of the same name. | |
324 | ||
278df34e NS |
325 | @cindex @code{-mwarn-deprecated} command line option, ARM |
326 | @item -mwarn-deprecated | |
327 | @itemx -mno-warn-deprecated | |
328 | Enable or disable warnings about using deprecated options or | |
329 | features. The default is to warn. | |
330 | ||
252b5132 RH |
331 | @end table |
332 | ||
333 | ||
334 | @node ARM Syntax | |
335 | @section Syntax | |
336 | @menu | |
cab7e4d9 | 337 | * ARM-Instruction-Set:: Instruction Set |
252b5132 RH |
338 | * ARM-Chars:: Special Characters |
339 | * ARM-Regs:: Register Names | |
b6895b4f | 340 | * ARM-Relocations:: Relocations |
252b5132 RH |
341 | @end menu |
342 | ||
cab7e4d9 NC |
343 | @node ARM-Instruction-Set |
344 | @subsection Instruction Set Syntax | |
345 | Two slightly different syntaxes are support for ARM and THUMB | |
346 | instructions. The default, @code{divided}, uses the old style where | |
347 | ARM and THUMB instructions had their own, separate syntaxes. The new, | |
348 | @code{unified} syntax, which can be selected via the @code{.syntax} | |
349 | directive, and has the following main features: | |
350 | ||
351 | @table @bullet | |
352 | @item | |
353 | Immediate operands do not require a @code{#} prefix. | |
354 | ||
355 | @item | |
356 | The @code{IT} instruction may appear, and if it does it is validated | |
357 | against subsequent conditional affixes. In ARM mode it does not | |
358 | generate machine code, in THUMB mode it does. | |
359 | ||
360 | @item | |
361 | For ARM instructions the conditional affixes always appear at the end | |
362 | of the instruction. For THUMB instructions conditional affixes can be | |
363 | used, but only inside the scope of an @code{IT} instruction. | |
364 | ||
365 | @item | |
366 | All of the instructions new to the V6T2 architecture (and later) are | |
367 | available. (Only a few such instructions can be written in the | |
368 | @code{divided} syntax). | |
369 | ||
370 | @item | |
371 | The @code{.N} and @code{.W} suffixes are recognized and honored. | |
372 | ||
373 | @item | |
374 | All instructions set the flags if and only if they have an @code{s} | |
375 | affix. | |
376 | @end table | |
377 | ||
252b5132 RH |
378 | @node ARM-Chars |
379 | @subsection Special Characters | |
380 | ||
381 | @cindex line comment character, ARM | |
382 | @cindex ARM line comment character | |
550262c4 NC |
383 | The presence of a @samp{@@} on a line indicates the start of a comment |
384 | that extends to the end of the current line. If a @samp{#} appears as | |
385 | the first character of a line, the whole line is treated as a comment. | |
386 | ||
387 | @cindex line separator, ARM | |
388 | @cindex statement separator, ARM | |
389 | @cindex ARM line separator | |
a349d9dd PB |
390 | The @samp{;} character can be used instead of a newline to separate |
391 | statements. | |
550262c4 NC |
392 | |
393 | @cindex immediate character, ARM | |
394 | @cindex ARM immediate character | |
395 | Either @samp{#} or @samp{$} can be used to indicate immediate operands. | |
252b5132 RH |
396 | |
397 | @cindex identifiers, ARM | |
398 | @cindex ARM identifiers | |
399 | *TODO* Explain about /data modifier on symbols. | |
400 | ||
401 | @node ARM-Regs | |
402 | @subsection Register Names | |
403 | ||
404 | @cindex ARM register names | |
405 | @cindex register names, ARM | |
406 | *TODO* Explain about ARM register naming, and the predefined names. | |
407 | ||
408 | @node ARM Floating Point | |
409 | @section Floating Point | |
410 | ||
411 | @cindex floating point, ARM (@sc{ieee}) | |
412 | @cindex ARM floating point (@sc{ieee}) | |
413 | The ARM family uses @sc{ieee} floating-point numbers. | |
414 | ||
b6895b4f PB |
415 | @node ARM-Relocations |
416 | @subsection ARM relocation generation | |
417 | ||
418 | @cindex data relocations, ARM | |
419 | @cindex ARM data relocations | |
420 | Specific data relocations can be generated by putting the relocation name | |
421 | in parentheses after the symbol name. For example: | |
422 | ||
423 | @smallexample | |
424 | .word foo(TARGET1) | |
425 | @end smallexample | |
426 | ||
427 | This will generate an @samp{R_ARM_TARGET1} relocation against the symbol | |
428 | @var{foo}. | |
429 | The following relocations are supported: | |
430 | @code{GOT}, | |
431 | @code{GOTOFF}, | |
432 | @code{TARGET1}, | |
433 | @code{TARGET2}, | |
434 | @code{SBREL}, | |
435 | @code{TLSGD}, | |
436 | @code{TLSLDM}, | |
437 | @code{TLSLDO}, | |
438 | @code{GOTTPOFF} | |
439 | and | |
440 | @code{TPOFF}. | |
441 | ||
442 | For compatibility with older toolchains the assembler also accepts | |
443 | @code{(PLT)} after branch targets. This will generate the deprecated | |
444 | @samp{R_ARM_PLT32} relocation. | |
445 | ||
446 | @cindex MOVW and MOVT relocations, ARM | |
447 | Relocations for @samp{MOVW} and @samp{MOVT} instructions can be generated | |
448 | by prefixing the value with @samp{#:lower16:} and @samp{#:upper16} | |
b45619c0 | 449 | respectively. For example to load the 32-bit address of foo into r0: |
252b5132 | 450 | |
b6895b4f PB |
451 | @smallexample |
452 | MOVW r0, #:lower16:foo | |
453 | MOVT r0, #:upper16:foo | |
454 | @end smallexample | |
252b5132 RH |
455 | |
456 | @node ARM Directives | |
457 | @section ARM Machine Directives | |
458 | ||
459 | @cindex machine directives, ARM | |
460 | @cindex ARM machine directives | |
461 | @table @code | |
462 | ||
4a6bc624 NS |
463 | @c AAAAAAAAAAAAAAAAAAAAAAAAA |
464 | ||
465 | @cindex @code{.2byte} directive, ARM | |
466 | @cindex @code{.4byte} directive, ARM | |
467 | @cindex @code{.8byte} directive, ARM | |
468 | @item .2byte @var{expression} [, @var{expression}]* | |
469 | @itemx .4byte @var{expression} [, @var{expression}]* | |
470 | @itemx .8byte @var{expression} [, @var{expression}]* | |
471 | These directives write 2, 4 or 8 byte values to the output section. | |
472 | ||
473 | @cindex @code{.align} directive, ARM | |
adcf07e6 NC |
474 | @item .align @var{expression} [, @var{expression}] |
475 | This is the generic @var{.align} directive. For the ARM however if the | |
476 | first argument is zero (ie no alignment is needed) the assembler will | |
477 | behave as if the argument had been 2 (ie pad to the next four byte | |
062b7c0c | 478 | boundary). This is for compatibility with ARM's own assembler. |
adcf07e6 | 479 | |
4a6bc624 NS |
480 | @cindex @code{.arch} directive, ARM |
481 | @item .arch @var{name} | |
482 | Select the target architecture. Valid values for @var{name} are the same as | |
483 | for the @option{-march} commandline option. | |
252b5132 | 484 | |
4a6bc624 NS |
485 | @cindex @code{.arm} directive, ARM |
486 | @item .arm | |
487 | This performs the same action as @var{.code 32}. | |
252b5132 | 488 | |
4a6bc624 NS |
489 | @anchor{arm_pad} |
490 | @cindex @code{.pad} directive, ARM | |
491 | @item .pad #@var{count} | |
492 | Generate unwinder annotations for a stack adjustment of @var{count} bytes. | |
493 | A positive value indicates the function prologue allocated stack space by | |
494 | decrementing the stack pointer. | |
0bbf2aa4 | 495 | |
4a6bc624 | 496 | @c BBBBBBBBBBBBBBBBBBBBBBBBBB |
0bbf2aa4 | 497 | |
4a6bc624 NS |
498 | @cindex @code{.bss} directive, ARM |
499 | @item .bss | |
500 | This directive switches to the @code{.bss} section. | |
0bbf2aa4 | 501 | |
4a6bc624 NS |
502 | @c CCCCCCCCCCCCCCCCCCCCCCCCCC |
503 | ||
504 | @cindex @code{.cantunwind} directive, ARM | |
505 | @item .cantunwind | |
506 | Prevents unwinding through the current function. No personality routine | |
507 | or exception table data is required or permitted. | |
508 | ||
509 | @cindex @code{.code} directive, ARM | |
510 | @item .code @code{[16|32]} | |
511 | This directive selects the instruction set being generated. The value 16 | |
512 | selects Thumb, with the value 32 selecting ARM. | |
513 | ||
514 | @cindex @code{.cpu} directive, ARM | |
515 | @item .cpu @var{name} | |
516 | Select the target processor. Valid values for @var{name} are the same as | |
517 | for the @option{-mcpu} commandline option. | |
518 | ||
519 | @c DDDDDDDDDDDDDDDDDDDDDDDDDD | |
520 | ||
521 | @cindex @code{.dn} and @code{.qn} directives, ARM | |
f467aa98 BE |
522 | @item @var{name} .dn @var{register name} [@var{.type}] [[@var{index}]] |
523 | @item @var{name} .qn @var{register name} [@var{.type}] [[@var{index}]] | |
23753660 JB |
524 | |
525 | The @code{dn} and @code{qn} directives are used to create typed | |
526 | and/or indexed register aliases for use in Advanced SIMD Extension | |
527 | (Neon) instructions. The former should be used to create aliases | |
528 | of double-precision registers, and the latter to create aliases of | |
529 | quad-precision registers. | |
530 | ||
531 | If these directives are used to create typed aliases, those aliases can | |
532 | be used in Neon instructions instead of writing types after the mnemonic | |
533 | or after each operand. For example: | |
534 | ||
535 | @smallexample | |
536 | x .dn d2.f32 | |
537 | y .dn d3.f32 | |
538 | z .dn d4.f32[1] | |
539 | vmul x,y,z | |
540 | @end smallexample | |
541 | ||
542 | This is equivalent to writing the following: | |
543 | ||
544 | @smallexample | |
545 | vmul.f32 d2,d3,d4[1] | |
546 | @end smallexample | |
547 | ||
548 | Aliases created using @code{dn} or @code{qn} can be destroyed using | |
549 | @code{unreq}. | |
550 | ||
4a6bc624 | 551 | @c EEEEEEEEEEEEEEEEEEEEEEEEEE |
252b5132 | 552 | |
4a6bc624 NS |
553 | @cindex @code{.eabi_attribute} directive, ARM |
554 | @item .eabi_attribute @var{tag}, @var{value} | |
555 | Set the EABI object attribute @var{tag} to @var{value}. | |
252b5132 | 556 | |
4a6bc624 NS |
557 | The @var{tag} is either an attribute number, or one of the following: |
558 | @code{Tag_CPU_raw_name}, @code{Tag_CPU_name}, @code{Tag_CPU_arch}, | |
559 | @code{Tag_CPU_arch_profile}, @code{Tag_ARM_ISA_use}, | |
560 | @code{Tag_THUMB_ISA_use}, @code{Tag_VFP_arch}, @code{Tag_WMMX_arch}, | |
561 | @code{Tag_Advanced_SIMD_arch}, @code{Tag_PCS_config}, | |
562 | @code{Tag_ABI_PCS_R9_use}, @code{Tag_ABI_PCS_RW_data}, | |
563 | @code{Tag_ABI_PCS_RO_data}, @code{Tag_ABI_PCS_GOT_use}, | |
564 | @code{Tag_ABI_PCS_wchar_t}, @code{Tag_ABI_FP_rounding}, | |
565 | @code{Tag_ABI_FP_denormal}, @code{Tag_ABI_FP_exceptions}, | |
566 | @code{Tag_ABI_FP_user_exceptions}, @code{Tag_ABI_FP_number_model}, | |
567 | @code{Tag_ABI_align8_needed}, @code{Tag_ABI_align8_preserved}, | |
568 | @code{Tag_ABI_enum_size}, @code{Tag_ABI_HardFP_use}, | |
569 | @code{Tag_ABI_VFP_args}, @code{Tag_ABI_WMMX_args}, | |
570 | @code{Tag_ABI_optimization_goals}, @code{Tag_ABI_FP_optimization_goals}, | |
571 | @code{Tag_compatibility}, @code{Tag_CPU_unaligned_access}, | |
572 | @code{Tag_VFP_HP_extension}, @code{Tag_ABI_FP_16bit_format}, | |
573 | @code{Tag_nodefaults}, @code{Tag_also_compatible_with}, | |
574 | @code{Tag_conformance}, @code{Tag_T2EE_use}, | |
575 | @code{Tag_Virtualization_use}, @code{Tag_MPextension_use} | |
576 | ||
577 | The @var{value} is either a @code{number}, @code{"string"}, or | |
578 | @code{number, "string"} depending on the tag. | |
579 | ||
580 | @cindex @code{.even} directive, ARM | |
581 | @item .even | |
582 | This directive aligns to an even-numbered address. | |
583 | ||
584 | @cindex @code{.extend} directive, ARM | |
585 | @cindex @code{.ldouble} directive, ARM | |
586 | @item .extend @var{expression} [, @var{expression}]* | |
587 | @itemx .ldouble @var{expression} [, @var{expression}]* | |
588 | These directives write 12byte long double floating-point values to the | |
589 | output section. These are not compatible with current ARM processors | |
590 | or ABIs. | |
591 | ||
592 | @c FFFFFFFFFFFFFFFFFFFFFFFFFF | |
593 | ||
594 | @anchor{arm_fnend} | |
595 | @cindex @code{.fnend} directive, ARM | |
596 | @item .fnend | |
597 | Marks the end of a function with an unwind table entry. The unwind index | |
598 | table entry is created when this directive is processed. | |
252b5132 | 599 | |
4a6bc624 NS |
600 | If no personality routine has been specified then standard personality |
601 | routine 0 or 1 will be used, depending on the number of unwind opcodes | |
602 | required. | |
603 | ||
604 | @anchor{arm_fnstart} | |
605 | @cindex @code{.fnstart} directive, ARM | |
606 | @item .fnstart | |
607 | Marks the start of a function with an unwind table entry. | |
608 | ||
609 | @cindex @code{.force_thumb} directive, ARM | |
252b5132 RH |
610 | @item .force_thumb |
611 | This directive forces the selection of Thumb instructions, even if the | |
612 | target processor does not support those instructions | |
613 | ||
4a6bc624 NS |
614 | @cindex @code{.fpu} directive, ARM |
615 | @item .fpu @var{name} | |
616 | Select the floating-point unit to assemble for. Valid values for @var{name} | |
617 | are the same as for the @option{-mfpu} commandline option. | |
252b5132 | 618 | |
4a6bc624 NS |
619 | @c GGGGGGGGGGGGGGGGGGGGGGGGGG |
620 | @c HHHHHHHHHHHHHHHHHHHHHHHHHH | |
e1da3f5b | 621 | |
4a6bc624 NS |
622 | @cindex @code{.handlerdata} directive, ARM |
623 | @item .handlerdata | |
624 | Marks the end of the current function, and the start of the exception table | |
625 | entry for that function. Anything between this directive and the | |
626 | @code{.fnend} directive will be added to the exception table entry. | |
627 | ||
628 | Must be preceded by a @code{.personality} or @code{.personalityindex} | |
629 | directive. | |
630 | ||
631 | @c IIIIIIIIIIIIIIIIIIIIIIIIII | |
c921be7d NC |
632 | |
633 | @cindex @code{.inst} directive, ARM | |
634 | @item .inst @var{opcode} [ , @dots{} ] | |
635 | @item .inst.n @var{opcode} [ , @dots{} ] | |
636 | @item .inst.w @var{opcode} [ , @dots{} ] | |
637 | Generates the instruction corresponding to the numerical value @var{opcode}. | |
638 | @code{.inst.n} and @code{.inst.w} allow the Thumb instruction size to be | |
639 | specified explicitly, overriding the normal encoding rules. | |
640 | ||
4a6bc624 NS |
641 | @c JJJJJJJJJJJJJJJJJJJJJJJJJJ |
642 | @c KKKKKKKKKKKKKKKKKKKKKKKKKK | |
643 | @c LLLLLLLLLLLLLLLLLLLLLLLLLL | |
644 | ||
645 | @item .ldouble @var{expression} [, @var{expression}]* | |
646 | See @code{.extend}. | |
5395a469 | 647 | |
252b5132 RH |
648 | @cindex @code{.ltorg} directive, ARM |
649 | @item .ltorg | |
650 | This directive causes the current contents of the literal pool to be | |
651 | dumped into the current section (which is assumed to be the .text | |
652 | section) at the current location (aligned to a word boundary). | |
3d0c9500 NC |
653 | @code{GAS} maintains a separate literal pool for each section and each |
654 | sub-section. The @code{.ltorg} directive will only affect the literal | |
655 | pool of the current section and sub-section. At the end of assembly | |
656 | all remaining, un-empty literal pools will automatically be dumped. | |
657 | ||
658 | Note - older versions of @code{GAS} would dump the current literal | |
659 | pool any time a section change occurred. This is no longer done, since | |
660 | it prevents accurate control of the placement of literal pools. | |
252b5132 | 661 | |
4a6bc624 | 662 | @c MMMMMMMMMMMMMMMMMMMMMMMMMM |
252b5132 | 663 | |
4a6bc624 NS |
664 | @cindex @code{.movsp} directive, ARM |
665 | @item .movsp @var{reg} [, #@var{offset}] | |
666 | Tell the unwinder that @var{reg} contains an offset from the current | |
667 | stack pointer. If @var{offset} is not specified then it is assumed to be | |
668 | zero. | |
7ed4c4c5 | 669 | |
4a6bc624 NS |
670 | @c NNNNNNNNNNNNNNNNNNNNNNNNNN |
671 | @c OOOOOOOOOOOOOOOOOOOOOOOOOO | |
7ed4c4c5 | 672 | |
4a6bc624 NS |
673 | @cindex @code{.object_arch} directive, ARM |
674 | @item .object_arch @var{name} | |
675 | Override the architecture recorded in the EABI object attribute section. | |
676 | Valid values for @var{name} are the same as for the @code{.arch} directive. | |
677 | Typically this is useful when code uses runtime detection of CPU features. | |
7ed4c4c5 | 678 | |
4a6bc624 NS |
679 | @c PPPPPPPPPPPPPPPPPPPPPPPPPP |
680 | ||
681 | @cindex @code{.packed} directive, ARM | |
682 | @item .packed @var{expression} [, @var{expression}]* | |
683 | This directive writes 12-byte packed floating-point values to the | |
684 | output section. These are not compatible with current ARM processors | |
685 | or ABIs. | |
686 | ||
687 | @cindex @code{.pad} directive, ARM | |
688 | @item .pad #@var{count} | |
689 | Generate unwinder annotations for a stack adjustment of @var{count} bytes. | |
690 | A positive value indicates the function prologue allocated stack space by | |
691 | decrementing the stack pointer. | |
7ed4c4c5 NC |
692 | |
693 | @cindex @code{.personality} directive, ARM | |
694 | @item .personality @var{name} | |
695 | Sets the personality routine for the current function to @var{name}. | |
696 | ||
697 | @cindex @code{.personalityindex} directive, ARM | |
698 | @item .personalityindex @var{index} | |
699 | Sets the personality routine for the current function to the EABI standard | |
700 | routine number @var{index} | |
701 | ||
4a6bc624 NS |
702 | @cindex @code{.pool} directive, ARM |
703 | @item .pool | |
704 | This is a synonym for .ltorg. | |
7ed4c4c5 | 705 | |
4a6bc624 NS |
706 | @c QQQQQQQQQQQQQQQQQQQQQQQQQQ |
707 | @c RRRRRRRRRRRRRRRRRRRRRRRRRR | |
708 | ||
709 | @cindex @code{.req} directive, ARM | |
710 | @item @var{name} .req @var{register name} | |
711 | This creates an alias for @var{register name} called @var{name}. For | |
712 | example: | |
713 | ||
714 | @smallexample | |
715 | foo .req r0 | |
716 | @end smallexample | |
717 | ||
718 | @c SSSSSSSSSSSSSSSSSSSSSSSSSS | |
7ed4c4c5 | 719 | |
7da4f750 | 720 | @anchor{arm_save} |
7ed4c4c5 NC |
721 | @cindex @code{.save} directive, ARM |
722 | @item .save @var{reglist} | |
723 | Generate unwinder annotations to restore the registers in @var{reglist}. | |
724 | The format of @var{reglist} is the same as the corresponding store-multiple | |
725 | instruction. | |
726 | ||
727 | @smallexample | |
728 | @exdent @emph{core registers} | |
729 | .save @{r4, r5, r6, lr@} | |
730 | stmfd sp!, @{r4, r5, r6, lr@} | |
731 | @exdent @emph{FPA registers} | |
732 | .save f4, 2 | |
733 | sfmfd f4, 2, [sp]! | |
734 | @exdent @emph{VFP registers} | |
735 | .save @{d8, d9, d10@} | |
fa073d69 | 736 | fstmdx sp!, @{d8, d9, d10@} |
7ed4c4c5 NC |
737 | @exdent @emph{iWMMXt registers} |
738 | .save @{wr10, wr11@} | |
739 | wstrd wr11, [sp, #-8]! | |
740 | wstrd wr10, [sp, #-8]! | |
741 | or | |
742 | .save wr11 | |
743 | wstrd wr11, [sp, #-8]! | |
744 | .save wr10 | |
745 | wstrd wr10, [sp, #-8]! | |
746 | @end smallexample | |
747 | ||
7da4f750 | 748 | @anchor{arm_setfp} |
7ed4c4c5 NC |
749 | @cindex @code{.setfp} directive, ARM |
750 | @item .setfp @var{fpreg}, @var{spreg} [, #@var{offset}] | |
4a6bc624 | 751 | Make all unwinder annotations relative to a frame pointer. Without this |
7ed4c4c5 NC |
752 | the unwinder will use offsets from the stack pointer. |
753 | ||
754 | The syntax of this directive is the same as the @code{sub} or @code{mov} | |
755 | instruction used to set the frame pointer. @var{spreg} must be either | |
756 | @code{sp} or mentioned in a previous @code{.movsp} directive. | |
757 | ||
758 | @smallexample | |
759 | .movsp ip | |
760 | mov ip, sp | |
761 | @dots{} | |
762 | .setfp fp, ip, #4 | |
763 | sub fp, ip, #4 | |
764 | @end smallexample | |
765 | ||
4a6bc624 NS |
766 | @cindex @code{.secrel32} directive, ARM |
767 | @item .secrel32 @var{expression} [, @var{expression}]* | |
768 | This directive emits relocations that evaluate to the section-relative | |
769 | offset of each expression's symbol. This directive is only supported | |
770 | for PE targets. | |
771 | ||
cab7e4d9 NC |
772 | @cindex @code{.syntax} directive, ARM |
773 | @item .syntax [@code{unified} | @code{divided}] | |
774 | This directive sets the Instruction Set Syntax as described in the | |
775 | @ref{ARM-Instruction-Set} section. | |
776 | ||
4a6bc624 NS |
777 | @c TTTTTTTTTTTTTTTTTTTTTTTTTT |
778 | ||
779 | @cindex @code{.thumb} directive, ARM | |
780 | @item .thumb | |
781 | This performs the same action as @var{.code 16}. | |
782 | ||
783 | @cindex @code{.thumb_func} directive, ARM | |
784 | @item .thumb_func | |
785 | This directive specifies that the following symbol is the name of a | |
786 | Thumb encoded function. This information is necessary in order to allow | |
787 | the assembler and linker to generate correct code for interworking | |
788 | between Arm and Thumb instructions and should be used even if | |
789 | interworking is not going to be performed. The presence of this | |
790 | directive also implies @code{.thumb} | |
791 | ||
792 | This directive is not neccessary when generating EABI objects. On these | |
793 | targets the encoding is implicit when generating Thumb code. | |
794 | ||
795 | @cindex @code{.thumb_set} directive, ARM | |
796 | @item .thumb_set | |
797 | This performs the equivalent of a @code{.set} directive in that it | |
798 | creates a symbol which is an alias for another symbol (possibly not yet | |
799 | defined). This directive also has the added property in that it marks | |
800 | the aliased symbol as being a thumb function entry point, in the same | |
801 | way that the @code{.thumb_func} directive does. | |
802 | ||
803 | @c UUUUUUUUUUUUUUUUUUUUUUUUUU | |
804 | ||
805 | @cindex @code{.unreq} directive, ARM | |
806 | @item .unreq @var{alias-name} | |
807 | This undefines a register alias which was previously defined using the | |
808 | @code{req}, @code{dn} or @code{qn} directives. For example: | |
809 | ||
810 | @smallexample | |
811 | foo .req r0 | |
812 | .unreq foo | |
813 | @end smallexample | |
814 | ||
815 | An error occurs if the name is undefined. Note - this pseudo op can | |
816 | be used to delete builtin in register name aliases (eg 'r0'). This | |
817 | should only be done if it is really necessary. | |
818 | ||
7ed4c4c5 | 819 | @cindex @code{.unwind_raw} directive, ARM |
4a6bc624 | 820 | @item .unwind_raw @var{offset}, @var{byte1}, @dots{} |
7ed4c4c5 NC |
821 | Insert one of more arbitary unwind opcode bytes, which are known to adjust |
822 | the stack pointer by @var{offset} bytes. | |
823 | ||
824 | For example @code{.unwind_raw 4, 0xb1, 0x01} is equivalent to | |
825 | @code{.save @{r0@}} | |
826 | ||
4a6bc624 | 827 | @c VVVVVVVVVVVVVVVVVVVVVVVVVV |
ee065d83 | 828 | |
4a6bc624 NS |
829 | @cindex @code{.vsave} directive, ARM |
830 | @item .vsave @var{vfp-reglist} | |
831 | Generate unwinder annotations to restore the VFP registers in @var{vfp-reglist} | |
832 | using FLDMD. Also works for VFPv3 registers | |
833 | that are to be restored using VLDM. | |
834 | The format of @var{vfp-reglist} is the same as the corresponding store-multiple | |
835 | instruction. | |
ee065d83 | 836 | |
4a6bc624 NS |
837 | @smallexample |
838 | @exdent @emph{VFP registers} | |
839 | .vsave @{d8, d9, d10@} | |
840 | fstmdd sp!, @{d8, d9, d10@} | |
841 | @exdent @emph{VFPv3 registers} | |
842 | .vsave @{d15, d16, d17@} | |
843 | vstm sp!, @{d15, d16, d17@} | |
844 | @end smallexample | |
e04befd0 | 845 | |
4a6bc624 NS |
846 | Since FLDMX and FSTMX are now deprecated, this directive should be |
847 | used in favour of @code{.save} for saving VFP registers for ARMv6 and above. | |
e04befd0 | 848 | |
4a6bc624 NS |
849 | @c WWWWWWWWWWWWWWWWWWWWWWWWWW |
850 | @c XXXXXXXXXXXXXXXXXXXXXXXXXX | |
851 | @c YYYYYYYYYYYYYYYYYYYYYYYYYY | |
852 | @c ZZZZZZZZZZZZZZZZZZZZZZZZZZ | |
ee065d83 | 853 | |
252b5132 RH |
854 | @end table |
855 | ||
856 | @node ARM Opcodes | |
857 | @section Opcodes | |
858 | ||
859 | @cindex ARM opcodes | |
860 | @cindex opcodes for ARM | |
49a5575c NC |
861 | @code{@value{AS}} implements all the standard ARM opcodes. It also |
862 | implements several pseudo opcodes, including several synthetic load | |
863 | instructions. | |
252b5132 | 864 | |
49a5575c NC |
865 | @table @code |
866 | ||
867 | @cindex @code{NOP} pseudo op, ARM | |
868 | @item NOP | |
869 | @smallexample | |
870 | nop | |
871 | @end smallexample | |
252b5132 | 872 | |
49a5575c NC |
873 | This pseudo op will always evaluate to a legal ARM instruction that does |
874 | nothing. Currently it will evaluate to MOV r0, r0. | |
252b5132 | 875 | |
49a5575c NC |
876 | @cindex @code{LDR reg,=<label>} pseudo op, ARM |
877 | @item LDR | |
252b5132 RH |
878 | @smallexample |
879 | ldr <register> , = <expression> | |
880 | @end smallexample | |
881 | ||
882 | If expression evaluates to a numeric constant then a MOV or MVN | |
883 | instruction will be used in place of the LDR instruction, if the | |
884 | constant can be generated by either of these instructions. Otherwise | |
885 | the constant will be placed into the nearest literal pool (if it not | |
886 | already there) and a PC relative LDR instruction will be generated. | |
887 | ||
49a5575c NC |
888 | @cindex @code{ADR reg,<label>} pseudo op, ARM |
889 | @item ADR | |
890 | @smallexample | |
891 | adr <register> <label> | |
892 | @end smallexample | |
893 | ||
894 | This instruction will load the address of @var{label} into the indicated | |
895 | register. The instruction will evaluate to a PC relative ADD or SUB | |
896 | instruction depending upon where the label is located. If the label is | |
897 | out of range, or if it is not defined in the same file (and section) as | |
898 | the ADR instruction, then an error will be generated. This instruction | |
899 | will not make use of the literal pool. | |
900 | ||
901 | @cindex @code{ADRL reg,<label>} pseudo op, ARM | |
902 | @item ADRL | |
903 | @smallexample | |
904 | adrl <register> <label> | |
905 | @end smallexample | |
906 | ||
907 | This instruction will load the address of @var{label} into the indicated | |
a349d9dd | 908 | register. The instruction will evaluate to one or two PC relative ADD |
49a5575c NC |
909 | or SUB instructions depending upon where the label is located. If a |
910 | second instruction is not needed a NOP instruction will be generated in | |
911 | its place, so that this instruction is always 8 bytes long. | |
912 | ||
913 | If the label is out of range, or if it is not defined in the same file | |
914 | (and section) as the ADRL instruction, then an error will be generated. | |
915 | This instruction will not make use of the literal pool. | |
916 | ||
917 | @end table | |
918 | ||
252b5132 RH |
919 | For information on the ARM or Thumb instruction sets, see @cite{ARM |
920 | Software Development Toolkit Reference Manual}, Advanced RISC Machines | |
921 | Ltd. | |
922 | ||
6057a28f NC |
923 | @node ARM Mapping Symbols |
924 | @section Mapping Symbols | |
925 | ||
926 | The ARM ELF specification requires that special symbols be inserted | |
927 | into object files to mark certain features: | |
928 | ||
929 | @table @code | |
930 | ||
931 | @cindex @code{$a} | |
932 | @item $a | |
933 | At the start of a region of code containing ARM instructions. | |
934 | ||
935 | @cindex @code{$t} | |
936 | @item $t | |
937 | At the start of a region of code containing THUMB instructions. | |
938 | ||
939 | @cindex @code{$d} | |
940 | @item $d | |
941 | At the start of a region of data. | |
942 | ||
943 | @end table | |
944 | ||
945 | The assembler will automatically insert these symbols for you - there | |
946 | is no need to code them yourself. Support for tagging symbols ($b, | |
947 | $f, $p and $m) which is also mentioned in the current ARM ELF | |
948 | specification is not implemented. This is because they have been | |
949 | dropped from the new EABI and so tools cannot rely upon their | |
950 | presence. | |
951 | ||
7da4f750 MM |
952 | @node ARM Unwinding Tutorial |
953 | @section Unwinding | |
954 | ||
955 | The ABI for the ARM Architecture specifies a standard format for | |
956 | exception unwind information. This information is used when an | |
957 | exception is thrown to determine where control should be transferred. | |
958 | In particular, the unwind information is used to determine which | |
959 | function called the function that threw the exception, and which | |
960 | function called that one, and so forth. This information is also used | |
961 | to restore the values of callee-saved registers in the function | |
962 | catching the exception. | |
963 | ||
964 | If you are writing functions in assembly code, and those functions | |
965 | call other functions that throw exceptions, you must use assembly | |
966 | pseudo ops to ensure that appropriate exception unwind information is | |
967 | generated. Otherwise, if one of the functions called by your assembly | |
968 | code throws an exception, the run-time library will be unable to | |
969 | unwind the stack through your assembly code and your program will not | |
970 | behave correctly. | |
971 | ||
972 | To illustrate the use of these pseudo ops, we will examine the code | |
973 | that G++ generates for the following C++ input: | |
974 | ||
975 | @verbatim | |
976 | void callee (int *); | |
977 | ||
978 | int | |
979 | caller () | |
980 | { | |
981 | int i; | |
982 | callee (&i); | |
983 | return i; | |
984 | } | |
985 | @end verbatim | |
986 | ||
987 | This example does not show how to throw or catch an exception from | |
988 | assembly code. That is a much more complex operation and should | |
989 | always be done in a high-level language, such as C++, that directly | |
990 | supports exceptions. | |
991 | ||
992 | The code generated by one particular version of G++ when compiling the | |
993 | example above is: | |
994 | ||
995 | @verbatim | |
996 | _Z6callerv: | |
997 | .fnstart | |
998 | .LFB2: | |
999 | @ Function supports interworking. | |
1000 | @ args = 0, pretend = 0, frame = 8 | |
1001 | @ frame_needed = 1, uses_anonymous_args = 0 | |
1002 | stmfd sp!, {fp, lr} | |
1003 | .save {fp, lr} | |
1004 | .LCFI0: | |
1005 | .setfp fp, sp, #4 | |
1006 | add fp, sp, #4 | |
1007 | .LCFI1: | |
1008 | .pad #8 | |
1009 | sub sp, sp, #8 | |
1010 | .LCFI2: | |
1011 | sub r3, fp, #8 | |
1012 | mov r0, r3 | |
1013 | bl _Z6calleePi | |
1014 | ldr r3, [fp, #-8] | |
1015 | mov r0, r3 | |
1016 | sub sp, fp, #4 | |
1017 | ldmfd sp!, {fp, lr} | |
1018 | bx lr | |
1019 | .LFE2: | |
1020 | .fnend | |
1021 | @end verbatim | |
1022 | ||
1023 | Of course, the sequence of instructions varies based on the options | |
1024 | you pass to GCC and on the version of GCC in use. The exact | |
1025 | instructions are not important since we are focusing on the pseudo ops | |
1026 | that are used to generate unwind information. | |
1027 | ||
1028 | An important assumption made by the unwinder is that the stack frame | |
1029 | does not change during the body of the function. In particular, since | |
1030 | we assume that the assembly code does not itself throw an exception, | |
1031 | the only point where an exception can be thrown is from a call, such | |
1032 | as the @code{bl} instruction above. At each call site, the same saved | |
1033 | registers (including @code{lr}, which indicates the return address) | |
1034 | must be located in the same locations relative to the frame pointer. | |
1035 | ||
1036 | The @code{.fnstart} (@pxref{arm_fnstart,,.fnstart pseudo op}) pseudo | |
1037 | op appears immediately before the first instruction of the function | |
1038 | while the @code{.fnend} (@pxref{arm_fnend,,.fnend pseudo op}) pseudo | |
1039 | op appears immediately after the last instruction of the function. | |
1040 | These pseudo ops specify the range of the function. | |
1041 | ||
1042 | Only the order of the other pseudos ops (e.g., @code{.setfp} or | |
1043 | @code{.pad}) matters; their exact locations are irrelevant. In the | |
1044 | example above, the compiler emits the pseudo ops with particular | |
1045 | instructions. That makes it easier to understand the code, but it is | |
1046 | not required for correctness. It would work just as well to emit all | |
1047 | of the pseudo ops other than @code{.fnend} in the same order, but | |
1048 | immediately after @code{.fnstart}. | |
1049 | ||
1050 | The @code{.save} (@pxref{arm_save,,.save pseudo op}) pseudo op | |
1051 | indicates registers that have been saved to the stack so that they can | |
1052 | be restored before the function returns. The argument to the | |
1053 | @code{.save} pseudo op is a list of registers to save. If a register | |
1054 | is ``callee-saved'' (as specified by the ABI) and is modified by the | |
1055 | function you are writing, then your code must save the value before it | |
1056 | is modified and restore the original value before the function | |
1057 | returns. If an exception is thrown, the run-time library restores the | |
1058 | values of these registers from their locations on the stack before | |
1059 | returning control to the exception handler. (Of course, if an | |
1060 | exception is not thrown, the function that contains the @code{.save} | |
1061 | pseudo op restores these registers in the function epilogue, as is | |
1062 | done with the @code{ldmfd} instruction above.) | |
1063 | ||
1064 | You do not have to save callee-saved registers at the very beginning | |
1065 | of the function and you do not need to use the @code{.save} pseudo op | |
1066 | immediately following the point at which the registers are saved. | |
1067 | However, if you modify a callee-saved register, you must save it on | |
1068 | the stack before modifying it and before calling any functions which | |
1069 | might throw an exception. And, you must use the @code{.save} pseudo | |
1070 | op to indicate that you have done so. | |
1071 | ||
1072 | The @code{.pad} (@pxref{arm_pad,,.pad}) pseudo op indicates a | |
1073 | modification of the stack pointer that does not save any registers. | |
1074 | The argument is the number of bytes (in decimal) that are subtracted | |
1075 | from the stack pointer. (On ARM CPUs, the stack grows downwards, so | |
1076 | subtracting from the stack pointer increases the size of the stack.) | |
1077 | ||
1078 | The @code{.setfp} (@pxref{arm_setfp,,.setfp pseudo op}) pseudo op | |
1079 | indicates the register that contains the frame pointer. The first | |
1080 | argument is the register that is set, which is typically @code{fp}. | |
1081 | The second argument indicates the register from which the frame | |
1082 | pointer takes its value. The third argument, if present, is the value | |
1083 | (in decimal) added to the register specified by the second argument to | |
1084 | compute the value of the frame pointer. You should not modify the | |
1085 | frame pointer in the body of the function. | |
1086 | ||
1087 | If you do not use a frame pointer, then you should not use the | |
1088 | @code{.setfp} pseudo op. If you do not use a frame pointer, then you | |
1089 | should avoid modifying the stack pointer outside of the function | |
1090 | prologue. Otherwise, the run-time library will be unable to find | |
1091 | saved registers when it is unwinding the stack. | |
1092 | ||
1093 | The pseudo ops described above are sufficient for writing assembly | |
1094 | code that calls functions which may throw exceptions. If you need to | |
1095 | know more about the object-file format used to represent unwind | |
1096 | information, you may consult the @cite{Exception Handling ABI for the | |
1097 | ARM Architecture} available from @uref{http://infocenter.arm.com}. |