1 \input texinfo @c -*-Texinfo-*-
2 @c Copyright (C) 1991-2015 Free Software Foundation, Inc.
3 @c UPDATE!! On future updates--
4 @c (1) check for new machine-dep cmdline options in
5 @c md_parse_option definitions in config/tc-*.c
6 @c (2) for platform-specific directives, examine md_pseudo_op
8 @c (3) for object-format specific directives, examine obj_pseudo_op
10 @c (4) portable directives in potable[] in read.c
14 @macro gcctabopt{body}
17 @c defaults, config file may override:
22 @include asconfig.texi
27 @c common OR combinations of conditions
53 @set abnormal-separator
57 @settitle Using @value{AS}
60 @settitle Using @value{AS} (@value{TARGET})
62 @setchapternewpage odd
67 @c WARE! Some of the machine-dependent sections contain tables of machine
68 @c instructions. Except in multi-column format, these tables look silly.
69 @c Unfortunately, Texinfo doesn't have a general-purpose multi-col format, so
70 @c the multi-col format is faked within @example sections.
72 @c Again unfortunately, the natural size that fits on a page, for these tables,
73 @c is different depending on whether or not smallbook is turned on.
74 @c This matters, because of order: text flow switches columns at each page
77 @c The format faked in this source works reasonably well for smallbook,
78 @c not well for the default large-page format. This manual expects that if you
79 @c turn on @smallbook, you will also uncomment the "@set SMALL" to enable the
80 @c tables in question. You can turn on one without the other at your
81 @c discretion, of course.
84 @c the insn tables look just as silly in info files regardless of smallbook,
85 @c might as well show 'em anyways.
89 @dircategory Software development
91 * As: (as). The GNU assembler.
92 * Gas: (as). The GNU assembler.
100 This file documents the GNU Assembler "@value{AS}".
102 @c man begin COPYRIGHT
103 Copyright @copyright{} 1991-2015 Free Software Foundation, Inc.
105 Permission is granted to copy, distribute and/or modify this document
106 under the terms of the GNU Free Documentation License, Version 1.3
107 or any later version published by the Free Software Foundation;
108 with no Invariant Sections, with no Front-Cover Texts, and with no
109 Back-Cover Texts. A copy of the license is included in the
110 section entitled ``GNU Free Documentation License''.
116 @title Using @value{AS}
117 @subtitle The @sc{gnu} Assembler
119 @subtitle for the @value{TARGET} family
121 @ifset VERSION_PACKAGE
123 @subtitle @value{VERSION_PACKAGE}
126 @subtitle Version @value{VERSION}
129 The Free Software Foundation Inc.@: thanks The Nice Computer
130 Company of Australia for loaning Dean Elsner to write the
131 first (Vax) version of @command{as} for Project @sc{gnu}.
132 The proprietors, management and staff of TNCCA thank FSF for
133 distracting the boss while they got some work
136 @author Dean Elsner, Jay Fenlason & friends
140 \hfill {\it Using {\tt @value{AS}}}\par
141 \hfill Edited by Cygnus Support\par
143 %"boxit" macro for figures:
144 %Modified from Knuth's ``boxit'' macro from TeXbook (answer to exercise 21.3)
145 \gdef\boxit#1#2{\vbox{\hrule\hbox{\vrule\kern3pt
146 \vbox{\parindent=0pt\parskip=0pt\hsize=#1\kern3pt\strut\hfil
147 #2\hfil\strut\kern3pt}\kern3pt\vrule}\hrule}}%box with visible outline
148 \gdef\ibox#1#2{\hbox to #1{#2\hfil}\kern8pt}% invisible box
151 @vskip 0pt plus 1filll
152 Copyright @copyright{} 1991-2015 Free Software Foundation, Inc.
154 Permission is granted to copy, distribute and/or modify this document
155 under the terms of the GNU Free Documentation License, Version 1.3
156 or any later version published by the Free Software Foundation;
157 with no Invariant Sections, with no Front-Cover Texts, and with no
158 Back-Cover Texts. A copy of the license is included in the
159 section entitled ``GNU Free Documentation License''.
166 @top Using @value{AS}
168 This file is a user guide to the @sc{gnu} assembler @command{@value{AS}}
169 @ifset VERSION_PACKAGE
170 @value{VERSION_PACKAGE}
172 version @value{VERSION}.
174 This version of the file describes @command{@value{AS}} configured to generate
175 code for @value{TARGET} architectures.
178 This document is distributed under the terms of the GNU Free
179 Documentation License. A copy of the license is included in the
180 section entitled ``GNU Free Documentation License''.
183 * Overview:: Overview
184 * Invoking:: Command-Line Options
186 * Sections:: Sections and Relocation
188 * Expressions:: Expressions
189 * Pseudo Ops:: Assembler Directives
191 * Object Attributes:: Object Attributes
193 * Machine Dependencies:: Machine Dependent Features
194 * Reporting Bugs:: Reporting Bugs
195 * Acknowledgements:: Who Did What
196 * GNU Free Documentation License:: GNU Free Documentation License
197 * AS Index:: AS Index
204 This manual is a user guide to the @sc{gnu} assembler @command{@value{AS}}.
206 This version of the manual describes @command{@value{AS}} configured to generate
207 code for @value{TARGET} architectures.
211 @cindex invocation summary
212 @cindex option summary
213 @cindex summary of options
214 Here is a brief summary of how to invoke @command{@value{AS}}. For details,
215 see @ref{Invoking,,Command-Line Options}.
217 @c man title AS the portable GNU assembler.
221 gcc(1), ld(1), and the Info entries for @file{binutils} and @file{ld}.
225 @c We don't use deffn and friends for the following because they seem
226 @c to be limited to one line for the header.
228 @c man begin SYNOPSIS
229 @value{AS} [@b{-a}[@b{cdghlns}][=@var{file}]] [@b{--alternate}] [@b{-D}]
230 [@b{--compress-debug-sections}] [@b{--nocompress-debug-sections}]
231 [@b{--debug-prefix-map} @var{old}=@var{new}]
232 [@b{--defsym} @var{sym}=@var{val}] [@b{-f}] [@b{-g}] [@b{--gstabs}]
233 [@b{--gstabs+}] [@b{--gdwarf-2}] [@b{--gdwarf-sections}]
234 [@b{--help}] [@b{-I} @var{dir}] [@b{-J}]
235 [@b{-K}] [@b{-L}] [@b{--listing-lhs-width}=@var{NUM}]
236 [@b{--listing-lhs-width2}=@var{NUM}] [@b{--listing-rhs-width}=@var{NUM}]
237 [@b{--listing-cont-lines}=@var{NUM}] [@b{--keep-locals}] [@b{-o}
238 @var{objfile}] [@b{-R}] [@b{--reduce-memory-overheads}] [@b{--statistics}]
239 [@b{-v}] [@b{-version}] [@b{--version}] [@b{-W}] [@b{--warn}]
240 [@b{--fatal-warnings}] [@b{-w}] [@b{-x}] [@b{-Z}] [@b{@@@var{FILE}}]
241 [@b{--size-check=[error|warning]}]
242 [@b{--target-help}] [@var{target-options}]
243 [@b{--}|@var{files} @dots{}]
245 @c Target dependent options are listed below. Keep the list sorted.
246 @c Add an empty line for separation.
249 @emph{Target AArch64 options:}
251 [@b{-mabi}=@var{ABI}]
255 @emph{Target Alpha options:}
257 [@b{-mdebug} | @b{-no-mdebug}]
258 [@b{-replace} | @b{-noreplace}]
259 [@b{-relax}] [@b{-g}] [@b{-G@var{size}}]
260 [@b{-F}] [@b{-32addr}]
264 @emph{Target ARC options:}
270 @emph{Target ARM options:}
271 @c Don't document the deprecated options
272 [@b{-mcpu}=@var{processor}[+@var{extension}@dots{}]]
273 [@b{-march}=@var{architecture}[+@var{extension}@dots{}]]
274 [@b{-mfpu}=@var{floating-point-format}]
275 [@b{-mfloat-abi}=@var{abi}]
276 [@b{-meabi}=@var{ver}]
279 [@b{-mapcs-32}|@b{-mapcs-26}|@b{-mapcs-float}|
280 @b{-mapcs-reentrant}]
281 [@b{-mthumb-interwork}] [@b{-k}]
285 @emph{Target Blackfin options:}
286 [@b{-mcpu}=@var{processor}[-@var{sirevision}]]
293 @emph{Target CRIS options:}
294 [@b{--underscore} | @b{--no-underscore}]
296 [@b{--emulation=criself} | @b{--emulation=crisaout}]
297 [@b{--march=v0_v10} | @b{--march=v10} | @b{--march=v32} | @b{--march=common_v10_v32}]
298 @c Deprecated -- deliberately not documented.
303 @emph{Target D10V options:}
308 @emph{Target D30V options:}
309 [@b{-O}|@b{-n}|@b{-N}]
313 @emph{Target EPIPHANY options:}
314 [@b{-mepiphany}|@b{-mepiphany16}]
318 @emph{Target H8/300 options:}
322 @c HPPA has no machine-dependent assembler options (yet).
326 @emph{Target i386 options:}
327 [@b{--32}|@b{--x32}|@b{--64}] [@b{-n}]
328 [@b{-march}=@var{CPU}[+@var{EXTENSION}@dots{}]] [@b{-mtune}=@var{CPU}]
332 @emph{Target i960 options:}
333 @c see md_parse_option in tc-i960.c
334 [@b{-ACA}|@b{-ACA_A}|@b{-ACB}|@b{-ACC}|@b{-AKA}|@b{-AKB}|
336 [@b{-b}] [@b{-no-relax}]
340 @emph{Target IA-64 options:}
341 [@b{-mconstant-gp}|@b{-mauto-pic}]
342 [@b{-milp32}|@b{-milp64}|@b{-mlp64}|@b{-mp64}]
344 [@b{-mtune=itanium1}|@b{-mtune=itanium2}]
345 [@b{-munwind-check=warning}|@b{-munwind-check=error}]
346 [@b{-mhint.b=ok}|@b{-mhint.b=warning}|@b{-mhint.b=error}]
347 [@b{-x}|@b{-xexplicit}] [@b{-xauto}] [@b{-xdebug}]
351 @emph{Target IP2K options:}
352 [@b{-mip2022}|@b{-mip2022ext}]
356 @emph{Target M32C options:}
357 [@b{-m32c}|@b{-m16c}] [-relax] [-h-tick-hex]
361 @emph{Target M32R options:}
362 [@b{--m32rx}|@b{--[no-]warn-explicit-parallel-conflicts}|
367 @emph{Target M680X0 options:}
368 [@b{-l}] [@b{-m68000}|@b{-m68010}|@b{-m68020}|@dots{}]
372 @emph{Target M68HC11 options:}
373 [@b{-m68hc11}|@b{-m68hc12}|@b{-m68hcs12}|@b{-mm9s12x}|@b{-mm9s12xg}]
374 [@b{-mshort}|@b{-mlong}]
375 [@b{-mshort-double}|@b{-mlong-double}]
376 [@b{--force-long-branches}] [@b{--short-branches}]
377 [@b{--strict-direct-mode}] [@b{--print-insn-syntax}]
378 [@b{--print-opcodes}] [@b{--generate-example}]
382 @emph{Target MCORE options:}
383 [@b{-jsri2bsr}] [@b{-sifilter}] [@b{-relax}]
384 [@b{-mcpu=[210|340]}]
388 @emph{Target Meta options:}
389 [@b{-mcpu=@var{cpu}}] [@b{-mfpu=@var{cpu}}] [@b{-mdsp=@var{cpu}}]
392 @emph{Target MICROBLAZE options:}
393 @c MicroBlaze has no machine-dependent assembler options.
397 @emph{Target MIPS options:}
398 [@b{-nocpp}] [@b{-EL}] [@b{-EB}] [@b{-O}[@var{optimization level}]]
399 [@b{-g}[@var{debug level}]] [@b{-G} @var{num}] [@b{-KPIC}] [@b{-call_shared}]
400 [@b{-non_shared}] [@b{-xgot} [@b{-mvxworks-pic}]
401 [@b{-mabi}=@var{ABI}] [@b{-32}] [@b{-n32}] [@b{-64}] [@b{-mfp32}] [@b{-mgp32}]
402 [@b{-mfp64}] [@b{-mgp64}] [@b{-mfpxx}]
403 [@b{-modd-spreg}] [@b{-mno-odd-spreg}]
404 [@b{-march}=@var{CPU}] [@b{-mtune}=@var{CPU}] [@b{-mips1}] [@b{-mips2}]
405 [@b{-mips3}] [@b{-mips4}] [@b{-mips5}] [@b{-mips32}] [@b{-mips32r2}]
406 [@b{-mips32r3}] [@b{-mips32r5}] [@b{-mips32r6}] [@b{-mips64}] [@b{-mips64r2}]
407 [@b{-mips64r3}] [@b{-mips64r5}] [@b{-mips64r6}]
408 [@b{-construct-floats}] [@b{-no-construct-floats}]
409 [@b{-mnan=@var{encoding}}]
410 [@b{-trap}] [@b{-no-break}] [@b{-break}] [@b{-no-trap}]
411 [@b{-mips16}] [@b{-no-mips16}]
412 [@b{-mmicromips}] [@b{-mno-micromips}]
413 [@b{-msmartmips}] [@b{-mno-smartmips}]
414 [@b{-mips3d}] [@b{-no-mips3d}]
415 [@b{-mdmx}] [@b{-no-mdmx}]
416 [@b{-mdsp}] [@b{-mno-dsp}]
417 [@b{-mdspr2}] [@b{-mno-dspr2}]
418 [@b{-mmsa}] [@b{-mno-msa}]
419 [@b{-mxpa}] [@b{-mno-xpa}]
420 [@b{-mmt}] [@b{-mno-mt}]
421 [@b{-mmcu}] [@b{-mno-mcu}]
422 [@b{-minsn32}] [@b{-mno-insn32}]
423 [@b{-mfix7000}] [@b{-mno-fix7000}]
424 [@b{-mfix-rm7000}] [@b{-mno-fix-rm7000}]
425 [@b{-mfix-vr4120}] [@b{-mno-fix-vr4120}]
426 [@b{-mfix-vr4130}] [@b{-mno-fix-vr4130}]
427 [@b{-mdebug}] [@b{-no-mdebug}]
428 [@b{-mpdr}] [@b{-mno-pdr}]
432 @emph{Target MMIX options:}
433 [@b{--fixed-special-register-names}] [@b{--globalize-symbols}]
434 [@b{--gnu-syntax}] [@b{--relax}] [@b{--no-predefined-symbols}]
435 [@b{--no-expand}] [@b{--no-merge-gregs}] [@b{-x}]
436 [@b{--linker-allocated-gregs}]
440 @emph{Target Nios II options:}
441 [@b{-relax-all}] [@b{-relax-section}] [@b{-no-relax}]
446 @emph{Target NDS32 options:}
447 [@b{-EL}] [@b{-EB}] [@b{-O}] [@b{-Os}] [@b{-mcpu=@var{cpu}}]
448 [@b{-misa=@var{isa}}] [@b{-mabi=@var{abi}}] [@b{-mall-ext}]
449 [@b{-m[no-]16-bit}] [@b{-m[no-]perf-ext}] [@b{-m[no-]perf2-ext}]
450 [@b{-m[no-]string-ext}] [@b{-m[no-]dsp-ext}] [@b{-m[no-]mac}] [@b{-m[no-]div}]
451 [@b{-m[no-]audio-isa-ext}] [@b{-m[no-]fpu-sp-ext}] [@b{-m[no-]fpu-dp-ext}]
452 [@b{-m[no-]fpu-fma}] [@b{-mfpu-freg=@var{FREG}}] [@b{-mreduced-regs}]
453 [@b{-mfull-regs}] [@b{-m[no-]dx-regs}] [@b{-mpic}] [@b{-mno-relax}]
458 @emph{Target PDP11 options:}
459 [@b{-mpic}|@b{-mno-pic}] [@b{-mall}] [@b{-mno-extensions}]
460 [@b{-m}@var{extension}|@b{-mno-}@var{extension}]
461 [@b{-m}@var{cpu}] [@b{-m}@var{machine}]
465 @emph{Target picoJava options:}
470 @emph{Target PowerPC options:}
472 [@b{-mpwrx}|@b{-mpwr2}|@b{-mpwr}|@b{-m601}|@b{-mppc}|@b{-mppc32}|@b{-m603}|@b{-m604}|@b{-m403}|@b{-m405}|
473 @b{-m440}|@b{-m464}|@b{-m476}|@b{-m7400}|@b{-m7410}|@b{-m7450}|@b{-m7455}|@b{-m750cl}|@b{-mppc64}|
474 @b{-m620}|@b{-me500}|@b{-e500x2}|@b{-me500mc}|@b{-me500mc64}|@b{-me5500}|@b{-me6500}|@b{-mppc64bridge}|
475 @b{-mbooke}|@b{-mpower4}|@b{-mpwr4}|@b{-mpower5}|@b{-mpwr5}|@b{-mpwr5x}|@b{-mpower6}|@b{-mpwr6}|
476 @b{-mpower7}|@b{-mpwr7}|@b{-mpower8}|@b{-mpwr8}|@b{-ma2}|@b{-mcell}|@b{-mspe}|@b{-mtitan}|@b{-me300}|@b{-mcom}]
477 [@b{-many}] [@b{-maltivec}|@b{-mvsx}|@b{-mhtm}|@b{-mvle}]
478 [@b{-mregnames}|@b{-mno-regnames}]
479 [@b{-mrelocatable}|@b{-mrelocatable-lib}|@b{-K PIC}] [@b{-memb}]
480 [@b{-mlittle}|@b{-mlittle-endian}|@b{-le}|@b{-mbig}|@b{-mbig-endian}|@b{-be}]
481 [@b{-msolaris}|@b{-mno-solaris}]
482 [@b{-nops=@var{count}}]
486 @emph{Target RL78 options:}
488 [@b{-m32bit-doubles}|@b{-m64bit-doubles}]
492 @emph{Target RX options:}
493 [@b{-mlittle-endian}|@b{-mbig-endian}]
494 [@b{-m32bit-doubles}|@b{-m64bit-doubles}]
495 [@b{-muse-conventional-section-names}]
496 [@b{-msmall-data-limit}]
499 [@b{-mint-register=@var{number}}]
500 [@b{-mgcc-abi}|@b{-mrx-abi}]
504 @emph{Target s390 options:}
505 [@b{-m31}|@b{-m64}] [@b{-mesa}|@b{-mzarch}] [@b{-march}=@var{CPU}]
506 [@b{-mregnames}|@b{-mno-regnames}]
507 [@b{-mwarn-areg-zero}]
511 @emph{Target SCORE options:}
512 [@b{-EB}][@b{-EL}][@b{-FIXDD}][@b{-NWARN}]
513 [@b{-SCORE5}][@b{-SCORE5U}][@b{-SCORE7}][@b{-SCORE3}]
514 [@b{-march=score7}][@b{-march=score3}]
515 [@b{-USE_R1}][@b{-KPIC}][@b{-O0}][@b{-G} @var{num}][@b{-V}]
519 @emph{Target SPARC options:}
520 @c The order here is important. See c-sparc.texi.
521 [@b{-Av6}|@b{-Av7}|@b{-Av8}|@b{-Asparclet}|@b{-Asparclite}
522 @b{-Av8plus}|@b{-Av8plusa}|@b{-Av9}|@b{-Av9a}]
523 [@b{-xarch=v8plus}|@b{-xarch=v8plusa}] [@b{-bump}]
528 @emph{Target TIC54X options:}
529 [@b{-mcpu=54[123589]}|@b{-mcpu=54[56]lp}] [@b{-mfar-mode}|@b{-mf}]
530 [@b{-merrors-to-file} @var{<filename>}|@b{-me} @var{<filename>}]
534 @emph{Target TIC6X options:}
535 [@b{-march=@var{arch}}] [@b{-mbig-endian}|@b{-mlittle-endian}]
536 [@b{-mdsbt}|@b{-mno-dsbt}] [@b{-mpid=no}|@b{-mpid=near}|@b{-mpid=far}]
537 [@b{-mpic}|@b{-mno-pic}]
541 @emph{Target TILE-Gx options:}
542 [@b{-m32}|@b{-m64}][@b{-EB}][@b{-EL}]
545 @c TILEPro has no machine-dependent assembler options
549 @emph{Target Visium options:}
550 [@b{-mtune=@var{arch}}]
554 @emph{Target Xtensa options:}
555 [@b{--[no-]text-section-literals}] [@b{--[no-]absolute-literals}]
556 [@b{--[no-]target-align}] [@b{--[no-]longcalls}]
557 [@b{--[no-]transform}]
558 [@b{--rename-section} @var{oldname}=@var{newname}]
559 [@b{--[no-]trampolines}]
563 @emph{Target Z80 options:}
564 [@b{-z80}] [@b{-r800}]
565 [@b{ -ignore-undocumented-instructions}] [@b{-Wnud}]
566 [@b{ -ignore-unportable-instructions}] [@b{-Wnup}]
567 [@b{ -warn-undocumented-instructions}] [@b{-Wud}]
568 [@b{ -warn-unportable-instructions}] [@b{-Wup}]
569 [@b{ -forbid-undocumented-instructions}] [@b{-Fud}]
570 [@b{ -forbid-unportable-instructions}] [@b{-Fup}]
574 @c Z8000 has no machine-dependent assembler options
583 @include at-file.texi
586 Turn on listings, in any of a variety of ways:
590 omit false conditionals
593 omit debugging directives
596 include general information, like @value{AS} version and options passed
599 include high-level source
605 include macro expansions
608 omit forms processing
614 set the name of the listing file
617 You may combine these options; for example, use @samp{-aln} for assembly
618 listing without forms processing. The @samp{=file} option, if used, must be
619 the last one. By itself, @samp{-a} defaults to @samp{-ahls}.
622 Begin in alternate macro mode.
624 @xref{Altmacro,,@code{.altmacro}}.
627 @item --compress-debug-sections
628 Compress DWARF debug sections using zlib. The debug sections are renamed
629 to begin with @samp{.zdebug}, and the resulting object file may not be
630 compatible with older linkers and object file utilities. Note if compression
631 would make a given section @emph{larger} then it is not compressed or renamed.
633 @item --nocompress-debug-sections
634 Do not compress DWARF debug sections. This is the default.
637 Ignored. This option is accepted for script compatibility with calls to
640 @item --debug-prefix-map @var{old}=@var{new}
641 When assembling files in directory @file{@var{old}}, record debugging
642 information describing them as in @file{@var{new}} instead.
644 @item --defsym @var{sym}=@var{value}
645 Define the symbol @var{sym} to be @var{value} before assembling the input file.
646 @var{value} must be an integer constant. As in C, a leading @samp{0x}
647 indicates a hexadecimal value, and a leading @samp{0} indicates an octal
648 value. The value of the symbol can be overridden inside a source file via the
649 use of a @code{.set} pseudo-op.
652 ``fast''---skip whitespace and comment preprocessing (assume source is
657 Generate debugging information for each assembler source line using whichever
658 debug format is preferred by the target. This currently means either STABS,
662 Generate stabs debugging information for each assembler line. This
663 may help debugging assembler code, if the debugger can handle it.
666 Generate stabs debugging information for each assembler line, with GNU
667 extensions that probably only gdb can handle, and that could make other
668 debuggers crash or refuse to read your program. This
669 may help debugging assembler code. Currently the only GNU extension is
670 the location of the current working directory at assembling time.
673 Generate DWARF2 debugging information for each assembler line. This
674 may help debugging assembler code, if the debugger can handle it. Note---this
675 option is only supported by some targets, not all of them.
677 @item --gdwarf-sections
678 Instead of creating a .debug_line section, create a series of
679 .debug_line.@var{foo} sections where @var{foo} is the name of the
680 corresponding code section. For example a code section called @var{.text.func}
681 will have its dwarf line number information placed into a section called
682 @var{.debug_line.text.func}. If the code section is just called @var{.text}
683 then debug line section will still be called just @var{.debug_line} without any
686 @item --size-check=error
687 @itemx --size-check=warning
688 Issue an error or warning for invalid ELF .size directive.
691 Print a summary of the command line options and exit.
694 Print a summary of all target specific options and exit.
697 Add directory @var{dir} to the search list for @code{.include} directives.
700 Don't warn about signed overflow.
703 @ifclear DIFF-TBL-KLUGE
704 This option is accepted but has no effect on the @value{TARGET} family.
706 @ifset DIFF-TBL-KLUGE
707 Issue warnings when difference tables altered for long displacements.
712 Keep (in the symbol table) local symbols. These symbols start with
713 system-specific local label prefixes, typically @samp{.L} for ELF systems
714 or @samp{L} for traditional a.out systems.
719 @item --listing-lhs-width=@var{number}
720 Set the maximum width, in words, of the output data column for an assembler
721 listing to @var{number}.
723 @item --listing-lhs-width2=@var{number}
724 Set the maximum width, in words, of the output data column for continuation
725 lines in an assembler listing to @var{number}.
727 @item --listing-rhs-width=@var{number}
728 Set the maximum width of an input source line, as displayed in a listing, to
731 @item --listing-cont-lines=@var{number}
732 Set the maximum number of lines printed in a listing for a single line of input
735 @item -o @var{objfile}
736 Name the object-file output from @command{@value{AS}} @var{objfile}.
739 Fold the data section into the text section.
741 @kindex --hash-size=@var{number}
742 Set the default size of GAS's hash tables to a prime number close to
743 @var{number}. Increasing this value can reduce the length of time it takes the
744 assembler to perform its tasks, at the expense of increasing the assembler's
745 memory requirements. Similarly reducing this value can reduce the memory
746 requirements at the expense of speed.
748 @item --reduce-memory-overheads
749 This option reduces GAS's memory requirements, at the expense of making the
750 assembly processes slower. Currently this switch is a synonym for
751 @samp{--hash-size=4051}, but in the future it may have other effects as well.
754 Print the maximum space (in bytes) and total time (in seconds) used by
757 @item --strip-local-absolute
758 Remove local absolute symbols from the outgoing symbol table.
762 Print the @command{as} version.
765 Print the @command{as} version and exit.
769 Suppress warning messages.
771 @item --fatal-warnings
772 Treat warnings as errors.
775 Don't suppress warning messages or treat them as errors.
784 Generate an object file even after errors.
786 @item -- | @var{files} @dots{}
787 Standard input, or source files to assemble.
795 @xref{AArch64 Options}, for the options available when @value{AS} is configured
796 for the 64-bit mode of the ARM Architecture (AArch64).
801 The following options are available when @value{AS} is configured for the
802 64-bit mode of the ARM Architecture (AArch64).
805 @include c-aarch64.texi
806 @c ended inside the included file
814 @xref{Alpha Options}, for the options available when @value{AS} is configured
815 for an Alpha processor.
820 The following options are available when @value{AS} is configured for an Alpha
824 @include c-alpha.texi
825 @c ended inside the included file
832 The following options are available when @value{AS} is configured for
837 This option selects the core processor variant.
839 Select either big-endian (-EB) or little-endian (-EL) output.
844 The following options are available when @value{AS} is configured for the ARM
848 @item -mcpu=@var{processor}[+@var{extension}@dots{}]
849 Specify which ARM processor variant is the target.
850 @item -march=@var{architecture}[+@var{extension}@dots{}]
851 Specify which ARM architecture variant is used by the target.
852 @item -mfpu=@var{floating-point-format}
853 Select which Floating Point architecture is the target.
854 @item -mfloat-abi=@var{abi}
855 Select which floating point ABI is in use.
857 Enable Thumb only instruction decoding.
858 @item -mapcs-32 | -mapcs-26 | -mapcs-float | -mapcs-reentrant
859 Select which procedure calling convention is in use.
861 Select either big-endian (-EB) or little-endian (-EL) output.
862 @item -mthumb-interwork
863 Specify that the code has been generated with interworking between Thumb and
866 Turns on CodeComposer Studio assembly syntax compatibility mode.
868 Specify that PIC code has been generated.
876 @xref{Blackfin Options}, for the options available when @value{AS} is
877 configured for the Blackfin processor family.
882 The following options are available when @value{AS} is configured for
883 the Blackfin processor family.
887 @c ended inside the included file
894 See the info pages for documentation of the CRIS-specific options.
898 The following options are available when @value{AS} is configured for
901 @cindex D10V optimization
902 @cindex optimization, D10V
904 Optimize output by parallelizing instructions.
909 The following options are available when @value{AS} is configured for a D30V
912 @cindex D30V optimization
913 @cindex optimization, D30V
915 Optimize output by parallelizing instructions.
919 Warn when nops are generated.
921 @cindex D30V nops after 32-bit multiply
923 Warn when a nop after a 32-bit multiply instruction is generated.
929 The following options are available when @value{AS} is configured for the
930 Adapteva EPIPHANY series.
933 @xref{Epiphany Options}, for the options available when @value{AS} is
934 configured for an Epiphany processor.
939 The following options are available when @value{AS} is configured for
940 an Epiphany processor.
943 @include c-epiphany.texi
944 @c ended inside the included file
952 @xref{H8/300 Options}, for the options available when @value{AS} is configured
953 for an H8/300 processor.
958 The following options are available when @value{AS} is configured for an H8/300
962 @include c-h8300.texi
963 @c ended inside the included file
971 @xref{i386-Options}, for the options available when @value{AS} is
972 configured for an i386 processor.
977 The following options are available when @value{AS} is configured for
982 @c ended inside the included file
989 The following options are available when @value{AS} is configured for the
990 Intel 80960 processor.
993 @item -ACA | -ACA_A | -ACB | -ACC | -AKA | -AKB | -AKC | -AMC
994 Specify which variant of the 960 architecture is the target.
997 Add code to collect statistics about branches taken.
1000 Do not alter compare-and-branch instructions for long displacements;
1007 The following options are available when @value{AS} is configured for the
1013 Specifies that the extended IP2022 instructions are allowed.
1016 Restores the default behaviour, which restricts the permitted instructions to
1017 just the basic IP2022 ones.
1023 The following options are available when @value{AS} is configured for the
1024 Renesas M32C and M16C processors.
1029 Assemble M32C instructions.
1032 Assemble M16C instructions (the default).
1035 Enable support for link-time relaxations.
1038 Support H'00 style hex constants in addition to 0x00 style.
1044 The following options are available when @value{AS} is configured for the
1045 Renesas M32R (formerly Mitsubishi M32R) series.
1050 Specify which processor in the M32R family is the target. The default
1051 is normally the M32R, but this option changes it to the M32RX.
1053 @item --warn-explicit-parallel-conflicts or --Wp
1054 Produce warning messages when questionable parallel constructs are
1057 @item --no-warn-explicit-parallel-conflicts or --Wnp
1058 Do not produce warning messages when questionable parallel constructs are
1065 The following options are available when @value{AS} is configured for the
1066 Motorola 68000 series.
1071 Shorten references to undefined symbols, to one word instead of two.
1073 @item -m68000 | -m68008 | -m68010 | -m68020 | -m68030
1074 @itemx | -m68040 | -m68060 | -m68302 | -m68331 | -m68332
1075 @itemx | -m68333 | -m68340 | -mcpu32 | -m5200
1076 Specify what processor in the 68000 family is the target. The default
1077 is normally the 68020, but this can be changed at configuration time.
1079 @item -m68881 | -m68882 | -mno-68881 | -mno-68882
1080 The target machine does (or does not) have a floating-point coprocessor.
1081 The default is to assume a coprocessor for 68020, 68030, and cpu32. Although
1082 the basic 68000 is not compatible with the 68881, a combination of the
1083 two can be specified, since it's possible to do emulation of the
1084 coprocessor instructions with the main processor.
1086 @item -m68851 | -mno-68851
1087 The target machine does (or does not) have a memory-management
1088 unit coprocessor. The default is to assume an MMU for 68020 and up.
1096 @xref{Nios II Options}, for the options available when @value{AS} is configured
1097 for an Altera Nios II processor.
1101 @c man begin OPTIONS
1102 The following options are available when @value{AS} is configured for an
1103 Altera Nios II processor.
1105 @c man begin INCLUDE
1106 @include c-nios2.texi
1107 @c ended inside the included file
1113 For details about the PDP-11 machine dependent features options,
1114 see @ref{PDP-11-Options}.
1117 @item -mpic | -mno-pic
1118 Generate position-independent (or position-dependent) code. The
1119 default is @option{-mpic}.
1122 @itemx -mall-extensions
1123 Enable all instruction set extensions. This is the default.
1125 @item -mno-extensions
1126 Disable all instruction set extensions.
1128 @item -m@var{extension} | -mno-@var{extension}
1129 Enable (or disable) a particular instruction set extension.
1132 Enable the instruction set extensions supported by a particular CPU, and
1133 disable all other extensions.
1135 @item -m@var{machine}
1136 Enable the instruction set extensions supported by a particular machine
1137 model, and disable all other extensions.
1143 The following options are available when @value{AS} is configured for
1144 a picoJava processor.
1148 @cindex PJ endianness
1149 @cindex endianness, PJ
1150 @cindex big endian output, PJ
1152 Generate ``big endian'' format output.
1154 @cindex little endian output, PJ
1156 Generate ``little endian'' format output.
1162 The following options are available when @value{AS} is configured for the
1163 Motorola 68HC11 or 68HC12 series.
1167 @item -m68hc11 | -m68hc12 | -m68hcs12 | -mm9s12x | -mm9s12xg
1168 Specify what processor is the target. The default is
1169 defined by the configuration option when building the assembler.
1171 @item --xgate-ramoffset
1172 Instruct the linker to offset RAM addresses from S12X address space into
1173 XGATE address space.
1176 Specify to use the 16-bit integer ABI.
1179 Specify to use the 32-bit integer ABI.
1181 @item -mshort-double
1182 Specify to use the 32-bit double ABI.
1185 Specify to use the 64-bit double ABI.
1187 @item --force-long-branches
1188 Relative branches are turned into absolute ones. This concerns
1189 conditional branches, unconditional branches and branches to a
1192 @item -S | --short-branches
1193 Do not turn relative branches into absolute ones
1194 when the offset is out of range.
1196 @item --strict-direct-mode
1197 Do not turn the direct addressing mode into extended addressing mode
1198 when the instruction does not support direct addressing mode.
1200 @item --print-insn-syntax
1201 Print the syntax of instruction in case of error.
1203 @item --print-opcodes
1204 Print the list of instructions with syntax and then exit.
1206 @item --generate-example
1207 Print an example of instruction for each possible instruction and then exit.
1208 This option is only useful for testing @command{@value{AS}}.
1214 The following options are available when @command{@value{AS}} is configured
1215 for the SPARC architecture:
1218 @item -Av6 | -Av7 | -Av8 | -Asparclet | -Asparclite
1219 @itemx -Av8plus | -Av8plusa | -Av9 | -Av9a
1220 Explicitly select a variant of the SPARC architecture.
1222 @samp{-Av8plus} and @samp{-Av8plusa} select a 32 bit environment.
1223 @samp{-Av9} and @samp{-Av9a} select a 64 bit environment.
1225 @samp{-Av8plusa} and @samp{-Av9a} enable the SPARC V9 instruction set with
1226 UltraSPARC extensions.
1228 @item -xarch=v8plus | -xarch=v8plusa
1229 For compatibility with the Solaris v9 assembler. These options are
1230 equivalent to -Av8plus and -Av8plusa, respectively.
1233 Warn when the assembler switches to another architecture.
1238 The following options are available when @value{AS} is configured for the 'c54x
1243 Enable extended addressing mode. All addresses and relocations will assume
1244 extended addressing (usually 23 bits).
1245 @item -mcpu=@var{CPU_VERSION}
1246 Sets the CPU version being compiled for.
1247 @item -merrors-to-file @var{FILENAME}
1248 Redirect error output to a file, for broken systems which don't support such
1249 behaviour in the shell.
1254 The following options are available when @value{AS} is configured for
1259 This option sets the largest size of an object that can be referenced
1260 implicitly with the @code{gp} register. It is only accepted for targets that
1261 use ECOFF format, such as a DECstation running Ultrix. The default value is 8.
1263 @cindex MIPS endianness
1264 @cindex endianness, MIPS
1265 @cindex big endian output, MIPS
1267 Generate ``big endian'' format output.
1269 @cindex little endian output, MIPS
1271 Generate ``little endian'' format output.
1289 Generate code for a particular MIPS Instruction Set Architecture level.
1290 @samp{-mips1} is an alias for @samp{-march=r3000}, @samp{-mips2} is an
1291 alias for @samp{-march=r6000}, @samp{-mips3} is an alias for
1292 @samp{-march=r4000} and @samp{-mips4} is an alias for @samp{-march=r8000}.
1293 @samp{-mips5}, @samp{-mips32}, @samp{-mips32r2}, @samp{-mips32r3},
1294 @samp{-mips32r5}, @samp{-mips32r6}, @samp{-mips64}, @samp{-mips64r2},
1295 @samp{-mips64r3}, @samp{-mips64r5}, and @samp{-mips64r6} correspond to generic
1296 MIPS V, MIPS32, MIPS32 Release 2, MIPS32 Release 3, MIPS32 Release 5, MIPS32
1297 Release 6, MIPS64, MIPS64 Release 2, MIPS64 Release 3, MIPS64 Release 5, and
1298 MIPS64 Release 6 ISA processors, respectively.
1300 @item -march=@var{cpu}
1301 Generate code for a particular MIPS CPU.
1303 @item -mtune=@var{cpu}
1304 Schedule and tune for a particular MIPS CPU.
1308 Cause nops to be inserted if the read of the destination register
1309 of an mfhi or mflo instruction occurs in the following two instructions.
1312 @itemx -mno-fix-rm7000
1313 Cause nops to be inserted if a dmult or dmultu instruction is
1314 followed by a load instruction.
1318 Cause stabs-style debugging output to go into an ECOFF-style .mdebug
1319 section instead of the standard ELF .stabs sections.
1323 Control generation of @code{.pdr} sections.
1327 The register sizes are normally inferred from the ISA and ABI, but these
1328 flags force a certain group of registers to be treated as 32 bits wide at
1329 all times. @samp{-mgp32} controls the size of general-purpose registers
1330 and @samp{-mfp32} controls the size of floating-point registers.
1334 The register sizes are normally inferred from the ISA and ABI, but these
1335 flags force a certain group of registers to be treated as 64 bits wide at
1336 all times. @samp{-mgp64} controls the size of general-purpose registers
1337 and @samp{-mfp64} controls the size of floating-point registers.
1340 The register sizes are normally inferred from the ISA and ABI, but using
1341 this flag in combination with @samp{-mabi=32} enables an ABI variant
1342 which will operate correctly with floating-point registers which are
1346 @itemx -mno-odd-spreg
1347 Enable use of floating-point operations on odd-numbered single-precision
1348 registers when supported by the ISA. @samp{-mfpxx} implies
1349 @samp{-mno-odd-spreg}, otherwise the default is @samp{-modd-spreg}.
1353 Generate code for the MIPS 16 processor. This is equivalent to putting
1354 @code{.set mips16} at the start of the assembly file. @samp{-no-mips16}
1355 turns off this option.
1358 @itemx -mno-micromips
1359 Generate code for the microMIPS processor. This is equivalent to putting
1360 @code{.set micromips} at the start of the assembly file. @samp{-mno-micromips}
1361 turns off this option. This is equivalent to putting @code{.set nomicromips}
1362 at the start of the assembly file.
1365 @itemx -mno-smartmips
1366 Enables the SmartMIPS extension to the MIPS32 instruction set. This is
1367 equivalent to putting @code{.set smartmips} at the start of the assembly file.
1368 @samp{-mno-smartmips} turns off this option.
1372 Generate code for the MIPS-3D Application Specific Extension.
1373 This tells the assembler to accept MIPS-3D instructions.
1374 @samp{-no-mips3d} turns off this option.
1378 Generate code for the MDMX Application Specific Extension.
1379 This tells the assembler to accept MDMX instructions.
1380 @samp{-no-mdmx} turns off this option.
1384 Generate code for the DSP Release 1 Application Specific Extension.
1385 This tells the assembler to accept DSP Release 1 instructions.
1386 @samp{-mno-dsp} turns off this option.
1390 Generate code for the DSP Release 2 Application Specific Extension.
1391 This option implies -mdsp.
1392 This tells the assembler to accept DSP Release 2 instructions.
1393 @samp{-mno-dspr2} turns off this option.
1397 Generate code for the MIPS SIMD Architecture Extension.
1398 This tells the assembler to accept MSA instructions.
1399 @samp{-mno-msa} turns off this option.
1403 Generate code for the MIPS eXtended Physical Address (XPA) Extension.
1404 This tells the assembler to accept XPA instructions.
1405 @samp{-mno-xpa} turns off this option.
1409 Generate code for the MT Application Specific Extension.
1410 This tells the assembler to accept MT instructions.
1411 @samp{-mno-mt} turns off this option.
1415 Generate code for the MCU Application Specific Extension.
1416 This tells the assembler to accept MCU instructions.
1417 @samp{-mno-mcu} turns off this option.
1421 Only use 32-bit instruction encodings when generating code for the
1422 microMIPS processor. This option inhibits the use of any 16-bit
1423 instructions. This is equivalent to putting @code{.set insn32} at
1424 the start of the assembly file. @samp{-mno-insn32} turns off this
1425 option. This is equivalent to putting @code{.set noinsn32} at the
1426 start of the assembly file. By default @samp{-mno-insn32} is
1427 selected, allowing all instructions to be used.
1429 @item --construct-floats
1430 @itemx --no-construct-floats
1431 The @samp{--no-construct-floats} option disables the construction of
1432 double width floating point constants by loading the two halves of the
1433 value into the two single width floating point registers that make up
1434 the double width register. By default @samp{--construct-floats} is
1435 selected, allowing construction of these floating point constants.
1437 @item --relax-branch
1438 @itemx --no-relax-branch
1439 The @samp{--relax-branch} option enables the relaxation of out-of-range
1440 branches. By default @samp{--no-relax-branch} is selected, causing any
1441 out-of-range branches to produce an error.
1443 @item -mnan=@var{encoding}
1444 Select between the IEEE 754-2008 (@option{-mnan=2008}) or the legacy
1445 (@option{-mnan=legacy}) NaN encoding format. The latter is the default.
1448 @item --emulation=@var{name}
1449 This option was formerly used to switch between ELF and ECOFF output
1450 on targets like IRIX 5 that supported both. MIPS ECOFF support was
1451 removed in GAS 2.24, so the option now serves little purpose.
1452 It is retained for backwards compatibility.
1454 The available configuration names are: @samp{mipself}, @samp{mipslelf} and
1455 @samp{mipsbelf}. Choosing @samp{mipself} now has no effect, since the output
1456 is always ELF. @samp{mipslelf} and @samp{mipsbelf} select little- and
1457 big-endian output respectively, but @samp{-EL} and @samp{-EB} are now the
1458 preferred options instead.
1461 @command{@value{AS}} ignores this option. It is accepted for compatibility with
1468 Control how to deal with multiplication overflow and division by zero.
1469 @samp{--trap} or @samp{--no-break} (which are synonyms) take a trap exception
1470 (and only work for Instruction Set Architecture level 2 and higher);
1471 @samp{--break} or @samp{--no-trap} (also synonyms, and the default) take a
1475 When this option is used, @command{@value{AS}} will issue a warning every
1476 time it generates a nop instruction from a macro.
1481 The following options are available when @value{AS} is configured for
1487 Enable or disable the JSRI to BSR transformation. By default this is enabled.
1488 The command line option @samp{-nojsri2bsr} can be used to disable it.
1492 Enable or disable the silicon filter behaviour. By default this is disabled.
1493 The default can be overridden by the @samp{-sifilter} command line option.
1496 Alter jump instructions for long displacements.
1498 @item -mcpu=[210|340]
1499 Select the cpu type on the target hardware. This controls which instructions
1503 Assemble for a big endian target.
1506 Assemble for a little endian target.
1515 @xref{Meta Options}, for the options available when @value{AS} is configured
1516 for a Meta processor.
1520 @c man begin OPTIONS
1521 The following options are available when @value{AS} is configured for a
1524 @c man begin INCLUDE
1525 @include c-metag.texi
1526 @c ended inside the included file
1531 @c man begin OPTIONS
1533 See the info pages for documentation of the MMIX-specific options.
1539 @xref{NDS32 Options}, for the options available when @value{AS} is configured
1540 for a NDS32 processor.
1542 @c ended inside the included file
1546 @c man begin OPTIONS
1547 The following options are available when @value{AS} is configured for a
1550 @c man begin INCLUDE
1551 @include c-nds32.texi
1552 @c ended inside the included file
1559 @xref{PowerPC-Opts}, for the options available when @value{AS} is configured
1560 for a PowerPC processor.
1564 @c man begin OPTIONS
1565 The following options are available when @value{AS} is configured for a
1568 @c man begin INCLUDE
1570 @c ended inside the included file
1575 @c man begin OPTIONS
1577 See the info pages for documentation of the RX-specific options.
1581 The following options are available when @value{AS} is configured for the s390
1587 Select the word size, either 31/32 bits or 64 bits.
1590 Select the architecture mode, either the Enterprise System
1591 Architecture (esa) or the z/Architecture mode (zarch).
1592 @item -march=@var{processor}
1593 Specify which s390 processor variant is the target, @samp{g6}, @samp{g6},
1594 @samp{z900}, @samp{z990}, @samp{z9-109}, @samp{z9-ec}, @samp{z10},
1595 @samp{z196}, @samp{zEC12}, or @samp{z13}.
1597 @itemx -mno-regnames
1598 Allow or disallow symbolic names for registers.
1599 @item -mwarn-areg-zero
1600 Warn whenever the operand for a base or index register has been specified
1601 but evaluates to zero.
1609 @xref{TIC6X Options}, for the options available when @value{AS} is configured
1610 for a TMS320C6000 processor.
1614 @c man begin OPTIONS
1615 The following options are available when @value{AS} is configured for a
1616 TMS320C6000 processor.
1618 @c man begin INCLUDE
1619 @include c-tic6x.texi
1620 @c ended inside the included file
1628 @xref{TILE-Gx Options}, for the options available when @value{AS} is configured
1629 for a TILE-Gx processor.
1633 @c man begin OPTIONS
1634 The following options are available when @value{AS} is configured for a TILE-Gx
1637 @c man begin INCLUDE
1638 @include c-tilegx.texi
1639 @c ended inside the included file
1647 @xref{Visium Options}, for the options available when @value{AS} is configured
1648 for a Visium processor.
1652 @c man begin OPTIONS
1653 The following option is available when @value{AS} is configured for a Visium
1656 @c man begin INCLUDE
1657 @include c-visium.texi
1658 @c ended inside the included file
1666 @xref{Xtensa Options}, for the options available when @value{AS} is configured
1667 for an Xtensa processor.
1671 @c man begin OPTIONS
1672 The following options are available when @value{AS} is configured for an
1675 @c man begin INCLUDE
1676 @include c-xtensa.texi
1677 @c ended inside the included file
1682 @c man begin OPTIONS
1685 The following options are available when @value{AS} is configured for
1686 a Z80 family processor.
1689 Assemble for Z80 processor.
1691 Assemble for R800 processor.
1692 @item -ignore-undocumented-instructions
1694 Assemble undocumented Z80 instructions that also work on R800 without warning.
1695 @item -ignore-unportable-instructions
1697 Assemble all undocumented Z80 instructions without warning.
1698 @item -warn-undocumented-instructions
1700 Issue a warning for undocumented Z80 instructions that also work on R800.
1701 @item -warn-unportable-instructions
1703 Issue a warning for undocumented Z80 instructions that do not work on R800.
1704 @item -forbid-undocumented-instructions
1706 Treat all undocumented instructions as errors.
1707 @item -forbid-unportable-instructions
1709 Treat undocumented Z80 instructions that do not work on R800 as errors.
1716 * Manual:: Structure of this Manual
1717 * GNU Assembler:: The GNU Assembler
1718 * Object Formats:: Object File Formats
1719 * Command Line:: Command Line
1720 * Input Files:: Input Files
1721 * Object:: Output (Object) File
1722 * Errors:: Error and Warning Messages
1726 @section Structure of this Manual
1728 @cindex manual, structure and purpose
1729 This manual is intended to describe what you need to know to use
1730 @sc{gnu} @command{@value{AS}}. We cover the syntax expected in source files, including
1731 notation for symbols, constants, and expressions; the directives that
1732 @command{@value{AS}} understands; and of course how to invoke @command{@value{AS}}.
1735 We also cover special features in the @value{TARGET}
1736 configuration of @command{@value{AS}}, including assembler directives.
1739 This manual also describes some of the machine-dependent features of
1740 various flavors of the assembler.
1743 @cindex machine instructions (not covered)
1744 On the other hand, this manual is @emph{not} intended as an introduction
1745 to programming in assembly language---let alone programming in general!
1746 In a similar vein, we make no attempt to introduce the machine
1747 architecture; we do @emph{not} describe the instruction set, standard
1748 mnemonics, registers or addressing modes that are standard to a
1749 particular architecture.
1751 You may want to consult the manufacturer's
1752 machine architecture manual for this information.
1756 For information on the H8/300 machine instruction set, see @cite{H8/300
1757 Series Programming Manual}. For the H8/300H, see @cite{H8/300H Series
1758 Programming Manual} (Renesas).
1761 For information on the Renesas (formerly Hitachi) / SuperH SH machine instruction set,
1762 see @cite{SH-Microcomputer User's Manual} (Renesas) or
1763 @cite{SH-4 32-bit CPU Core Architecture} (SuperH) and
1764 @cite{SuperH (SH) 64-Bit RISC Series} (SuperH).
1767 For information on the Z8000 machine instruction set, see @cite{Z8000 CPU Technical Manual}
1771 @c I think this is premature---doc@cygnus.com, 17jan1991
1773 Throughout this manual, we assume that you are running @dfn{GNU},
1774 the portable operating system from the @dfn{Free Software
1775 Foundation, Inc.}. This restricts our attention to certain kinds of
1776 computer (in particular, the kinds of computers that @sc{gnu} can run on);
1777 once this assumption is granted examples and definitions need less
1780 @command{@value{AS}} is part of a team of programs that turn a high-level
1781 human-readable series of instructions into a low-level
1782 computer-readable series of instructions. Different versions of
1783 @command{@value{AS}} are used for different kinds of computer.
1786 @c There used to be a section "Terminology" here, which defined
1787 @c "contents", "byte", "word", and "long". Defining "word" to any
1788 @c particular size is confusing when the .word directive may generate 16
1789 @c bits on one machine and 32 bits on another; in general, for the user
1790 @c version of this manual, none of these terms seem essential to define.
1791 @c They were used very little even in the former draft of the manual;
1792 @c this draft makes an effort to avoid them (except in names of
1796 @section The GNU Assembler
1798 @c man begin DESCRIPTION
1800 @sc{gnu} @command{as} is really a family of assemblers.
1802 This manual describes @command{@value{AS}}, a member of that family which is
1803 configured for the @value{TARGET} architectures.
1805 If you use (or have used) the @sc{gnu} assembler on one architecture, you
1806 should find a fairly similar environment when you use it on another
1807 architecture. Each version has much in common with the others,
1808 including object file formats, most assembler directives (often called
1809 @dfn{pseudo-ops}) and assembler syntax.@refill
1811 @cindex purpose of @sc{gnu} assembler
1812 @command{@value{AS}} is primarily intended to assemble the output of the
1813 @sc{gnu} C compiler @code{@value{GCC}} for use by the linker
1814 @code{@value{LD}}. Nevertheless, we've tried to make @command{@value{AS}}
1815 assemble correctly everything that other assemblers for the same
1816 machine would assemble.
1818 Any exceptions are documented explicitly (@pxref{Machine Dependencies}).
1821 @c This remark should appear in generic version of manual; assumption
1822 @c here is that generic version sets M680x0.
1823 This doesn't mean @command{@value{AS}} always uses the same syntax as another
1824 assembler for the same architecture; for example, we know of several
1825 incompatible versions of 680x0 assembly language syntax.
1830 Unlike older assemblers, @command{@value{AS}} is designed to assemble a source
1831 program in one pass of the source file. This has a subtle impact on the
1832 @kbd{.org} directive (@pxref{Org,,@code{.org}}).
1834 @node Object Formats
1835 @section Object File Formats
1837 @cindex object file format
1838 The @sc{gnu} assembler can be configured to produce several alternative
1839 object file formats. For the most part, this does not affect how you
1840 write assembly language programs; but directives for debugging symbols
1841 are typically different in different file formats. @xref{Symbol
1842 Attributes,,Symbol Attributes}.
1845 For the @value{TARGET} target, @command{@value{AS}} is configured to produce
1846 @value{OBJ-NAME} format object files.
1848 @c The following should exhaust all configs that set MULTI-OBJ, ideally
1850 On the @value{TARGET}, @command{@value{AS}} can be configured to produce either
1851 @code{b.out} or COFF format object files.
1854 On the @value{TARGET}, @command{@value{AS}} can be configured to produce either
1855 SOM or ELF format object files.
1860 @section Command Line
1862 @cindex command line conventions
1864 After the program name @command{@value{AS}}, the command line may contain
1865 options and file names. Options may appear in any order, and may be
1866 before, after, or between file names. The order of file names is
1869 @cindex standard input, as input file
1871 @file{--} (two hyphens) by itself names the standard input file
1872 explicitly, as one of the files for @command{@value{AS}} to assemble.
1874 @cindex options, command line
1875 Except for @samp{--} any command line argument that begins with a
1876 hyphen (@samp{-}) is an option. Each option changes the behavior of
1877 @command{@value{AS}}. No option changes the way another option works. An
1878 option is a @samp{-} followed by one or more letters; the case of
1879 the letter is important. All options are optional.
1881 Some options expect exactly one file name to follow them. The file
1882 name may either immediately follow the option's letter (compatible
1883 with older assemblers) or it may be the next command argument (@sc{gnu}
1884 standard). These two command lines are equivalent:
1887 @value{AS} -o my-object-file.o mumble.s
1888 @value{AS} -omy-object-file.o mumble.s
1892 @section Input Files
1895 @cindex source program
1896 @cindex files, input
1897 We use the phrase @dfn{source program}, abbreviated @dfn{source}, to
1898 describe the program input to one run of @command{@value{AS}}. The program may
1899 be in one or more files; how the source is partitioned into files
1900 doesn't change the meaning of the source.
1902 @c I added "con" prefix to "catenation" just to prove I can overcome my
1903 @c APL training... doc@cygnus.com
1904 The source program is a concatenation of the text in all the files, in the
1907 @c man begin DESCRIPTION
1908 Each time you run @command{@value{AS}} it assembles exactly one source
1909 program. The source program is made up of one or more files.
1910 (The standard input is also a file.)
1912 You give @command{@value{AS}} a command line that has zero or more input file
1913 names. The input files are read (from left file name to right). A
1914 command line argument (in any position) that has no special meaning
1915 is taken to be an input file name.
1917 If you give @command{@value{AS}} no file names it attempts to read one input file
1918 from the @command{@value{AS}} standard input, which is normally your terminal. You
1919 may have to type @key{ctl-D} to tell @command{@value{AS}} there is no more program
1922 Use @samp{--} if you need to explicitly name the standard input file
1923 in your command line.
1925 If the source is empty, @command{@value{AS}} produces a small, empty object
1930 @subheading Filenames and Line-numbers
1932 @cindex input file linenumbers
1933 @cindex line numbers, in input files
1934 There are two ways of locating a line in the input file (or files) and
1935 either may be used in reporting error messages. One way refers to a line
1936 number in a physical file; the other refers to a line number in a
1937 ``logical'' file. @xref{Errors, ,Error and Warning Messages}.
1939 @dfn{Physical files} are those files named in the command line given
1940 to @command{@value{AS}}.
1942 @dfn{Logical files} are simply names declared explicitly by assembler
1943 directives; they bear no relation to physical files. Logical file names help
1944 error messages reflect the original source file, when @command{@value{AS}} source
1945 is itself synthesized from other files. @command{@value{AS}} understands the
1946 @samp{#} directives emitted by the @code{@value{GCC}} preprocessor. See also
1947 @ref{File,,@code{.file}}.
1950 @section Output (Object) File
1956 Every time you run @command{@value{AS}} it produces an output file, which is
1957 your assembly language program translated into numbers. This file
1958 is the object file. Its default name is
1966 @code{b.out} when @command{@value{AS}} is configured for the Intel 80960.
1968 You can give it another name by using the @option{-o} option. Conventionally,
1969 object file names end with @file{.o}. The default name is used for historical
1970 reasons: older assemblers were capable of assembling self-contained programs
1971 directly into a runnable program. (For some formats, this isn't currently
1972 possible, but it can be done for the @code{a.out} format.)
1976 The object file is meant for input to the linker @code{@value{LD}}. It contains
1977 assembled program code, information to help @code{@value{LD}} integrate
1978 the assembled program into a runnable file, and (optionally) symbolic
1979 information for the debugger.
1981 @c link above to some info file(s) like the description of a.out.
1982 @c don't forget to describe @sc{gnu} info as well as Unix lossage.
1985 @section Error and Warning Messages
1987 @c man begin DESCRIPTION
1989 @cindex error messages
1990 @cindex warning messages
1991 @cindex messages from assembler
1992 @command{@value{AS}} may write warnings and error messages to the standard error
1993 file (usually your terminal). This should not happen when a compiler
1994 runs @command{@value{AS}} automatically. Warnings report an assumption made so
1995 that @command{@value{AS}} could keep assembling a flawed program; errors report a
1996 grave problem that stops the assembly.
2000 @cindex format of warning messages
2001 Warning messages have the format
2004 file_name:@b{NNN}:Warning Message Text
2008 @cindex line numbers, in warnings/errors
2009 (where @b{NNN} is a line number). If a logical file name has been given
2010 (@pxref{File,,@code{.file}}) it is used for the filename, otherwise the name of
2011 the current input file is used. If a logical line number was given
2013 (@pxref{Line,,@code{.line}})
2015 then it is used to calculate the number printed,
2016 otherwise the actual line in the current source file is printed. The
2017 message text is intended to be self explanatory (in the grand Unix
2020 @cindex format of error messages
2021 Error messages have the format
2023 file_name:@b{NNN}:FATAL:Error Message Text
2025 The file name and line number are derived as for warning
2026 messages. The actual message text may be rather less explanatory
2027 because many of them aren't supposed to happen.
2030 @chapter Command-Line Options
2032 @cindex options, all versions of assembler
2033 This chapter describes command-line options available in @emph{all}
2034 versions of the @sc{gnu} assembler; see @ref{Machine Dependencies},
2035 for options specific
2037 to the @value{TARGET} target.
2040 to particular machine architectures.
2043 @c man begin DESCRIPTION
2045 If you are invoking @command{@value{AS}} via the @sc{gnu} C compiler,
2046 you can use the @samp{-Wa} option to pass arguments through to the assembler.
2047 The assembler arguments must be separated from each other (and the @samp{-Wa})
2048 by commas. For example:
2051 gcc -c -g -O -Wa,-alh,-L file.c
2055 This passes two options to the assembler: @samp{-alh} (emit a listing to
2056 standard output with high-level and assembly source) and @samp{-L} (retain
2057 local symbols in the symbol table).
2059 Usually you do not need to use this @samp{-Wa} mechanism, since many compiler
2060 command-line options are automatically passed to the assembler by the compiler.
2061 (You can call the @sc{gnu} compiler driver with the @samp{-v} option to see
2062 precisely what options it passes to each compilation pass, including the
2068 * a:: -a[cdghlns] enable listings
2069 * alternate:: --alternate enable alternate macro syntax
2070 * D:: -D for compatibility
2071 * f:: -f to work faster
2072 * I:: -I for .include search path
2073 @ifclear DIFF-TBL-KLUGE
2074 * K:: -K for compatibility
2076 @ifset DIFF-TBL-KLUGE
2077 * K:: -K for difference tables
2080 * L:: -L to retain local symbols
2081 * listing:: --listing-XXX to configure listing output
2082 * M:: -M or --mri to assemble in MRI compatibility mode
2083 * MD:: --MD for dependency tracking
2084 * o:: -o to name the object file
2085 * R:: -R to join data and text sections
2086 * statistics:: --statistics to see statistics about assembly
2087 * traditional-format:: --traditional-format for compatible output
2088 * v:: -v to announce version
2089 * W:: -W, --no-warn, --warn, --fatal-warnings to control warnings
2090 * Z:: -Z to make object file even after errors
2094 @section Enable Listings: @option{-a[cdghlns]}
2104 @cindex listings, enabling
2105 @cindex assembly listings, enabling
2107 These options enable listing output from the assembler. By itself,
2108 @samp{-a} requests high-level, assembly, and symbols listing.
2109 You can use other letters to select specific options for the list:
2110 @samp{-ah} requests a high-level language listing,
2111 @samp{-al} requests an output-program assembly listing, and
2112 @samp{-as} requests a symbol table listing.
2113 High-level listings require that a compiler debugging option like
2114 @samp{-g} be used, and that assembly listings (@samp{-al}) be requested
2117 Use the @samp{-ag} option to print a first section with general assembly
2118 information, like @value{AS} version, switches passed, or time stamp.
2120 Use the @samp{-ac} option to omit false conditionals from a listing. Any lines
2121 which are not assembled because of a false @code{.if} (or @code{.ifdef}, or any
2122 other conditional), or a true @code{.if} followed by an @code{.else}, will be
2123 omitted from the listing.
2125 Use the @samp{-ad} option to omit debugging directives from the
2128 Once you have specified one of these options, you can further control
2129 listing output and its appearance using the directives @code{.list},
2130 @code{.nolist}, @code{.psize}, @code{.eject}, @code{.title}, and
2132 The @samp{-an} option turns off all forms processing.
2133 If you do not request listing output with one of the @samp{-a} options, the
2134 listing-control directives have no effect.
2136 The letters after @samp{-a} may be combined into one option,
2137 @emph{e.g.}, @samp{-aln}.
2139 Note if the assembler source is coming from the standard input (e.g.,
2141 is being created by @code{@value{GCC}} and the @samp{-pipe} command line switch
2142 is being used) then the listing will not contain any comments or preprocessor
2143 directives. This is because the listing code buffers input source lines from
2144 stdin only after they have been preprocessed by the assembler. This reduces
2145 memory usage and makes the code more efficient.
2148 @section @option{--alternate}
2151 Begin in alternate macro mode, see @ref{Altmacro,,@code{.altmacro}}.
2154 @section @option{-D}
2157 This option has no effect whatsoever, but it is accepted to make it more
2158 likely that scripts written for other assemblers also work with
2159 @command{@value{AS}}.
2162 @section Work Faster: @option{-f}
2165 @cindex trusted compiler
2166 @cindex faster processing (@option{-f})
2167 @samp{-f} should only be used when assembling programs written by a
2168 (trusted) compiler. @samp{-f} stops the assembler from doing whitespace
2169 and comment preprocessing on
2170 the input file(s) before assembling them. @xref{Preprocessing,
2174 @emph{Warning:} if you use @samp{-f} when the files actually need to be
2175 preprocessed (if they contain comments, for example), @command{@value{AS}} does
2180 @section @code{.include} Search Path: @option{-I} @var{path}
2182 @kindex -I @var{path}
2183 @cindex paths for @code{.include}
2184 @cindex search path for @code{.include}
2185 @cindex @code{include} directive search path
2186 Use this option to add a @var{path} to the list of directories
2187 @command{@value{AS}} searches for files specified in @code{.include}
2188 directives (@pxref{Include,,@code{.include}}). You may use @option{-I} as
2189 many times as necessary to include a variety of paths. The current
2190 working directory is always searched first; after that, @command{@value{AS}}
2191 searches any @samp{-I} directories in the same order as they were
2192 specified (left to right) on the command line.
2195 @section Difference Tables: @option{-K}
2198 @ifclear DIFF-TBL-KLUGE
2199 On the @value{TARGET} family, this option is allowed, but has no effect. It is
2200 permitted for compatibility with the @sc{gnu} assembler on other platforms,
2201 where it can be used to warn when the assembler alters the machine code
2202 generated for @samp{.word} directives in difference tables. The @value{TARGET}
2203 family does not have the addressing limitations that sometimes lead to this
2204 alteration on other platforms.
2207 @ifset DIFF-TBL-KLUGE
2208 @cindex difference tables, warning
2209 @cindex warning for altered difference tables
2210 @command{@value{AS}} sometimes alters the code emitted for directives of the
2211 form @samp{.word @var{sym1}-@var{sym2}}. @xref{Word,,@code{.word}}.
2212 You can use the @samp{-K} option if you want a warning issued when this
2217 @section Include Local Symbols: @option{-L}
2220 @cindex local symbols, retaining in output
2221 Symbols beginning with system-specific local label prefixes, typically
2222 @samp{.L} for ELF systems or @samp{L} for traditional a.out systems, are
2223 called @dfn{local symbols}. @xref{Symbol Names}. Normally you do not see
2224 such symbols when debugging, because they are intended for the use of
2225 programs (like compilers) that compose assembler programs, not for your
2226 notice. Normally both @command{@value{AS}} and @code{@value{LD}} discard
2227 such symbols, so you do not normally debug with them.
2229 This option tells @command{@value{AS}} to retain those local symbols
2230 in the object file. Usually if you do this you also tell the linker
2231 @code{@value{LD}} to preserve those symbols.
2234 @section Configuring listing output: @option{--listing}
2236 The listing feature of the assembler can be enabled via the command line switch
2237 @samp{-a} (@pxref{a}). This feature combines the input source file(s) with a
2238 hex dump of the corresponding locations in the output object file, and displays
2239 them as a listing file. The format of this listing can be controlled by
2240 directives inside the assembler source (i.e., @code{.list} (@pxref{List}),
2241 @code{.title} (@pxref{Title}), @code{.sbttl} (@pxref{Sbttl}),
2242 @code{.psize} (@pxref{Psize}), and
2243 @code{.eject} (@pxref{Eject}) and also by the following switches:
2246 @item --listing-lhs-width=@samp{number}
2247 @kindex --listing-lhs-width
2248 @cindex Width of first line disassembly output
2249 Sets the maximum width, in words, of the first line of the hex byte dump. This
2250 dump appears on the left hand side of the listing output.
2252 @item --listing-lhs-width2=@samp{number}
2253 @kindex --listing-lhs-width2
2254 @cindex Width of continuation lines of disassembly output
2255 Sets the maximum width, in words, of any further lines of the hex byte dump for
2256 a given input source line. If this value is not specified, it defaults to being
2257 the same as the value specified for @samp{--listing-lhs-width}. If neither
2258 switch is used the default is to one.
2260 @item --listing-rhs-width=@samp{number}
2261 @kindex --listing-rhs-width
2262 @cindex Width of source line output
2263 Sets the maximum width, in characters, of the source line that is displayed
2264 alongside the hex dump. The default value for this parameter is 100. The
2265 source line is displayed on the right hand side of the listing output.
2267 @item --listing-cont-lines=@samp{number}
2268 @kindex --listing-cont-lines
2269 @cindex Maximum number of continuation lines
2270 Sets the maximum number of continuation lines of hex dump that will be
2271 displayed for a given single line of source input. The default value is 4.
2275 @section Assemble in MRI Compatibility Mode: @option{-M}
2278 @cindex MRI compatibility mode
2279 The @option{-M} or @option{--mri} option selects MRI compatibility mode. This
2280 changes the syntax and pseudo-op handling of @command{@value{AS}} to make it
2281 compatible with the @code{ASM68K} or the @code{ASM960} (depending upon the
2282 configured target) assembler from Microtec Research. The exact nature of the
2283 MRI syntax will not be documented here; see the MRI manuals for more
2284 information. Note in particular that the handling of macros and macro
2285 arguments is somewhat different. The purpose of this option is to permit
2286 assembling existing MRI assembler code using @command{@value{AS}}.
2288 The MRI compatibility is not complete. Certain operations of the MRI assembler
2289 depend upon its object file format, and can not be supported using other object
2290 file formats. Supporting these would require enhancing each object file format
2291 individually. These are:
2294 @item global symbols in common section
2296 The m68k MRI assembler supports common sections which are merged by the linker.
2297 Other object file formats do not support this. @command{@value{AS}} handles
2298 common sections by treating them as a single common symbol. It permits local
2299 symbols to be defined within a common section, but it can not support global
2300 symbols, since it has no way to describe them.
2302 @item complex relocations
2304 The MRI assemblers support relocations against a negated section address, and
2305 relocations which combine the start addresses of two or more sections. These
2306 are not support by other object file formats.
2308 @item @code{END} pseudo-op specifying start address
2310 The MRI @code{END} pseudo-op permits the specification of a start address.
2311 This is not supported by other object file formats. The start address may
2312 instead be specified using the @option{-e} option to the linker, or in a linker
2315 @item @code{IDNT}, @code{.ident} and @code{NAME} pseudo-ops
2317 The MRI @code{IDNT}, @code{.ident} and @code{NAME} pseudo-ops assign a module
2318 name to the output file. This is not supported by other object file formats.
2320 @item @code{ORG} pseudo-op
2322 The m68k MRI @code{ORG} pseudo-op begins an absolute section at a given
2323 address. This differs from the usual @command{@value{AS}} @code{.org} pseudo-op,
2324 which changes the location within the current section. Absolute sections are
2325 not supported by other object file formats. The address of a section may be
2326 assigned within a linker script.
2329 There are some other features of the MRI assembler which are not supported by
2330 @command{@value{AS}}, typically either because they are difficult or because they
2331 seem of little consequence. Some of these may be supported in future releases.
2335 @item EBCDIC strings
2337 EBCDIC strings are not supported.
2339 @item packed binary coded decimal
2341 Packed binary coded decimal is not supported. This means that the @code{DC.P}
2342 and @code{DCB.P} pseudo-ops are not supported.
2344 @item @code{FEQU} pseudo-op
2346 The m68k @code{FEQU} pseudo-op is not supported.
2348 @item @code{NOOBJ} pseudo-op
2350 The m68k @code{NOOBJ} pseudo-op is not supported.
2352 @item @code{OPT} branch control options
2354 The m68k @code{OPT} branch control options---@code{B}, @code{BRS}, @code{BRB},
2355 @code{BRL}, and @code{BRW}---are ignored. @command{@value{AS}} automatically
2356 relaxes all branches, whether forward or backward, to an appropriate size, so
2357 these options serve no purpose.
2359 @item @code{OPT} list control options
2361 The following m68k @code{OPT} list control options are ignored: @code{C},
2362 @code{CEX}, @code{CL}, @code{CRE}, @code{E}, @code{G}, @code{I}, @code{M},
2363 @code{MEX}, @code{MC}, @code{MD}, @code{X}.
2365 @item other @code{OPT} options
2367 The following m68k @code{OPT} options are ignored: @code{NEST}, @code{O},
2368 @code{OLD}, @code{OP}, @code{P}, @code{PCO}, @code{PCR}, @code{PCS}, @code{R}.
2370 @item @code{OPT} @code{D} option is default
2372 The m68k @code{OPT} @code{D} option is the default, unlike the MRI assembler.
2373 @code{OPT NOD} may be used to turn it off.
2375 @item @code{XREF} pseudo-op.
2377 The m68k @code{XREF} pseudo-op is ignored.
2379 @item @code{.debug} pseudo-op
2381 The i960 @code{.debug} pseudo-op is not supported.
2383 @item @code{.extended} pseudo-op
2385 The i960 @code{.extended} pseudo-op is not supported.
2387 @item @code{.list} pseudo-op.
2389 The various options of the i960 @code{.list} pseudo-op are not supported.
2391 @item @code{.optimize} pseudo-op
2393 The i960 @code{.optimize} pseudo-op is not supported.
2395 @item @code{.output} pseudo-op
2397 The i960 @code{.output} pseudo-op is not supported.
2399 @item @code{.setreal} pseudo-op
2401 The i960 @code{.setreal} pseudo-op is not supported.
2406 @section Dependency Tracking: @option{--MD}
2409 @cindex dependency tracking
2412 @command{@value{AS}} can generate a dependency file for the file it creates. This
2413 file consists of a single rule suitable for @code{make} describing the
2414 dependencies of the main source file.
2416 The rule is written to the file named in its argument.
2418 This feature is used in the automatic updating of makefiles.
2421 @section Name the Object File: @option{-o}
2424 @cindex naming object file
2425 @cindex object file name
2426 There is always one object file output when you run @command{@value{AS}}. By
2427 default it has the name
2430 @file{a.out} (or @file{b.out}, for Intel 960 targets only).
2444 You use this option (which takes exactly one filename) to give the
2445 object file a different name.
2447 Whatever the object file is called, @command{@value{AS}} overwrites any
2448 existing file of the same name.
2451 @section Join Data and Text Sections: @option{-R}
2454 @cindex data and text sections, joining
2455 @cindex text and data sections, joining
2456 @cindex joining text and data sections
2457 @cindex merging text and data sections
2458 @option{-R} tells @command{@value{AS}} to write the object file as if all
2459 data-section data lives in the text section. This is only done at
2460 the very last moment: your binary data are the same, but data
2461 section parts are relocated differently. The data section part of
2462 your object file is zero bytes long because all its bytes are
2463 appended to the text section. (@xref{Sections,,Sections and Relocation}.)
2465 When you specify @option{-R} it would be possible to generate shorter
2466 address displacements (because we do not have to cross between text and
2467 data section). We refrain from doing this simply for compatibility with
2468 older versions of @command{@value{AS}}. In future, @option{-R} may work this way.
2471 When @command{@value{AS}} is configured for COFF or ELF output,
2472 this option is only useful if you use sections named @samp{.text} and
2477 @option{-R} is not supported for any of the HPPA targets. Using
2478 @option{-R} generates a warning from @command{@value{AS}}.
2482 @section Display Assembly Statistics: @option{--statistics}
2484 @kindex --statistics
2485 @cindex statistics, about assembly
2486 @cindex time, total for assembly
2487 @cindex space used, maximum for assembly
2488 Use @samp{--statistics} to display two statistics about the resources used by
2489 @command{@value{AS}}: the maximum amount of space allocated during the assembly
2490 (in bytes), and the total execution time taken for the assembly (in @sc{cpu}
2493 @node traditional-format
2494 @section Compatible Output: @option{--traditional-format}
2496 @kindex --traditional-format
2497 For some targets, the output of @command{@value{AS}} is different in some ways
2498 from the output of some existing assembler. This switch requests
2499 @command{@value{AS}} to use the traditional format instead.
2501 For example, it disables the exception frame optimizations which
2502 @command{@value{AS}} normally does by default on @code{@value{GCC}} output.
2505 @section Announce Version: @option{-v}
2509 @cindex assembler version
2510 @cindex version of assembler
2511 You can find out what version of as is running by including the
2512 option @samp{-v} (which you can also spell as @samp{-version}) on the
2516 @section Control Warnings: @option{-W}, @option{--warn}, @option{--no-warn}, @option{--fatal-warnings}
2518 @command{@value{AS}} should never give a warning or error message when
2519 assembling compiler output. But programs written by people often
2520 cause @command{@value{AS}} to give a warning that a particular assumption was
2521 made. All such warnings are directed to the standard error file.
2525 @cindex suppressing warnings
2526 @cindex warnings, suppressing
2527 If you use the @option{-W} and @option{--no-warn} options, no warnings are issued.
2528 This only affects the warning messages: it does not change any particular of
2529 how @command{@value{AS}} assembles your file. Errors, which stop the assembly,
2532 @kindex --fatal-warnings
2533 @cindex errors, caused by warnings
2534 @cindex warnings, causing error
2535 If you use the @option{--fatal-warnings} option, @command{@value{AS}} considers
2536 files that generate warnings to be in error.
2539 @cindex warnings, switching on
2540 You can switch these options off again by specifying @option{--warn}, which
2541 causes warnings to be output as usual.
2544 @section Generate Object File in Spite of Errors: @option{-Z}
2545 @cindex object file, after errors
2546 @cindex errors, continuing after
2547 After an error message, @command{@value{AS}} normally produces no output. If for
2548 some reason you are interested in object file output even after
2549 @command{@value{AS}} gives an error message on your program, use the @samp{-Z}
2550 option. If there are any errors, @command{@value{AS}} continues anyways, and
2551 writes an object file after a final warning message of the form @samp{@var{n}
2552 errors, @var{m} warnings, generating bad object file.}
2557 @cindex machine-independent syntax
2558 @cindex syntax, machine-independent
2559 This chapter describes the machine-independent syntax allowed in a
2560 source file. @command{@value{AS}} syntax is similar to what many other
2561 assemblers use; it is inspired by the BSD 4.2
2566 assembler, except that @command{@value{AS}} does not assemble Vax bit-fields.
2570 * Preprocessing:: Preprocessing
2571 * Whitespace:: Whitespace
2572 * Comments:: Comments
2573 * Symbol Intro:: Symbols
2574 * Statements:: Statements
2575 * Constants:: Constants
2579 @section Preprocessing
2581 @cindex preprocessing
2582 The @command{@value{AS}} internal preprocessor:
2584 @cindex whitespace, removed by preprocessor
2586 adjusts and removes extra whitespace. It leaves one space or tab before
2587 the keywords on a line, and turns any other whitespace on the line into
2590 @cindex comments, removed by preprocessor
2592 removes all comments, replacing them with a single space, or an
2593 appropriate number of newlines.
2595 @cindex constants, converted by preprocessor
2597 converts character constants into the appropriate numeric values.
2600 It does not do macro processing, include file handling, or
2601 anything else you may get from your C compiler's preprocessor. You can
2602 do include file processing with the @code{.include} directive
2603 (@pxref{Include,,@code{.include}}). You can use the @sc{gnu} C compiler driver
2604 to get other ``CPP'' style preprocessing by giving the input file a
2605 @samp{.S} suffix. @xref{Overall Options, ,Options Controlling the Kind of
2606 Output, gcc.info, Using GNU CC}.
2608 Excess whitespace, comments, and character constants
2609 cannot be used in the portions of the input text that are not
2612 @cindex turning preprocessing on and off
2613 @cindex preprocessing, turning on and off
2616 If the first line of an input file is @code{#NO_APP} or if you use the
2617 @samp{-f} option, whitespace and comments are not removed from the input file.
2618 Within an input file, you can ask for whitespace and comment removal in
2619 specific portions of the by putting a line that says @code{#APP} before the
2620 text that may contain whitespace or comments, and putting a line that says
2621 @code{#NO_APP} after this text. This feature is mainly intend to support
2622 @code{asm} statements in compilers whose output is otherwise free of comments
2629 @dfn{Whitespace} is one or more blanks or tabs, in any order.
2630 Whitespace is used to separate symbols, and to make programs neater for
2631 people to read. Unless within character constants
2632 (@pxref{Characters,,Character Constants}), any whitespace means the same
2633 as exactly one space.
2639 There are two ways of rendering comments to @command{@value{AS}}. In both
2640 cases the comment is equivalent to one space.
2642 Anything from @samp{/*} through the next @samp{*/} is a comment.
2643 This means you may not nest these comments.
2647 The only way to include a newline ('\n') in a comment
2648 is to use this sort of comment.
2651 /* This sort of comment does not nest. */
2654 @cindex line comment character
2655 Anything from a @dfn{line comment} character up to the next newline is
2656 considered a comment and is ignored. The line comment character is target
2657 specific, and some targets multiple comment characters. Some targets also have
2658 line comment characters that only work if they are the first character on a
2659 line. Some targets use a sequence of two characters to introduce a line
2660 comment. Some targets can also change their line comment characters depending
2661 upon command line options that have been used. For more details see the
2662 @emph{Syntax} section in the documentation for individual targets.
2664 If the line comment character is the hash sign (@samp{#}) then it still has the
2665 special ability to enable and disable preprocessing (@pxref{Preprocessing}) and
2666 to specify logical line numbers:
2669 @cindex lines starting with @code{#}
2670 @cindex logical line numbers
2671 To be compatible with past assemblers, lines that begin with @samp{#} have a
2672 special interpretation. Following the @samp{#} should be an absolute
2673 expression (@pxref{Expressions}): the logical line number of the @emph{next}
2674 line. Then a string (@pxref{Strings, ,Strings}) is allowed: if present it is a
2675 new logical file name. The rest of the line, if any, should be whitespace.
2677 If the first non-whitespace characters on the line are not numeric,
2678 the line is ignored. (Just like a comment.)
2681 # This is an ordinary comment.
2682 # 42-6 "new_file_name" # New logical file name
2683 # This is logical line # 36.
2685 This feature is deprecated, and may disappear from future versions
2686 of @command{@value{AS}}.
2691 @cindex characters used in symbols
2692 @ifclear SPECIAL-SYMS
2693 A @dfn{symbol} is one or more characters chosen from the set of all
2694 letters (both upper and lower case), digits and the three characters
2700 A @dfn{symbol} is one or more characters chosen from the set of all
2701 letters (both upper and lower case), digits and the three characters
2702 @samp{._$}. (Save that, on the H8/300 only, you may not use @samp{$} in
2708 On most machines, you can also use @code{$} in symbol names; exceptions
2709 are noted in @ref{Machine Dependencies}.
2711 No symbol may begin with a digit. Case is significant.
2712 There is no length limit: all characters are significant. Multibyte characters
2713 are supported. Symbols are delimited by characters not in that set, or by the
2714 beginning of a file (since the source program must end with a newline, the end
2715 of a file is not a possible symbol delimiter). @xref{Symbols}.
2716 @cindex length of symbols
2721 @cindex statements, structure of
2722 @cindex line separator character
2723 @cindex statement separator character
2725 A @dfn{statement} ends at a newline character (@samp{\n}) or a
2726 @dfn{line separator character}. The line separator character is target
2727 specific and described in the @emph{Syntax} section of each
2728 target's documentation. Not all targets support a line separator character.
2729 The newline or line separator character is considered to be part of the
2730 preceding statement. Newlines and separators within character constants are an
2731 exception: they do not end statements.
2733 @cindex newline, required at file end
2734 @cindex EOF, newline must precede
2735 It is an error to end any statement with end-of-file: the last
2736 character of any input file should be a newline.@refill
2738 An empty statement is allowed, and may include whitespace. It is ignored.
2740 @cindex instructions and directives
2741 @cindex directives and instructions
2742 @c "key symbol" is not used elsewhere in the document; seems pedantic to
2743 @c @defn{} it in that case, as was done previously... doc@cygnus.com,
2745 A statement begins with zero or more labels, optionally followed by a
2746 key symbol which determines what kind of statement it is. The key
2747 symbol determines the syntax of the rest of the statement. If the
2748 symbol begins with a dot @samp{.} then the statement is an assembler
2749 directive: typically valid for any computer. If the symbol begins with
2750 a letter the statement is an assembly language @dfn{instruction}: it
2751 assembles into a machine language instruction.
2753 Different versions of @command{@value{AS}} for different computers
2754 recognize different instructions. In fact, the same symbol may
2755 represent a different instruction in a different computer's assembly
2759 @cindex @code{:} (label)
2760 @cindex label (@code{:})
2761 A label is a symbol immediately followed by a colon (@code{:}).
2762 Whitespace before a label or after a colon is permitted, but you may not
2763 have whitespace between a label's symbol and its colon. @xref{Labels}.
2766 For HPPA targets, labels need not be immediately followed by a colon, but
2767 the definition of a label must begin in column zero. This also implies that
2768 only one label may be defined on each line.
2772 label: .directive followed by something
2773 another_label: # This is an empty statement.
2774 instruction operand_1, operand_2, @dots{}
2781 A constant is a number, written so that its value is known by
2782 inspection, without knowing any context. Like this:
2785 .byte 74, 0112, 092, 0x4A, 0X4a, 'J, '\J # All the same value.
2786 .ascii "Ring the bell\7" # A string constant.
2787 .octa 0x123456789abcdef0123456789ABCDEF0 # A bignum.
2788 .float 0f-314159265358979323846264338327\
2789 95028841971.693993751E-40 # - pi, a flonum.
2794 * Characters:: Character Constants
2795 * Numbers:: Number Constants
2799 @subsection Character Constants
2801 @cindex character constants
2802 @cindex constants, character
2803 There are two kinds of character constants. A @dfn{character} stands
2804 for one character in one byte and its value may be used in
2805 numeric expressions. String constants (properly called string
2806 @emph{literals}) are potentially many bytes and their values may not be
2807 used in arithmetic expressions.
2811 * Chars:: Characters
2815 @subsubsection Strings
2817 @cindex string constants
2818 @cindex constants, string
2819 A @dfn{string} is written between double-quotes. It may contain
2820 double-quotes or null characters. The way to get special characters
2821 into a string is to @dfn{escape} these characters: precede them with
2822 a backslash @samp{\} character. For example @samp{\\} represents
2823 one backslash: the first @code{\} is an escape which tells
2824 @command{@value{AS}} to interpret the second character literally as a backslash
2825 (which prevents @command{@value{AS}} from recognizing the second @code{\} as an
2826 escape character). The complete list of escapes follows.
2828 @cindex escape codes, character
2829 @cindex character escape codes
2832 @c Mnemonic for ACKnowledge; for ASCII this is octal code 007.
2834 @cindex @code{\b} (backspace character)
2835 @cindex backspace (@code{\b})
2837 Mnemonic for backspace; for ASCII this is octal code 010.
2840 @c Mnemonic for EOText; for ASCII this is octal code 004.
2842 @cindex @code{\f} (formfeed character)
2843 @cindex formfeed (@code{\f})
2845 Mnemonic for FormFeed; for ASCII this is octal code 014.
2847 @cindex @code{\n} (newline character)
2848 @cindex newline (@code{\n})
2850 Mnemonic for newline; for ASCII this is octal code 012.
2853 @c Mnemonic for prefix; for ASCII this is octal code 033, usually known as @code{escape}.
2855 @cindex @code{\r} (carriage return character)
2856 @cindex carriage return (@code{\r})
2858 Mnemonic for carriage-Return; for ASCII this is octal code 015.
2861 @c Mnemonic for space; for ASCII this is octal code 040. Included for compliance with
2862 @c other assemblers.
2864 @cindex @code{\t} (tab)
2865 @cindex tab (@code{\t})
2867 Mnemonic for horizontal Tab; for ASCII this is octal code 011.
2870 @c Mnemonic for Vertical tab; for ASCII this is octal code 013.
2871 @c @item \x @var{digit} @var{digit} @var{digit}
2872 @c A hexadecimal character code. The numeric code is 3 hexadecimal digits.
2874 @cindex @code{\@var{ddd}} (octal character code)
2875 @cindex octal character code (@code{\@var{ddd}})
2876 @item \ @var{digit} @var{digit} @var{digit}
2877 An octal character code. The numeric code is 3 octal digits.
2878 For compatibility with other Unix systems, 8 and 9 are accepted as digits:
2879 for example, @code{\008} has the value 010, and @code{\009} the value 011.
2881 @cindex @code{\@var{xd...}} (hex character code)
2882 @cindex hex character code (@code{\@var{xd...}})
2883 @item \@code{x} @var{hex-digits...}
2884 A hex character code. All trailing hex digits are combined. Either upper or
2885 lower case @code{x} works.
2887 @cindex @code{\\} (@samp{\} character)
2888 @cindex backslash (@code{\\})
2890 Represents one @samp{\} character.
2893 @c Represents one @samp{'} (accent acute) character.
2894 @c This is needed in single character literals
2895 @c (@xref{Characters,,Character Constants}.) to represent
2898 @cindex @code{\"} (doublequote character)
2899 @cindex doublequote (@code{\"})
2901 Represents one @samp{"} character. Needed in strings to represent
2902 this character, because an unescaped @samp{"} would end the string.
2904 @item \ @var{anything-else}
2905 Any other character when escaped by @kbd{\} gives a warning, but
2906 assembles as if the @samp{\} was not present. The idea is that if
2907 you used an escape sequence you clearly didn't want the literal
2908 interpretation of the following character. However @command{@value{AS}} has no
2909 other interpretation, so @command{@value{AS}} knows it is giving you the wrong
2910 code and warns you of the fact.
2913 Which characters are escapable, and what those escapes represent,
2914 varies widely among assemblers. The current set is what we think
2915 the BSD 4.2 assembler recognizes, and is a subset of what most C
2916 compilers recognize. If you are in doubt, do not use an escape
2920 @subsubsection Characters
2922 @cindex single character constant
2923 @cindex character, single
2924 @cindex constant, single character
2925 A single character may be written as a single quote immediately
2926 followed by that character. The same escapes apply to characters as
2927 to strings. So if you want to write the character backslash, you
2928 must write @kbd{'\\} where the first @code{\} escapes the second
2929 @code{\}. As you can see, the quote is an acute accent, not a
2930 grave accent. A newline
2932 @ifclear abnormal-separator
2933 (or semicolon @samp{;})
2935 @ifset abnormal-separator
2937 (or dollar sign @samp{$}, for the H8/300; or semicolon @samp{;} for the
2942 immediately following an acute accent is taken as a literal character
2943 and does not count as the end of a statement. The value of a character
2944 constant in a numeric expression is the machine's byte-wide code for
2945 that character. @command{@value{AS}} assumes your character code is ASCII:
2946 @kbd{'A} means 65, @kbd{'B} means 66, and so on. @refill
2949 @subsection Number Constants
2951 @cindex constants, number
2952 @cindex number constants
2953 @command{@value{AS}} distinguishes three kinds of numbers according to how they
2954 are stored in the target machine. @emph{Integers} are numbers that
2955 would fit into an @code{int} in the C language. @emph{Bignums} are
2956 integers, but they are stored in more than 32 bits. @emph{Flonums}
2957 are floating point numbers, described below.
2960 * Integers:: Integers
2965 * Bit Fields:: Bit Fields
2971 @subsubsection Integers
2973 @cindex constants, integer
2975 @cindex binary integers
2976 @cindex integers, binary
2977 A binary integer is @samp{0b} or @samp{0B} followed by zero or more of
2978 the binary digits @samp{01}.
2980 @cindex octal integers
2981 @cindex integers, octal
2982 An octal integer is @samp{0} followed by zero or more of the octal
2983 digits (@samp{01234567}).
2985 @cindex decimal integers
2986 @cindex integers, decimal
2987 A decimal integer starts with a non-zero digit followed by zero or
2988 more digits (@samp{0123456789}).
2990 @cindex hexadecimal integers
2991 @cindex integers, hexadecimal
2992 A hexadecimal integer is @samp{0x} or @samp{0X} followed by one or
2993 more hexadecimal digits chosen from @samp{0123456789abcdefABCDEF}.
2995 Integers have the usual values. To denote a negative integer, use
2996 the prefix operator @samp{-} discussed under expressions
2997 (@pxref{Prefix Ops,,Prefix Operators}).
3000 @subsubsection Bignums
3003 @cindex constants, bignum
3004 A @dfn{bignum} has the same syntax and semantics as an integer
3005 except that the number (or its negative) takes more than 32 bits to
3006 represent in binary. The distinction is made because in some places
3007 integers are permitted while bignums are not.
3010 @subsubsection Flonums
3012 @cindex floating point numbers
3013 @cindex constants, floating point
3015 @cindex precision, floating point
3016 A @dfn{flonum} represents a floating point number. The translation is
3017 indirect: a decimal floating point number from the text is converted by
3018 @command{@value{AS}} to a generic binary floating point number of more than
3019 sufficient precision. This generic floating point number is converted
3020 to a particular computer's floating point format (or formats) by a
3021 portion of @command{@value{AS}} specialized to that computer.
3023 A flonum is written by writing (in order)
3028 (@samp{0} is optional on the HPPA.)
3032 A letter, to tell @command{@value{AS}} the rest of the number is a flonum.
3034 @kbd{e} is recommended. Case is not important.
3036 @c FIXME: verify if flonum syntax really this vague for most cases
3037 (Any otherwise illegal letter works here, but that might be changed. Vax BSD
3038 4.2 assembler seems to allow any of @samp{defghDEFGH}.)
3041 On the H8/300, Renesas / SuperH SH,
3042 and AMD 29K architectures, the letter must be
3043 one of the letters @samp{DFPRSX} (in upper or lower case).
3045 On the ARC, the letter must be one of the letters @samp{DFRS}
3046 (in upper or lower case).
3048 On the Intel 960 architecture, the letter must be
3049 one of the letters @samp{DFT} (in upper or lower case).
3051 On the HPPA architecture, the letter must be @samp{E} (upper case only).
3055 One of the letters @samp{DFRS} (in upper or lower case).
3058 One of the letters @samp{DFPRSX} (in upper or lower case).
3061 The letter @samp{E} (upper case only).
3064 One of the letters @samp{DFT} (in upper or lower case).
3069 An optional sign: either @samp{+} or @samp{-}.
3072 An optional @dfn{integer part}: zero or more decimal digits.
3075 An optional @dfn{fractional part}: @samp{.} followed by zero
3076 or more decimal digits.
3079 An optional exponent, consisting of:
3083 An @samp{E} or @samp{e}.
3084 @c I can't find a config where "EXP_CHARS" is other than 'eE', but in
3085 @c principle this can perfectly well be different on different targets.
3087 Optional sign: either @samp{+} or @samp{-}.
3089 One or more decimal digits.
3094 At least one of the integer part or the fractional part must be
3095 present. The floating point number has the usual base-10 value.
3097 @command{@value{AS}} does all processing using integers. Flonums are computed
3098 independently of any floating point hardware in the computer running
3099 @command{@value{AS}}.
3103 @c Bit fields are written as a general facility but are also controlled
3104 @c by a conditional-compilation flag---which is as of now (21mar91)
3105 @c turned on only by the i960 config of GAS.
3107 @subsubsection Bit Fields
3110 @cindex constants, bit field
3111 You can also define numeric constants as @dfn{bit fields}.
3112 Specify two numbers separated by a colon---
3114 @var{mask}:@var{value}
3117 @command{@value{AS}} applies a bitwise @sc{and} between @var{mask} and
3120 The resulting number is then packed
3122 @c this conditional paren in case bit fields turned on elsewhere than 960
3123 (in host-dependent byte order)
3125 into a field whose width depends on which assembler directive has the
3126 bit-field as its argument. Overflow (a result from the bitwise and
3127 requiring more binary digits to represent) is not an error; instead,
3128 more constants are generated, of the specified width, beginning with the
3129 least significant digits.@refill
3131 The directives @code{.byte}, @code{.hword}, @code{.int}, @code{.long},
3132 @code{.short}, and @code{.word} accept bit-field arguments.
3137 @chapter Sections and Relocation
3142 * Secs Background:: Background
3143 * Ld Sections:: Linker Sections
3144 * As Sections:: Assembler Internal Sections
3145 * Sub-Sections:: Sub-Sections
3149 @node Secs Background
3152 Roughly, a section is a range of addresses, with no gaps; all data
3153 ``in'' those addresses is treated the same for some particular purpose.
3154 For example there may be a ``read only'' section.
3156 @cindex linker, and assembler
3157 @cindex assembler, and linker
3158 The linker @code{@value{LD}} reads many object files (partial programs) and
3159 combines their contents to form a runnable program. When @command{@value{AS}}
3160 emits an object file, the partial program is assumed to start at address 0.
3161 @code{@value{LD}} assigns the final addresses for the partial program, so that
3162 different partial programs do not overlap. This is actually an
3163 oversimplification, but it suffices to explain how @command{@value{AS}} uses
3166 @code{@value{LD}} moves blocks of bytes of your program to their run-time
3167 addresses. These blocks slide to their run-time addresses as rigid
3168 units; their length does not change and neither does the order of bytes
3169 within them. Such a rigid unit is called a @emph{section}. Assigning
3170 run-time addresses to sections is called @dfn{relocation}. It includes
3171 the task of adjusting mentions of object-file addresses so they refer to
3172 the proper run-time addresses.
3174 For the H8/300, and for the Renesas / SuperH SH,
3175 @command{@value{AS}} pads sections if needed to
3176 ensure they end on a word (sixteen bit) boundary.
3179 @cindex standard assembler sections
3180 An object file written by @command{@value{AS}} has at least three sections, any
3181 of which may be empty. These are named @dfn{text}, @dfn{data} and
3186 When it generates COFF or ELF output,
3188 @command{@value{AS}} can also generate whatever other named sections you specify
3189 using the @samp{.section} directive (@pxref{Section,,@code{.section}}).
3190 If you do not use any directives that place output in the @samp{.text}
3191 or @samp{.data} sections, these sections still exist, but are empty.
3196 When @command{@value{AS}} generates SOM or ELF output for the HPPA,
3198 @command{@value{AS}} can also generate whatever other named sections you
3199 specify using the @samp{.space} and @samp{.subspace} directives. See
3200 @cite{HP9000 Series 800 Assembly Language Reference Manual}
3201 (HP 92432-90001) for details on the @samp{.space} and @samp{.subspace}
3202 assembler directives.
3205 Additionally, @command{@value{AS}} uses different names for the standard
3206 text, data, and bss sections when generating SOM output. Program text
3207 is placed into the @samp{$CODE$} section, data into @samp{$DATA$}, and
3208 BSS into @samp{$BSS$}.
3212 Within the object file, the text section starts at address @code{0}, the
3213 data section follows, and the bss section follows the data section.
3216 When generating either SOM or ELF output files on the HPPA, the text
3217 section starts at address @code{0}, the data section at address
3218 @code{0x4000000}, and the bss section follows the data section.
3221 To let @code{@value{LD}} know which data changes when the sections are
3222 relocated, and how to change that data, @command{@value{AS}} also writes to the
3223 object file details of the relocation needed. To perform relocation
3224 @code{@value{LD}} must know, each time an address in the object
3228 Where in the object file is the beginning of this reference to
3231 How long (in bytes) is this reference?
3233 Which section does the address refer to? What is the numeric value of
3235 (@var{address}) @minus{} (@var{start-address of section})?
3238 Is the reference to an address ``Program-Counter relative''?
3241 @cindex addresses, format of
3242 @cindex section-relative addressing
3243 In fact, every address @command{@value{AS}} ever uses is expressed as
3245 (@var{section}) + (@var{offset into section})
3248 Further, most expressions @command{@value{AS}} computes have this section-relative
3251 (For some object formats, such as SOM for the HPPA, some expressions are
3252 symbol-relative instead.)
3255 In this manual we use the notation @{@var{secname} @var{N}@} to mean ``offset
3256 @var{N} into section @var{secname}.''
3258 Apart from text, data and bss sections you need to know about the
3259 @dfn{absolute} section. When @code{@value{LD}} mixes partial programs,
3260 addresses in the absolute section remain unchanged. For example, address
3261 @code{@{absolute 0@}} is ``relocated'' to run-time address 0 by
3262 @code{@value{LD}}. Although the linker never arranges two partial programs'
3263 data sections with overlapping addresses after linking, @emph{by definition}
3264 their absolute sections must overlap. Address @code{@{absolute@ 239@}} in one
3265 part of a program is always the same address when the program is running as
3266 address @code{@{absolute@ 239@}} in any other part of the program.
3268 The idea of sections is extended to the @dfn{undefined} section. Any
3269 address whose section is unknown at assembly time is by definition
3270 rendered @{undefined @var{U}@}---where @var{U} is filled in later.
3271 Since numbers are always defined, the only way to generate an undefined
3272 address is to mention an undefined symbol. A reference to a named
3273 common block would be such a symbol: its value is unknown at assembly
3274 time so it has section @emph{undefined}.
3276 By analogy the word @emph{section} is used to describe groups of sections in
3277 the linked program. @code{@value{LD}} puts all partial programs' text
3278 sections in contiguous addresses in the linked program. It is
3279 customary to refer to the @emph{text section} of a program, meaning all
3280 the addresses of all partial programs' text sections. Likewise for
3281 data and bss sections.
3283 Some sections are manipulated by @code{@value{LD}}; others are invented for
3284 use of @command{@value{AS}} and have no meaning except during assembly.
3287 @section Linker Sections
3288 @code{@value{LD}} deals with just four kinds of sections, summarized below.
3293 @cindex named sections
3294 @cindex sections, named
3295 @item named sections
3298 @cindex text section
3299 @cindex data section
3303 These sections hold your program. @command{@value{AS}} and @code{@value{LD}} treat them as
3304 separate but equal sections. Anything you can say of one section is
3307 When the program is running, however, it is
3308 customary for the text section to be unalterable. The
3309 text section is often shared among processes: it contains
3310 instructions, constants and the like. The data section of a running
3311 program is usually alterable: for example, C variables would be stored
3312 in the data section.
3317 This section contains zeroed bytes when your program begins running. It
3318 is used to hold uninitialized variables or common storage. The length of
3319 each partial program's bss section is important, but because it starts
3320 out containing zeroed bytes there is no need to store explicit zero
3321 bytes in the object file. The bss section was invented to eliminate
3322 those explicit zeros from object files.
3324 @cindex absolute section
3325 @item absolute section
3326 Address 0 of this section is always ``relocated'' to runtime address 0.
3327 This is useful if you want to refer to an address that @code{@value{LD}} must
3328 not change when relocating. In this sense we speak of absolute
3329 addresses being ``unrelocatable'': they do not change during relocation.
3331 @cindex undefined section
3332 @item undefined section
3333 This ``section'' is a catch-all for address references to objects not in
3334 the preceding sections.
3335 @c FIXME: ref to some other doc on obj-file formats could go here.
3338 @cindex relocation example
3339 An idealized example of three relocatable sections follows.
3341 The example uses the traditional section names @samp{.text} and @samp{.data}.
3343 Memory addresses are on the horizontal axis.
3347 @c END TEXI2ROFF-KILL
3350 partial program # 1: |ttttt|dddd|00|
3357 partial program # 2: |TTT|DDD|000|
3360 +--+---+-----+--+----+---+-----+~~
3361 linked program: | |TTT|ttttt| |dddd|DDD|00000|
3362 +--+---+-----+--+----+---+-----+~~
3364 addresses: 0 @dots{}
3371 \line{\it Partial program \#1: \hfil}
3372 \line{\ibox{2.5cm}{\tt text}\ibox{2cm}{\tt data}\ibox{1cm}{\tt bss}\hfil}
3373 \line{\boxit{2.5cm}{\tt ttttt}\boxit{2cm}{\tt dddd}\boxit{1cm}{\tt 00}\hfil}
3375 \line{\it Partial program \#2: \hfil}
3376 \line{\ibox{1cm}{\tt text}\ibox{1.5cm}{\tt data}\ibox{1cm}{\tt bss}\hfil}
3377 \line{\boxit{1cm}{\tt TTT}\boxit{1.5cm}{\tt DDDD}\boxit{1cm}{\tt 000}\hfil}
3379 \line{\it linked program: \hfil}
3380 \line{\ibox{.5cm}{}\ibox{1cm}{\tt text}\ibox{2.5cm}{}\ibox{.75cm}{}\ibox{2cm}{\tt data}\ibox{1.5cm}{}\ibox{2cm}{\tt bss}\hfil}
3381 \line{\boxit{.5cm}{}\boxit{1cm}{\tt TTT}\boxit{2.5cm}{\tt
3382 ttttt}\boxit{.75cm}{}\boxit{2cm}{\tt dddd}\boxit{1.5cm}{\tt
3383 DDDD}\boxit{2cm}{\tt 00000}\ \dots\hfil}
3385 \line{\it addresses: \hfil}
3389 @c END TEXI2ROFF-KILL
3392 @section Assembler Internal Sections
3394 @cindex internal assembler sections
3395 @cindex sections in messages, internal
3396 These sections are meant only for the internal use of @command{@value{AS}}. They
3397 have no meaning at run-time. You do not really need to know about these
3398 sections for most purposes; but they can be mentioned in @command{@value{AS}}
3399 warning messages, so it might be helpful to have an idea of their
3400 meanings to @command{@value{AS}}. These sections are used to permit the
3401 value of every expression in your assembly language program to be a
3402 section-relative address.
3405 @cindex assembler internal logic error
3406 @item ASSEMBLER-INTERNAL-LOGIC-ERROR!
3407 An internal assembler logic error has been found. This means there is a
3408 bug in the assembler.
3410 @cindex expr (internal section)
3412 The assembler stores complex expression internally as combinations of
3413 symbols. When it needs to represent an expression as a symbol, it puts
3414 it in the expr section.
3416 @c FIXME item transfer[t] vector preload
3417 @c FIXME item transfer[t] vector postload
3418 @c FIXME item register
3422 @section Sub-Sections
3424 @cindex numbered subsections
3425 @cindex grouping data
3431 fall into two sections: text and data.
3433 You may have separate groups of
3435 data in named sections
3439 data in named sections
3445 that you want to end up near to each other in the object file, even though they
3446 are not contiguous in the assembler source. @command{@value{AS}} allows you to
3447 use @dfn{subsections} for this purpose. Within each section, there can be
3448 numbered subsections with values from 0 to 8192. Objects assembled into the
3449 same subsection go into the object file together with other objects in the same
3450 subsection. For example, a compiler might want to store constants in the text
3451 section, but might not want to have them interspersed with the program being
3452 assembled. In this case, the compiler could issue a @samp{.text 0} before each
3453 section of code being output, and a @samp{.text 1} before each group of
3454 constants being output.
3456 Subsections are optional. If you do not use subsections, everything
3457 goes in subsection number zero.
3460 Each subsection is zero-padded up to a multiple of four bytes.
3461 (Subsections may be padded a different amount on different flavors
3462 of @command{@value{AS}}.)
3466 On the H8/300 platform, each subsection is zero-padded to a word
3467 boundary (two bytes).
3468 The same is true on the Renesas SH.
3471 @c FIXME section padding (alignment)?
3472 @c Rich Pixley says padding here depends on target obj code format; that
3473 @c doesn't seem particularly useful to say without further elaboration,
3474 @c so for now I say nothing about it. If this is a generic BFD issue,
3475 @c these paragraphs might need to vanish from this manual, and be
3476 @c discussed in BFD chapter of binutils (or some such).
3480 Subsections appear in your object file in numeric order, lowest numbered
3481 to highest. (All this to be compatible with other people's assemblers.)
3482 The object file contains no representation of subsections; @code{@value{LD}} and
3483 other programs that manipulate object files see no trace of them.
3484 They just see all your text subsections as a text section, and all your
3485 data subsections as a data section.
3487 To specify which subsection you want subsequent statements assembled
3488 into, use a numeric argument to specify it, in a @samp{.text
3489 @var{expression}} or a @samp{.data @var{expression}} statement.
3492 When generating COFF output, you
3497 can also use an extra subsection
3498 argument with arbitrary named sections: @samp{.section @var{name},
3503 When generating ELF output, you
3508 can also use the @code{.subsection} directive (@pxref{SubSection})
3509 to specify a subsection: @samp{.subsection @var{expression}}.
3511 @var{Expression} should be an absolute expression
3512 (@pxref{Expressions}). If you just say @samp{.text} then @samp{.text 0}
3513 is assumed. Likewise @samp{.data} means @samp{.data 0}. Assembly
3514 begins in @code{text 0}. For instance:
3516 .text 0 # The default subsection is text 0 anyway.
3517 .ascii "This lives in the first text subsection. *"
3519 .ascii "But this lives in the second text subsection."
3521 .ascii "This lives in the data section,"
3522 .ascii "in the first data subsection."
3524 .ascii "This lives in the first text section,"
3525 .ascii "immediately following the asterisk (*)."
3528 Each section has a @dfn{location counter} incremented by one for every byte
3529 assembled into that section. Because subsections are merely a convenience
3530 restricted to @command{@value{AS}} there is no concept of a subsection location
3531 counter. There is no way to directly manipulate a location counter---but the
3532 @code{.align} directive changes it, and any label definition captures its
3533 current value. The location counter of the section where statements are being
3534 assembled is said to be the @dfn{active} location counter.
3537 @section bss Section
3540 @cindex common variable storage
3541 The bss section is used for local common variable storage.
3542 You may allocate address space in the bss section, but you may
3543 not dictate data to load into it before your program executes. When
3544 your program starts running, all the contents of the bss
3545 section are zeroed bytes.
3547 The @code{.lcomm} pseudo-op defines a symbol in the bss section; see
3548 @ref{Lcomm,,@code{.lcomm}}.
3550 The @code{.comm} pseudo-op may be used to declare a common symbol, which is
3551 another form of uninitialized symbol; see @ref{Comm,,@code{.comm}}.
3554 When assembling for a target which supports multiple sections, such as ELF or
3555 COFF, you may switch into the @code{.bss} section and define symbols as usual;
3556 see @ref{Section,,@code{.section}}. You may only assemble zero values into the
3557 section. Typically the section will only contain symbol definitions and
3558 @code{.skip} directives (@pxref{Skip,,@code{.skip}}).
3565 Symbols are a central concept: the programmer uses symbols to name
3566 things, the linker uses symbols to link, and the debugger uses symbols
3570 @cindex debuggers, and symbol order
3571 @emph{Warning:} @command{@value{AS}} does not place symbols in the object file in
3572 the same order they were declared. This may break some debuggers.
3577 * Setting Symbols:: Giving Symbols Other Values
3578 * Symbol Names:: Symbol Names
3579 * Dot:: The Special Dot Symbol
3580 * Symbol Attributes:: Symbol Attributes
3587 A @dfn{label} is written as a symbol immediately followed by a colon
3588 @samp{:}. The symbol then represents the current value of the
3589 active location counter, and is, for example, a suitable instruction
3590 operand. You are warned if you use the same symbol to represent two
3591 different locations: the first definition overrides any other
3595 On the HPPA, the usual form for a label need not be immediately followed by a
3596 colon, but instead must start in column zero. Only one label may be defined on
3597 a single line. To work around this, the HPPA version of @command{@value{AS}} also
3598 provides a special directive @code{.label} for defining labels more flexibly.
3601 @node Setting Symbols
3602 @section Giving Symbols Other Values
3604 @cindex assigning values to symbols
3605 @cindex symbol values, assigning
3606 A symbol can be given an arbitrary value by writing a symbol, followed
3607 by an equals sign @samp{=}, followed by an expression
3608 (@pxref{Expressions}). This is equivalent to using the @code{.set}
3609 directive. @xref{Set,,@code{.set}}. In the same way, using a double
3610 equals sign @samp{=}@samp{=} here represents an equivalent of the
3611 @code{.eqv} directive. @xref{Eqv,,@code{.eqv}}.
3614 Blackfin does not support symbol assignment with @samp{=}.
3618 @section Symbol Names
3620 @cindex symbol names
3621 @cindex names, symbol
3622 @ifclear SPECIAL-SYMS
3623 Symbol names begin with a letter or with one of @samp{._}. On most
3624 machines, you can also use @code{$} in symbol names; exceptions are
3625 noted in @ref{Machine Dependencies}. That character may be followed by any
3626 string of digits, letters, dollar signs (unless otherwise noted for a
3627 particular target machine), and underscores.
3631 Symbol names begin with a letter or with one of @samp{._}. On the
3632 Renesas SH you can also use @code{$} in symbol names. That
3633 character may be followed by any string of digits, letters, dollar signs (save
3634 on the H8/300), and underscores.
3638 Case of letters is significant: @code{foo} is a different symbol name
3641 Multibyte characters are supported. To generate a symbol name containing
3642 multibyte characters enclose it within double quotes and use escape codes. cf
3643 @xref{Strings}. Generating a multibyte symbol name from a label is not
3644 currently supported.
3646 Each symbol has exactly one name. Each name in an assembly language program
3647 refers to exactly one symbol. You may use that symbol name any number of times
3650 @subheading Local Symbol Names
3652 @cindex local symbol names
3653 @cindex symbol names, local
3654 A local symbol is any symbol beginning with certain local label prefixes.
3655 By default, the local label prefix is @samp{.L} for ELF systems or
3656 @samp{L} for traditional a.out systems, but each target may have its own
3657 set of local label prefixes.
3659 On the HPPA local symbols begin with @samp{L$}.
3662 Local symbols are defined and used within the assembler, but they are
3663 normally not saved in object files. Thus, they are not visible when debugging.
3664 You may use the @samp{-L} option (@pxref{L, ,Include Local Symbols})
3665 to retain the local symbols in the object files.
3667 @subheading Local Labels
3669 @cindex local labels
3670 @cindex temporary symbol names
3671 @cindex symbol names, temporary
3672 Local labels help compilers and programmers use names temporarily.
3673 They create symbols which are guaranteed to be unique over the entire scope of
3674 the input source code and which can be referred to by a simple notation.
3675 To define a local label, write a label of the form @samp{@b{N}:} (where @b{N}
3676 represents any positive integer). To refer to the most recent previous
3677 definition of that label write @samp{@b{N}b}, using the same number as when
3678 you defined the label. To refer to the next definition of a local label, write
3679 @samp{@b{N}f}---the @samp{b} stands for ``backwards'' and the @samp{f} stands
3682 There is no restriction on how you can use these labels, and you can reuse them
3683 too. So that it is possible to repeatedly define the same local label (using
3684 the same number @samp{@b{N}}), although you can only refer to the most recently
3685 defined local label of that number (for a backwards reference) or the next
3686 definition of a specific local label for a forward reference. It is also worth
3687 noting that the first 10 local labels (@samp{@b{0:}}@dots{}@samp{@b{9:}}) are
3688 implemented in a slightly more efficient manner than the others.
3699 Which is the equivalent of:
3702 label_1: branch label_3
3703 label_2: branch label_1
3704 label_3: branch label_4
3705 label_4: branch label_3
3708 Local label names are only a notational device. They are immediately
3709 transformed into more conventional symbol names before the assembler uses them.
3710 The symbol names are stored in the symbol table, appear in error messages, and
3711 are optionally emitted to the object file. The names are constructed using
3715 @item @emph{local label prefix}
3716 All local symbols begin with the system-specific local label prefix.
3717 Normally both @command{@value{AS}} and @code{@value{LD}} forget symbols
3718 that start with the local label prefix. These labels are
3719 used for symbols you are never intended to see. If you use the
3720 @samp{-L} option then @command{@value{AS}} retains these symbols in the
3721 object file. If you also instruct @code{@value{LD}} to retain these symbols,
3722 you may use them in debugging.
3725 This is the number that was used in the local label definition. So if the
3726 label is written @samp{55:} then the number is @samp{55}.
3729 This unusual character is included so you do not accidentally invent a symbol
3730 of the same name. The character has ASCII value of @samp{\002} (control-B).
3732 @item @emph{ordinal number}
3733 This is a serial number to keep the labels distinct. The first definition of
3734 @samp{0:} gets the number @samp{1}. The 15th definition of @samp{0:} gets the
3735 number @samp{15}, and so on. Likewise the first definition of @samp{1:} gets
3736 the number @samp{1} and its 15th definition gets @samp{15} as well.
3739 So for example, the first @code{1:} may be named @code{.L1@kbd{C-B}1}, and
3740 the 44th @code{3:} may be named @code{.L3@kbd{C-B}44}.
3742 @subheading Dollar Local Labels
3743 @cindex dollar local symbols
3745 @code{@value{AS}} also supports an even more local form of local labels called
3746 dollar labels. These labels go out of scope (i.e., they become undefined) as
3747 soon as a non-local label is defined. Thus they remain valid for only a small
3748 region of the input source code. Normal local labels, by contrast, remain in
3749 scope for the entire file, or until they are redefined by another occurrence of
3750 the same local label.
3752 Dollar labels are defined in exactly the same way as ordinary local labels,
3753 except that they have a dollar sign suffix to their numeric value, e.g.,
3756 They can also be distinguished from ordinary local labels by their transformed
3757 names which use ASCII character @samp{\001} (control-A) as the magic character
3758 to distinguish them from ordinary labels. For example, the fifth definition of
3759 @samp{6$} may be named @samp{.L6@kbd{C-A}5}.
3762 @section The Special Dot Symbol
3764 @cindex dot (symbol)
3765 @cindex @code{.} (symbol)
3766 @cindex current address
3767 @cindex location counter
3768 The special symbol @samp{.} refers to the current address that
3769 @command{@value{AS}} is assembling into. Thus, the expression @samp{melvin:
3770 .long .} defines @code{melvin} to contain its own address.
3771 Assigning a value to @code{.} is treated the same as a @code{.org}
3773 @ifclear no-space-dir
3774 Thus, the expression @samp{.=.+4} is the same as saying
3778 @node Symbol Attributes
3779 @section Symbol Attributes
3781 @cindex symbol attributes
3782 @cindex attributes, symbol
3783 Every symbol has, as well as its name, the attributes ``Value'' and
3784 ``Type''. Depending on output format, symbols can also have auxiliary
3787 The detailed definitions are in @file{a.out.h}.
3790 If you use a symbol without defining it, @command{@value{AS}} assumes zero for
3791 all these attributes, and probably won't warn you. This makes the
3792 symbol an externally defined symbol, which is generally what you
3796 * Symbol Value:: Value
3797 * Symbol Type:: Type
3800 * a.out Symbols:: Symbol Attributes: @code{a.out}
3804 * a.out Symbols:: Symbol Attributes: @code{a.out}
3807 * a.out Symbols:: Symbol Attributes: @code{a.out}, @code{b.out}
3812 * COFF Symbols:: Symbol Attributes for COFF
3815 * SOM Symbols:: Symbol Attributes for SOM
3822 @cindex value of a symbol
3823 @cindex symbol value
3824 The value of a symbol is (usually) 32 bits. For a symbol which labels a
3825 location in the text, data, bss or absolute sections the value is the
3826 number of addresses from the start of that section to the label.
3827 Naturally for text, data and bss sections the value of a symbol changes
3828 as @code{@value{LD}} changes section base addresses during linking. Absolute
3829 symbols' values do not change during linking: that is why they are
3832 The value of an undefined symbol is treated in a special way. If it is
3833 0 then the symbol is not defined in this assembler source file, and
3834 @code{@value{LD}} tries to determine its value from other files linked into the
3835 same program. You make this kind of symbol simply by mentioning a symbol
3836 name without defining it. A non-zero value represents a @code{.comm}
3837 common declaration. The value is how much common storage to reserve, in
3838 bytes (addresses). The symbol refers to the first address of the
3844 @cindex type of a symbol
3846 The type attribute of a symbol contains relocation (section)
3847 information, any flag settings indicating that a symbol is external, and
3848 (optionally), other information for linkers and debuggers. The exact
3849 format depends on the object-code output format in use.
3854 @c The following avoids a "widow" subsection title. @group would be
3855 @c better if it were available outside examples.
3858 @subsection Symbol Attributes: @code{a.out}, @code{b.out}
3860 @cindex @code{b.out} symbol attributes
3861 @cindex symbol attributes, @code{b.out}
3862 These symbol attributes appear only when @command{@value{AS}} is configured for
3863 one of the Berkeley-descended object output formats---@code{a.out} or
3869 @subsection Symbol Attributes: @code{a.out}
3871 @cindex @code{a.out} symbol attributes
3872 @cindex symbol attributes, @code{a.out}
3878 @subsection Symbol Attributes: @code{a.out}
3880 @cindex @code{a.out} symbol attributes
3881 @cindex symbol attributes, @code{a.out}
3885 * Symbol Desc:: Descriptor
3886 * Symbol Other:: Other
3890 @subsubsection Descriptor
3892 @cindex descriptor, of @code{a.out} symbol
3893 This is an arbitrary 16-bit value. You may establish a symbol's
3894 descriptor value by using a @code{.desc} statement
3895 (@pxref{Desc,,@code{.desc}}). A descriptor value means nothing to
3896 @command{@value{AS}}.
3899 @subsubsection Other
3901 @cindex other attribute, of @code{a.out} symbol
3902 This is an arbitrary 8-bit value. It means nothing to @command{@value{AS}}.
3907 @subsection Symbol Attributes for COFF
3909 @cindex COFF symbol attributes
3910 @cindex symbol attributes, COFF
3912 The COFF format supports a multitude of auxiliary symbol attributes;
3913 like the primary symbol attributes, they are set between @code{.def} and
3914 @code{.endef} directives.
3916 @subsubsection Primary Attributes
3918 @cindex primary attributes, COFF symbols
3919 The symbol name is set with @code{.def}; the value and type,
3920 respectively, with @code{.val} and @code{.type}.
3922 @subsubsection Auxiliary Attributes
3924 @cindex auxiliary attributes, COFF symbols
3925 The @command{@value{AS}} directives @code{.dim}, @code{.line}, @code{.scl},
3926 @code{.size}, @code{.tag}, and @code{.weak} can generate auxiliary symbol
3927 table information for COFF.
3932 @subsection Symbol Attributes for SOM
3934 @cindex SOM symbol attributes
3935 @cindex symbol attributes, SOM
3937 The SOM format for the HPPA supports a multitude of symbol attributes set with
3938 the @code{.EXPORT} and @code{.IMPORT} directives.
3940 The attributes are described in @cite{HP9000 Series 800 Assembly
3941 Language Reference Manual} (HP 92432-90001) under the @code{IMPORT} and
3942 @code{EXPORT} assembler directive documentation.
3946 @chapter Expressions
3950 @cindex numeric values
3951 An @dfn{expression} specifies an address or numeric value.
3952 Whitespace may precede and/or follow an expression.
3954 The result of an expression must be an absolute number, or else an offset into
3955 a particular section. If an expression is not absolute, and there is not
3956 enough information when @command{@value{AS}} sees the expression to know its
3957 section, a second pass over the source program might be necessary to interpret
3958 the expression---but the second pass is currently not implemented.
3959 @command{@value{AS}} aborts with an error message in this situation.
3962 * Empty Exprs:: Empty Expressions
3963 * Integer Exprs:: Integer Expressions
3967 @section Empty Expressions
3969 @cindex empty expressions
3970 @cindex expressions, empty
3971 An empty expression has no value: it is just whitespace or null.
3972 Wherever an absolute expression is required, you may omit the
3973 expression, and @command{@value{AS}} assumes a value of (absolute) 0. This
3974 is compatible with other assemblers.
3977 @section Integer Expressions
3979 @cindex integer expressions
3980 @cindex expressions, integer
3981 An @dfn{integer expression} is one or more @emph{arguments} delimited
3982 by @emph{operators}.
3985 * Arguments:: Arguments
3986 * Operators:: Operators
3987 * Prefix Ops:: Prefix Operators
3988 * Infix Ops:: Infix Operators
3992 @subsection Arguments
3994 @cindex expression arguments
3995 @cindex arguments in expressions
3996 @cindex operands in expressions
3997 @cindex arithmetic operands
3998 @dfn{Arguments} are symbols, numbers or subexpressions. In other
3999 contexts arguments are sometimes called ``arithmetic operands''. In
4000 this manual, to avoid confusing them with the ``instruction operands'' of
4001 the machine language, we use the term ``argument'' to refer to parts of
4002 expressions only, reserving the word ``operand'' to refer only to machine
4003 instruction operands.
4005 Symbols are evaluated to yield @{@var{section} @var{NNN}@} where
4006 @var{section} is one of text, data, bss, absolute,
4007 or undefined. @var{NNN} is a signed, 2's complement 32 bit
4010 Numbers are usually integers.
4012 A number can be a flonum or bignum. In this case, you are warned
4013 that only the low order 32 bits are used, and @command{@value{AS}} pretends
4014 these 32 bits are an integer. You may write integer-manipulating
4015 instructions that act on exotic constants, compatible with other
4018 @cindex subexpressions
4019 Subexpressions are a left parenthesis @samp{(} followed by an integer
4020 expression, followed by a right parenthesis @samp{)}; or a prefix
4021 operator followed by an argument.
4024 @subsection Operators
4026 @cindex operators, in expressions
4027 @cindex arithmetic functions
4028 @cindex functions, in expressions
4029 @dfn{Operators} are arithmetic functions, like @code{+} or @code{%}. Prefix
4030 operators are followed by an argument. Infix operators appear
4031 between their arguments. Operators may be preceded and/or followed by
4035 @subsection Prefix Operator
4037 @cindex prefix operators
4038 @command{@value{AS}} has the following @dfn{prefix operators}. They each take
4039 one argument, which must be absolute.
4041 @c the tex/end tex stuff surrounding this small table is meant to make
4042 @c it align, on the printed page, with the similar table in the next
4043 @c section (which is inside an enumerate).
4045 \global\advance\leftskip by \itemindent
4050 @dfn{Negation}. Two's complement negation.
4052 @dfn{Complementation}. Bitwise not.
4056 \global\advance\leftskip by -\itemindent
4060 @subsection Infix Operators
4062 @cindex infix operators
4063 @cindex operators, permitted arguments
4064 @dfn{Infix operators} take two arguments, one on either side. Operators
4065 have precedence, but operations with equal precedence are performed left
4066 to right. Apart from @code{+} or @option{-}, both arguments must be
4067 absolute, and the result is absolute.
4070 @cindex operator precedence
4071 @cindex precedence of operators
4078 @dfn{Multiplication}.
4081 @dfn{Division}. Truncation is the same as the C operator @samp{/}
4087 @dfn{Shift Left}. Same as the C operator @samp{<<}.
4090 @dfn{Shift Right}. Same as the C operator @samp{>>}.
4094 Intermediate precedence
4099 @dfn{Bitwise Inclusive Or}.
4105 @dfn{Bitwise Exclusive Or}.
4108 @dfn{Bitwise Or Not}.
4115 @cindex addition, permitted arguments
4116 @cindex plus, permitted arguments
4117 @cindex arguments for addition
4119 @dfn{Addition}. If either argument is absolute, the result has the section of
4120 the other argument. You may not add together arguments from different
4123 @cindex subtraction, permitted arguments
4124 @cindex minus, permitted arguments
4125 @cindex arguments for subtraction
4127 @dfn{Subtraction}. If the right argument is absolute, the
4128 result has the section of the left argument.
4129 If both arguments are in the same section, the result is absolute.
4130 You may not subtract arguments from different sections.
4131 @c FIXME is there still something useful to say about undefined - undefined ?
4133 @cindex comparison expressions
4134 @cindex expressions, comparison
4139 @dfn{Is Not Equal To}
4143 @dfn{Is Greater Than}
4145 @dfn{Is Greater Than Or Equal To}
4147 @dfn{Is Less Than Or Equal To}
4149 The comparison operators can be used as infix operators. A true results has a
4150 value of -1 whereas a false result has a value of 0. Note, these operators
4151 perform signed comparisons.
4154 @item Lowest Precedence
4163 These two logical operations can be used to combine the results of sub
4164 expressions. Note, unlike the comparison operators a true result returns a
4165 value of 1 but a false results does still return 0. Also note that the logical
4166 or operator has a slightly lower precedence than logical and.
4171 In short, it's only meaningful to add or subtract the @emph{offsets} in an
4172 address; you can only have a defined section in one of the two arguments.
4175 @chapter Assembler Directives
4177 @cindex directives, machine independent
4178 @cindex pseudo-ops, machine independent
4179 @cindex machine independent directives
4180 All assembler directives have names that begin with a period (@samp{.}).
4181 The rest of the name is letters, usually in lower case.
4183 This chapter discusses directives that are available regardless of the
4184 target machine configuration for the @sc{gnu} assembler.
4186 Some machine configurations provide additional directives.
4187 @xref{Machine Dependencies}.
4190 @ifset machine-directives
4191 @xref{Machine Dependencies}, for additional directives.
4196 * Abort:: @code{.abort}
4198 * ABORT (COFF):: @code{.ABORT}
4201 * Align:: @code{.align @var{abs-expr} , @var{abs-expr}}
4202 * Altmacro:: @code{.altmacro}
4203 * Ascii:: @code{.ascii "@var{string}"}@dots{}
4204 * Asciz:: @code{.asciz "@var{string}"}@dots{}
4205 * Balign:: @code{.balign @var{abs-expr} , @var{abs-expr}}
4206 * Bundle directives:: @code{.bundle_align_mode @var{abs-expr}}, @code{.bundle_lock}, @code{.bundle_unlock}
4207 * Byte:: @code{.byte @var{expressions}}
4208 * CFI directives:: @code{.cfi_startproc [simple]}, @code{.cfi_endproc}, etc.
4209 * Comm:: @code{.comm @var{symbol} , @var{length} }
4210 * Data:: @code{.data @var{subsection}}
4212 * Def:: @code{.def @var{name}}
4215 * Desc:: @code{.desc @var{symbol}, @var{abs-expression}}
4221 * Double:: @code{.double @var{flonums}}
4222 * Eject:: @code{.eject}
4223 * Else:: @code{.else}
4224 * Elseif:: @code{.elseif}
4227 * Endef:: @code{.endef}
4230 * Endfunc:: @code{.endfunc}
4231 * Endif:: @code{.endif}
4232 * Equ:: @code{.equ @var{symbol}, @var{expression}}
4233 * Equiv:: @code{.equiv @var{symbol}, @var{expression}}
4234 * Eqv:: @code{.eqv @var{symbol}, @var{expression}}
4236 * Error:: @code{.error @var{string}}
4237 * Exitm:: @code{.exitm}
4238 * Extern:: @code{.extern}
4239 * Fail:: @code{.fail}
4240 * File:: @code{.file}
4241 * Fill:: @code{.fill @var{repeat} , @var{size} , @var{value}}
4242 * Float:: @code{.float @var{flonums}}
4243 * Func:: @code{.func}
4244 * Global:: @code{.global @var{symbol}}, @code{.globl @var{symbol}}
4246 * Gnu_attribute:: @code{.gnu_attribute @var{tag},@var{value}}
4247 * Hidden:: @code{.hidden @var{names}}
4250 * hword:: @code{.hword @var{expressions}}
4251 * Ident:: @code{.ident}
4252 * If:: @code{.if @var{absolute expression}}
4253 * Incbin:: @code{.incbin "@var{file}"[,@var{skip}[,@var{count}]]}
4254 * Include:: @code{.include "@var{file}"}
4255 * Int:: @code{.int @var{expressions}}
4257 * Internal:: @code{.internal @var{names}}
4260 * Irp:: @code{.irp @var{symbol},@var{values}}@dots{}
4261 * Irpc:: @code{.irpc @var{symbol},@var{values}}@dots{}
4262 * Lcomm:: @code{.lcomm @var{symbol} , @var{length}}
4263 * Lflags:: @code{.lflags}
4264 @ifclear no-line-dir
4265 * Line:: @code{.line @var{line-number}}
4268 * Linkonce:: @code{.linkonce [@var{type}]}
4269 * List:: @code{.list}
4270 * Ln:: @code{.ln @var{line-number}}
4271 * Loc:: @code{.loc @var{fileno} @var{lineno}}
4272 * Loc_mark_labels:: @code{.loc_mark_labels @var{enable}}
4274 * Local:: @code{.local @var{names}}
4277 * Long:: @code{.long @var{expressions}}
4279 * Lsym:: @code{.lsym @var{symbol}, @var{expression}}
4282 * Macro:: @code{.macro @var{name} @var{args}}@dots{}
4283 * MRI:: @code{.mri @var{val}}
4284 * Noaltmacro:: @code{.noaltmacro}
4285 * Nolist:: @code{.nolist}
4286 * Octa:: @code{.octa @var{bignums}}
4287 * Offset:: @code{.offset @var{loc}}
4288 * Org:: @code{.org @var{new-lc}, @var{fill}}
4289 * P2align:: @code{.p2align @var{abs-expr}, @var{abs-expr}, @var{abs-expr}}
4291 * PopSection:: @code{.popsection}
4292 * Previous:: @code{.previous}
4295 * Print:: @code{.print @var{string}}
4297 * Protected:: @code{.protected @var{names}}
4300 * Psize:: @code{.psize @var{lines}, @var{columns}}
4301 * Purgem:: @code{.purgem @var{name}}
4303 * PushSection:: @code{.pushsection @var{name}}
4306 * Quad:: @code{.quad @var{bignums}}
4307 * Reloc:: @code{.reloc @var{offset}, @var{reloc_name}[, @var{expression}]}
4308 * Rept:: @code{.rept @var{count}}
4309 * Sbttl:: @code{.sbttl "@var{subheading}"}
4311 * Scl:: @code{.scl @var{class}}
4314 * Section:: @code{.section @var{name}[, @var{flags}]}
4317 * Set:: @code{.set @var{symbol}, @var{expression}}
4318 * Short:: @code{.short @var{expressions}}
4319 * Single:: @code{.single @var{flonums}}
4321 * Size:: @code{.size [@var{name} , @var{expression}]}
4323 @ifclear no-space-dir
4324 * Skip:: @code{.skip @var{size} , @var{fill}}
4327 * Sleb128:: @code{.sleb128 @var{expressions}}
4328 @ifclear no-space-dir
4329 * Space:: @code{.space @var{size} , @var{fill}}
4332 * Stab:: @code{.stabd, .stabn, .stabs}
4335 * String:: @code{.string "@var{str}"}, @code{.string8 "@var{str}"}, @code{.string16 "@var{str}"}, @code{.string32 "@var{str}"}, @code{.string64 "@var{str}"}
4336 * Struct:: @code{.struct @var{expression}}
4338 * SubSection:: @code{.subsection}
4339 * Symver:: @code{.symver @var{name},@var{name2@@nodename}}
4343 * Tag:: @code{.tag @var{structname}}
4346 * Text:: @code{.text @var{subsection}}
4347 * Title:: @code{.title "@var{heading}"}
4349 * Type:: @code{.type <@var{int} | @var{name} , @var{type description}>}
4352 * Uleb128:: @code{.uleb128 @var{expressions}}
4354 * Val:: @code{.val @var{addr}}
4358 * Version:: @code{.version "@var{string}"}
4359 * VTableEntry:: @code{.vtable_entry @var{table}, @var{offset}}
4360 * VTableInherit:: @code{.vtable_inherit @var{child}, @var{parent}}
4363 * Warning:: @code{.warning @var{string}}
4364 * Weak:: @code{.weak @var{names}}
4365 * Weakref:: @code{.weakref @var{alias}, @var{symbol}}
4366 * Word:: @code{.word @var{expressions}}
4367 * Deprecated:: Deprecated Directives
4371 @section @code{.abort}
4373 @cindex @code{abort} directive
4374 @cindex stopping the assembly
4375 This directive stops the assembly immediately. It is for
4376 compatibility with other assemblers. The original idea was that the
4377 assembly language source would be piped into the assembler. If the sender
4378 of the source quit, it could use this directive tells @command{@value{AS}} to
4379 quit also. One day @code{.abort} will not be supported.
4383 @section @code{.ABORT} (COFF)
4385 @cindex @code{ABORT} directive
4386 When producing COFF output, @command{@value{AS}} accepts this directive as a
4387 synonym for @samp{.abort}.
4390 When producing @code{b.out} output, @command{@value{AS}} accepts this directive,
4396 @section @code{.align @var{abs-expr}, @var{abs-expr}, @var{abs-expr}}
4398 @cindex padding the location counter
4399 @cindex @code{align} directive
4400 Pad the location counter (in the current subsection) to a particular storage
4401 boundary. The first expression (which must be absolute) is the alignment
4402 required, as described below.
4404 The second expression (also absolute) gives the fill value to be stored in the
4405 padding bytes. It (and the comma) may be omitted. If it is omitted, the
4406 padding bytes are normally zero. However, on some systems, if the section is
4407 marked as containing code and the fill value is omitted, the space is filled
4408 with no-op instructions.
4410 The third expression is also absolute, and is also optional. If it is present,
4411 it is the maximum number of bytes that should be skipped by this alignment
4412 directive. If doing the alignment would require skipping more bytes than the
4413 specified maximum, then the alignment is not done at all. You can omit the
4414 fill value (the second argument) entirely by simply using two commas after the
4415 required alignment; this can be useful if you want the alignment to be filled
4416 with no-op instructions when appropriate.
4418 The way the required alignment is specified varies from system to system.
4419 For the arc, hppa, i386 using ELF, i860, iq2000, m68k, or1k,
4420 s390, sparc, tic4x, tic80 and xtensa, the first expression is the
4421 alignment request in bytes. For example @samp{.align 8} advances
4422 the location counter until it is a multiple of 8. If the location counter
4423 is already a multiple of 8, no change is needed. For the tic54x, the
4424 first expression is the alignment request in words.
4426 For other systems, including ppc, i386 using a.out format, arm and
4427 strongarm, it is the
4428 number of low-order zero bits the location counter must have after
4429 advancement. For example @samp{.align 3} advances the location
4430 counter until it a multiple of 8. If the location counter is already a
4431 multiple of 8, no change is needed.
4433 This inconsistency is due to the different behaviors of the various
4434 native assemblers for these systems which GAS must emulate.
4435 GAS also provides @code{.balign} and @code{.p2align} directives,
4436 described later, which have a consistent behavior across all
4437 architectures (but are specific to GAS).
4440 @section @code{.altmacro}
4441 Enable alternate macro mode, enabling:
4444 @item LOCAL @var{name} [ , @dots{} ]
4445 One additional directive, @code{LOCAL}, is available. It is used to
4446 generate a string replacement for each of the @var{name} arguments, and
4447 replace any instances of @var{name} in each macro expansion. The
4448 replacement string is unique in the assembly, and different for each
4449 separate macro expansion. @code{LOCAL} allows you to write macros that
4450 define symbols, without fear of conflict between separate macro expansions.
4452 @item String delimiters
4453 You can write strings delimited in these other ways besides
4454 @code{"@var{string}"}:
4457 @item '@var{string}'
4458 You can delimit strings with single-quote characters.
4460 @item <@var{string}>
4461 You can delimit strings with matching angle brackets.
4464 @item single-character string escape
4465 To include any single character literally in a string (even if the
4466 character would otherwise have some special meaning), you can prefix the
4467 character with @samp{!} (an exclamation mark). For example, you can
4468 write @samp{<4.3 !> 5.4!!>} to get the literal text @samp{4.3 > 5.4!}.
4470 @item Expression results as strings
4471 You can write @samp{%@var{expr}} to evaluate the expression @var{expr}
4472 and use the result as a string.
4476 @section @code{.ascii "@var{string}"}@dots{}
4478 @cindex @code{ascii} directive
4479 @cindex string literals
4480 @code{.ascii} expects zero or more string literals (@pxref{Strings})
4481 separated by commas. It assembles each string (with no automatic
4482 trailing zero byte) into consecutive addresses.
4485 @section @code{.asciz "@var{string}"}@dots{}
4487 @cindex @code{asciz} directive
4488 @cindex zero-terminated strings
4489 @cindex null-terminated strings
4490 @code{.asciz} is just like @code{.ascii}, but each string is followed by
4491 a zero byte. The ``z'' in @samp{.asciz} stands for ``zero''.
4494 @section @code{.balign[wl] @var{abs-expr}, @var{abs-expr}, @var{abs-expr}}
4496 @cindex padding the location counter given number of bytes
4497 @cindex @code{balign} directive
4498 Pad the location counter (in the current subsection) to a particular
4499 storage boundary. The first expression (which must be absolute) is the
4500 alignment request in bytes. For example @samp{.balign 8} advances
4501 the location counter until it is a multiple of 8. If the location counter
4502 is already a multiple of 8, no change is needed.
4504 The second expression (also absolute) gives the fill value to be stored in the
4505 padding bytes. It (and the comma) may be omitted. If it is omitted, the
4506 padding bytes are normally zero. However, on some systems, if the section is
4507 marked as containing code and the fill value is omitted, the space is filled
4508 with no-op instructions.
4510 The third expression is also absolute, and is also optional. If it is present,
4511 it is the maximum number of bytes that should be skipped by this alignment
4512 directive. If doing the alignment would require skipping more bytes than the
4513 specified maximum, then the alignment is not done at all. You can omit the
4514 fill value (the second argument) entirely by simply using two commas after the
4515 required alignment; this can be useful if you want the alignment to be filled
4516 with no-op instructions when appropriate.
4518 @cindex @code{balignw} directive
4519 @cindex @code{balignl} directive
4520 The @code{.balignw} and @code{.balignl} directives are variants of the
4521 @code{.balign} directive. The @code{.balignw} directive treats the fill
4522 pattern as a two byte word value. The @code{.balignl} directives treats the
4523 fill pattern as a four byte longword value. For example, @code{.balignw
4524 4,0x368d} will align to a multiple of 4. If it skips two bytes, they will be
4525 filled in with the value 0x368d (the exact placement of the bytes depends upon
4526 the endianness of the processor). If it skips 1 or 3 bytes, the fill value is
4529 @node Bundle directives
4530 @section @code{.bundle_align_mode @var{abs-expr}}
4531 @cindex @code{bundle_align_mode} directive
4533 @cindex instruction bundle
4534 @cindex aligned instruction bundle
4535 @code{.bundle_align_mode} enables or disables @dfn{aligned instruction
4536 bundle} mode. In this mode, sequences of adjacent instructions are grouped
4537 into fixed-sized @dfn{bundles}. If the argument is zero, this mode is
4538 disabled (which is the default state). If the argument it not zero, it
4539 gives the size of an instruction bundle as a power of two (as for the
4540 @code{.p2align} directive, @pxref{P2align}).
4542 For some targets, it's an ABI requirement that no instruction may span a
4543 certain aligned boundary. A @dfn{bundle} is simply a sequence of
4544 instructions that starts on an aligned boundary. For example, if
4545 @var{abs-expr} is @code{5} then the bundle size is 32, so each aligned
4546 chunk of 32 bytes is a bundle. When aligned instruction bundle mode is in
4547 effect, no single instruction may span a boundary between bundles. If an
4548 instruction would start too close to the end of a bundle for the length of
4549 that particular instruction to fit within the bundle, then the space at the
4550 end of that bundle is filled with no-op instructions so the instruction
4551 starts in the next bundle. As a corollary, it's an error if any single
4552 instruction's encoding is longer than the bundle size.
4554 @section @code{.bundle_lock} and @code{.bundle_unlock}
4555 @cindex @code{bundle_lock} directive
4556 @cindex @code{bundle_unlock} directive
4557 The @code{.bundle_lock} and directive @code{.bundle_unlock} directives
4558 allow explicit control over instruction bundle padding. These directives
4559 are only valid when @code{.bundle_align_mode} has been used to enable
4560 aligned instruction bundle mode. It's an error if they appear when
4561 @code{.bundle_align_mode} has not been used at all, or when the last
4562 directive was @w{@code{.bundle_align_mode 0}}.
4564 @cindex bundle-locked
4565 For some targets, it's an ABI requirement that certain instructions may
4566 appear only as part of specified permissible sequences of multiple
4567 instructions, all within the same bundle. A pair of @code{.bundle_lock}
4568 and @code{.bundle_unlock} directives define a @dfn{bundle-locked}
4569 instruction sequence. For purposes of aligned instruction bundle mode, a
4570 sequence starting with @code{.bundle_lock} and ending with
4571 @code{.bundle_unlock} is treated as a single instruction. That is, the
4572 entire sequence must fit into a single bundle and may not span a bundle
4573 boundary. If necessary, no-op instructions will be inserted before the
4574 first instruction of the sequence so that the whole sequence starts on an
4575 aligned bundle boundary. It's an error if the sequence is longer than the
4578 For convenience when using @code{.bundle_lock} and @code{.bundle_unlock}
4579 inside assembler macros (@pxref{Macro}), bundle-locked sequences may be
4580 nested. That is, a second @code{.bundle_lock} directive before the next
4581 @code{.bundle_unlock} directive has no effect except that it must be
4582 matched by another closing @code{.bundle_unlock} so that there is the
4583 same number of @code{.bundle_lock} and @code{.bundle_unlock} directives.
4586 @section @code{.byte @var{expressions}}
4588 @cindex @code{byte} directive
4589 @cindex integers, one byte
4590 @code{.byte} expects zero or more expressions, separated by commas.
4591 Each expression is assembled into the next byte.
4593 @node CFI directives
4594 @section @code{.cfi_sections @var{section_list}}
4595 @cindex @code{cfi_sections} directive
4596 @code{.cfi_sections} may be used to specify whether CFI directives
4597 should emit @code{.eh_frame} section and/or @code{.debug_frame} section.
4598 If @var{section_list} is @code{.eh_frame}, @code{.eh_frame} is emitted,
4599 if @var{section_list} is @code{.debug_frame}, @code{.debug_frame} is emitted.
4600 To emit both use @code{.eh_frame, .debug_frame}. The default if this
4601 directive is not used is @code{.cfi_sections .eh_frame}.
4603 @section @code{.cfi_startproc [simple]}
4604 @cindex @code{cfi_startproc} directive
4605 @code{.cfi_startproc} is used at the beginning of each function that
4606 should have an entry in @code{.eh_frame}. It initializes some internal
4607 data structures. Don't forget to close the function by
4608 @code{.cfi_endproc}.
4610 Unless @code{.cfi_startproc} is used along with parameter @code{simple}
4611 it also emits some architecture dependent initial CFI instructions.
4613 @section @code{.cfi_endproc}
4614 @cindex @code{cfi_endproc} directive
4615 @code{.cfi_endproc} is used at the end of a function where it closes its
4616 unwind entry previously opened by
4617 @code{.cfi_startproc}, and emits it to @code{.eh_frame}.
4619 @section @code{.cfi_personality @var{encoding} [, @var{exp}]}
4620 @code{.cfi_personality} defines personality routine and its encoding.
4621 @var{encoding} must be a constant determining how the personality
4622 should be encoded. If it is 255 (@code{DW_EH_PE_omit}), second
4623 argument is not present, otherwise second argument should be
4624 a constant or a symbol name. When using indirect encodings,
4625 the symbol provided should be the location where personality
4626 can be loaded from, not the personality routine itself.
4627 The default after @code{.cfi_startproc} is @code{.cfi_personality 0xff},
4628 no personality routine.
4630 @section @code{.cfi_lsda @var{encoding} [, @var{exp}]}
4631 @code{.cfi_lsda} defines LSDA and its encoding.
4632 @var{encoding} must be a constant determining how the LSDA
4633 should be encoded. If it is 255 (@code{DW_EH_PE_omit}), second
4634 argument is not present, otherwise second argument should be a constant
4635 or a symbol name. The default after @code{.cfi_startproc} is @code{.cfi_lsda 0xff},
4638 @section @code{.cfi_def_cfa @var{register}, @var{offset}}
4639 @code{.cfi_def_cfa} defines a rule for computing CFA as: @i{take
4640 address from @var{register} and add @var{offset} to it}.
4642 @section @code{.cfi_def_cfa_register @var{register}}
4643 @code{.cfi_def_cfa_register} modifies a rule for computing CFA. From
4644 now on @var{register} will be used instead of the old one. Offset
4647 @section @code{.cfi_def_cfa_offset @var{offset}}
4648 @code{.cfi_def_cfa_offset} modifies a rule for computing CFA. Register
4649 remains the same, but @var{offset} is new. Note that it is the
4650 absolute offset that will be added to a defined register to compute
4653 @section @code{.cfi_adjust_cfa_offset @var{offset}}
4654 Same as @code{.cfi_def_cfa_offset} but @var{offset} is a relative
4655 value that is added/substracted from the previous offset.
4657 @section @code{.cfi_offset @var{register}, @var{offset}}
4658 Previous value of @var{register} is saved at offset @var{offset} from
4661 @section @code{.cfi_rel_offset @var{register}, @var{offset}}
4662 Previous value of @var{register} is saved at offset @var{offset} from
4663 the current CFA register. This is transformed to @code{.cfi_offset}
4664 using the known displacement of the CFA register from the CFA.
4665 This is often easier to use, because the number will match the
4666 code it's annotating.
4668 @section @code{.cfi_register @var{register1}, @var{register2}}
4669 Previous value of @var{register1} is saved in register @var{register2}.
4671 @section @code{.cfi_restore @var{register}}
4672 @code{.cfi_restore} says that the rule for @var{register} is now the
4673 same as it was at the beginning of the function, after all initial
4674 instruction added by @code{.cfi_startproc} were executed.
4676 @section @code{.cfi_undefined @var{register}}
4677 From now on the previous value of @var{register} can't be restored anymore.
4679 @section @code{.cfi_same_value @var{register}}
4680 Current value of @var{register} is the same like in the previous frame,
4681 i.e. no restoration needed.
4683 @section @code{.cfi_remember_state},
4684 First save all current rules for all registers by @code{.cfi_remember_state},
4685 then totally screw them up by subsequent @code{.cfi_*} directives and when
4686 everything is hopelessly bad, use @code{.cfi_restore_state} to restore
4687 the previous saved state.
4689 @section @code{.cfi_return_column @var{register}}
4690 Change return column @var{register}, i.e. the return address is either
4691 directly in @var{register} or can be accessed by rules for @var{register}.
4693 @section @code{.cfi_signal_frame}
4694 Mark current function as signal trampoline.
4696 @section @code{.cfi_window_save}
4697 SPARC register window has been saved.
4699 @section @code{.cfi_escape} @var{expression}[, @dots{}]
4700 Allows the user to add arbitrary bytes to the unwind info. One
4701 might use this to add OS-specific CFI opcodes, or generic CFI
4702 opcodes that GAS does not yet support.
4704 @section @code{.cfi_val_encoded_addr @var{register}, @var{encoding}, @var{label}}
4705 The current value of @var{register} is @var{label}. The value of @var{label}
4706 will be encoded in the output file according to @var{encoding}; see the
4707 description of @code{.cfi_personality} for details on this encoding.
4709 The usefulness of equating a register to a fixed label is probably
4710 limited to the return address register. Here, it can be useful to
4711 mark a code segment that has only one return address which is reached
4712 by a direct branch and no copy of the return address exists in memory
4713 or another register.
4716 @section @code{.comm @var{symbol} , @var{length} }
4718 @cindex @code{comm} directive
4719 @cindex symbol, common
4720 @code{.comm} declares a common symbol named @var{symbol}. When linking, a
4721 common symbol in one object file may be merged with a defined or common symbol
4722 of the same name in another object file. If @code{@value{LD}} does not see a
4723 definition for the symbol--just one or more common symbols--then it will
4724 allocate @var{length} bytes of uninitialized memory. @var{length} must be an
4725 absolute expression. If @code{@value{LD}} sees multiple common symbols with
4726 the same name, and they do not all have the same size, it will allocate space
4727 using the largest size.
4730 When using ELF or (as a GNU extension) PE, the @code{.comm} directive takes
4731 an optional third argument. This is the desired alignment of the symbol,
4732 specified for ELF as a byte boundary (for example, an alignment of 16 means
4733 that the least significant 4 bits of the address should be zero), and for PE
4734 as a power of two (for example, an alignment of 5 means aligned to a 32-byte
4735 boundary). The alignment must be an absolute expression, and it must be a
4736 power of two. If @code{@value{LD}} allocates uninitialized memory for the
4737 common symbol, it will use the alignment when placing the symbol. If no
4738 alignment is specified, @command{@value{AS}} will set the alignment to the
4739 largest power of two less than or equal to the size of the symbol, up to a
4740 maximum of 16 on ELF, or the default section alignment of 4 on PE@footnote{This
4741 is not the same as the executable image file alignment controlled by @code{@value{LD}}'s
4742 @samp{--section-alignment} option; image file sections in PE are aligned to
4743 multiples of 4096, which is far too large an alignment for ordinary variables.
4744 It is rather the default alignment for (non-debug) sections within object
4745 (@samp{*.o}) files, which are less strictly aligned.}.
4749 The syntax for @code{.comm} differs slightly on the HPPA. The syntax is
4750 @samp{@var{symbol} .comm, @var{length}}; @var{symbol} is optional.
4754 @section @code{.data @var{subsection}}
4756 @cindex @code{data} directive
4757 @code{.data} tells @command{@value{AS}} to assemble the following statements onto the
4758 end of the data subsection numbered @var{subsection} (which is an
4759 absolute expression). If @var{subsection} is omitted, it defaults
4764 @section @code{.def @var{name}}
4766 @cindex @code{def} directive
4767 @cindex COFF symbols, debugging
4768 @cindex debugging COFF symbols
4769 Begin defining debugging information for a symbol @var{name}; the
4770 definition extends until the @code{.endef} directive is encountered.
4773 This directive is only observed when @command{@value{AS}} is configured for COFF
4774 format output; when producing @code{b.out}, @samp{.def} is recognized,
4781 @section @code{.desc @var{symbol}, @var{abs-expression}}
4783 @cindex @code{desc} directive
4784 @cindex COFF symbol descriptor
4785 @cindex symbol descriptor, COFF
4786 This directive sets the descriptor of the symbol (@pxref{Symbol Attributes})
4787 to the low 16 bits of an absolute expression.
4790 The @samp{.desc} directive is not available when @command{@value{AS}} is
4791 configured for COFF output; it is only for @code{a.out} or @code{b.out}
4792 object format. For the sake of compatibility, @command{@value{AS}} accepts
4793 it, but produces no output, when configured for COFF.
4799 @section @code{.dim}
4801 @cindex @code{dim} directive
4802 @cindex COFF auxiliary symbol information
4803 @cindex auxiliary symbol information, COFF
4804 This directive is generated by compilers to include auxiliary debugging
4805 information in the symbol table. It is only permitted inside
4806 @code{.def}/@code{.endef} pairs.
4809 @samp{.dim} is only meaningful when generating COFF format output; when
4810 @command{@value{AS}} is generating @code{b.out}, it accepts this directive but
4816 @section @code{.double @var{flonums}}
4818 @cindex @code{double} directive
4819 @cindex floating point numbers (double)
4820 @code{.double} expects zero or more flonums, separated by commas. It
4821 assembles floating point numbers.
4823 The exact kind of floating point numbers emitted depends on how
4824 @command{@value{AS}} is configured. @xref{Machine Dependencies}.
4828 On the @value{TARGET} family @samp{.double} emits 64-bit floating-point numbers
4829 in @sc{ieee} format.
4834 @section @code{.eject}
4836 @cindex @code{eject} directive
4837 @cindex new page, in listings
4838 @cindex page, in listings
4839 @cindex listing control: new page
4840 Force a page break at this point, when generating assembly listings.
4843 @section @code{.else}
4845 @cindex @code{else} directive
4846 @code{.else} is part of the @command{@value{AS}} support for conditional
4847 assembly; see @ref{If,,@code{.if}}. It marks the beginning of a section
4848 of code to be assembled if the condition for the preceding @code{.if}
4852 @section @code{.elseif}
4854 @cindex @code{elseif} directive
4855 @code{.elseif} is part of the @command{@value{AS}} support for conditional
4856 assembly; see @ref{If,,@code{.if}}. It is shorthand for beginning a new
4857 @code{.if} block that would otherwise fill the entire @code{.else} section.
4860 @section @code{.end}
4862 @cindex @code{end} directive
4863 @code{.end} marks the end of the assembly file. @command{@value{AS}} does not
4864 process anything in the file past the @code{.end} directive.
4868 @section @code{.endef}
4870 @cindex @code{endef} directive
4871 This directive flags the end of a symbol definition begun with
4875 @samp{.endef} is only meaningful when generating COFF format output; if
4876 @command{@value{AS}} is configured to generate @code{b.out}, it accepts this
4877 directive but ignores it.
4882 @section @code{.endfunc}
4883 @cindex @code{endfunc} directive
4884 @code{.endfunc} marks the end of a function specified with @code{.func}.
4887 @section @code{.endif}
4889 @cindex @code{endif} directive
4890 @code{.endif} is part of the @command{@value{AS}} support for conditional assembly;
4891 it marks the end of a block of code that is only assembled
4892 conditionally. @xref{If,,@code{.if}}.
4895 @section @code{.equ @var{symbol}, @var{expression}}
4897 @cindex @code{equ} directive
4898 @cindex assigning values to symbols
4899 @cindex symbols, assigning values to
4900 This directive sets the value of @var{symbol} to @var{expression}.
4901 It is synonymous with @samp{.set}; see @ref{Set,,@code{.set}}.
4904 The syntax for @code{equ} on the HPPA is
4905 @samp{@var{symbol} .equ @var{expression}}.
4909 The syntax for @code{equ} on the Z80 is
4910 @samp{@var{symbol} equ @var{expression}}.
4911 On the Z80 it is an eror if @var{symbol} is already defined,
4912 but the symbol is not protected from later redefinition.
4913 Compare @ref{Equiv}.
4917 @section @code{.equiv @var{symbol}, @var{expression}}
4918 @cindex @code{equiv} directive
4919 The @code{.equiv} directive is like @code{.equ} and @code{.set}, except that
4920 the assembler will signal an error if @var{symbol} is already defined. Note a
4921 symbol which has been referenced but not actually defined is considered to be
4924 Except for the contents of the error message, this is roughly equivalent to
4931 plus it protects the symbol from later redefinition.
4934 @section @code{.eqv @var{symbol}, @var{expression}}
4935 @cindex @code{eqv} directive
4936 The @code{.eqv} directive is like @code{.equiv}, but no attempt is made to
4937 evaluate the expression or any part of it immediately. Instead each time
4938 the resulting symbol is used in an expression, a snapshot of its current
4942 @section @code{.err}
4943 @cindex @code{err} directive
4944 If @command{@value{AS}} assembles a @code{.err} directive, it will print an error
4945 message and, unless the @option{-Z} option was used, it will not generate an
4946 object file. This can be used to signal an error in conditionally compiled code.
4949 @section @code{.error "@var{string}"}
4950 @cindex error directive
4952 Similarly to @code{.err}, this directive emits an error, but you can specify a
4953 string that will be emitted as the error message. If you don't specify the
4954 message, it defaults to @code{".error directive invoked in source file"}.
4955 @xref{Errors, ,Error and Warning Messages}.
4958 .error "This code has not been assembled and tested."
4962 @section @code{.exitm}
4963 Exit early from the current macro definition. @xref{Macro}.
4966 @section @code{.extern}
4968 @cindex @code{extern} directive
4969 @code{.extern} is accepted in the source program---for compatibility
4970 with other assemblers---but it is ignored. @command{@value{AS}} treats
4971 all undefined symbols as external.
4974 @section @code{.fail @var{expression}}
4976 @cindex @code{fail} directive
4977 Generates an error or a warning. If the value of the @var{expression} is 500
4978 or more, @command{@value{AS}} will print a warning message. If the value is less
4979 than 500, @command{@value{AS}} will print an error message. The message will
4980 include the value of @var{expression}. This can occasionally be useful inside
4981 complex nested macros or conditional assembly.
4984 @section @code{.file}
4985 @cindex @code{file} directive
4987 @ifclear no-file-dir
4988 There are two different versions of the @code{.file} directive. Targets
4989 that support DWARF2 line number information use the DWARF2 version of
4990 @code{.file}. Other targets use the default version.
4992 @subheading Default Version
4994 @cindex logical file name
4995 @cindex file name, logical
4996 This version of the @code{.file} directive tells @command{@value{AS}} that we
4997 are about to start a new logical file. The syntax is:
5003 @var{string} is the new file name. In general, the filename is
5004 recognized whether or not it is surrounded by quotes @samp{"}; but if you wish
5005 to specify an empty file name, you must give the quotes--@code{""}. This
5006 statement may go away in future: it is only recognized to be compatible with
5007 old @command{@value{AS}} programs.
5009 @subheading DWARF2 Version
5012 When emitting DWARF2 line number information, @code{.file} assigns filenames
5013 to the @code{.debug_line} file name table. The syntax is:
5016 .file @var{fileno} @var{filename}
5019 The @var{fileno} operand should be a unique positive integer to use as the
5020 index of the entry in the table. The @var{filename} operand is a C string
5023 The detail of filename indices is exposed to the user because the filename
5024 table is shared with the @code{.debug_info} section of the DWARF2 debugging
5025 information, and thus the user must know the exact indices that table
5029 @section @code{.fill @var{repeat} , @var{size} , @var{value}}
5031 @cindex @code{fill} directive
5032 @cindex writing patterns in memory
5033 @cindex patterns, writing in memory
5034 @var{repeat}, @var{size} and @var{value} are absolute expressions.
5035 This emits @var{repeat} copies of @var{size} bytes. @var{Repeat}
5036 may be zero or more. @var{Size} may be zero or more, but if it is
5037 more than 8, then it is deemed to have the value 8, compatible with
5038 other people's assemblers. The contents of each @var{repeat} bytes
5039 is taken from an 8-byte number. The highest order 4 bytes are
5040 zero. The lowest order 4 bytes are @var{value} rendered in the
5041 byte-order of an integer on the computer @command{@value{AS}} is assembling for.
5042 Each @var{size} bytes in a repetition is taken from the lowest order
5043 @var{size} bytes of this number. Again, this bizarre behavior is
5044 compatible with other people's assemblers.
5046 @var{size} and @var{value} are optional.
5047 If the second comma and @var{value} are absent, @var{value} is
5048 assumed zero. If the first comma and following tokens are absent,
5049 @var{size} is assumed to be 1.
5052 @section @code{.float @var{flonums}}
5054 @cindex floating point numbers (single)
5055 @cindex @code{float} directive
5056 This directive assembles zero or more flonums, separated by commas. It
5057 has the same effect as @code{.single}.
5059 The exact kind of floating point numbers emitted depends on how
5060 @command{@value{AS}} is configured.
5061 @xref{Machine Dependencies}.
5065 On the @value{TARGET} family, @code{.float} emits 32-bit floating point numbers
5066 in @sc{ieee} format.
5071 @section @code{.func @var{name}[,@var{label}]}
5072 @cindex @code{func} directive
5073 @code{.func} emits debugging information to denote function @var{name}, and
5074 is ignored unless the file is assembled with debugging enabled.
5075 Only @samp{--gstabs[+]} is currently supported.
5076 @var{label} is the entry point of the function and if omitted @var{name}
5077 prepended with the @samp{leading char} is used.
5078 @samp{leading char} is usually @code{_} or nothing, depending on the target.
5079 All functions are currently defined to have @code{void} return type.
5080 The function must be terminated with @code{.endfunc}.
5083 @section @code{.global @var{symbol}}, @code{.globl @var{symbol}}
5085 @cindex @code{global} directive
5086 @cindex symbol, making visible to linker
5087 @code{.global} makes the symbol visible to @code{@value{LD}}. If you define
5088 @var{symbol} in your partial program, its value is made available to
5089 other partial programs that are linked with it. Otherwise,
5090 @var{symbol} takes its attributes from a symbol of the same name
5091 from another file linked into the same program.
5093 Both spellings (@samp{.globl} and @samp{.global}) are accepted, for
5094 compatibility with other assemblers.
5097 On the HPPA, @code{.global} is not always enough to make it accessible to other
5098 partial programs. You may need the HPPA-only @code{.EXPORT} directive as well.
5099 @xref{HPPA Directives, ,HPPA Assembler Directives}.
5104 @section @code{.gnu_attribute @var{tag},@var{value}}
5105 Record a @sc{gnu} object attribute for this file. @xref{Object Attributes}.
5108 @section @code{.hidden @var{names}}
5110 @cindex @code{hidden} directive
5112 This is one of the ELF visibility directives. The other two are
5113 @code{.internal} (@pxref{Internal,,@code{.internal}}) and
5114 @code{.protected} (@pxref{Protected,,@code{.protected}}).
5116 This directive overrides the named symbols default visibility (which is set by
5117 their binding: local, global or weak). The directive sets the visibility to
5118 @code{hidden} which means that the symbols are not visible to other components.
5119 Such symbols are always considered to be @code{protected} as well.
5123 @section @code{.hword @var{expressions}}
5125 @cindex @code{hword} directive
5126 @cindex integers, 16-bit
5127 @cindex numbers, 16-bit
5128 @cindex sixteen bit integers
5129 This expects zero or more @var{expressions}, and emits
5130 a 16 bit number for each.
5133 This directive is a synonym for @samp{.short}; depending on the target
5134 architecture, it may also be a synonym for @samp{.word}.
5138 This directive is a synonym for @samp{.short}.
5141 This directive is a synonym for both @samp{.short} and @samp{.word}.
5146 @section @code{.ident}
5148 @cindex @code{ident} directive
5150 This directive is used by some assemblers to place tags in object files. The
5151 behavior of this directive varies depending on the target. When using the
5152 a.out object file format, @command{@value{AS}} simply accepts the directive for
5153 source-file compatibility with existing assemblers, but does not emit anything
5154 for it. When using COFF, comments are emitted to the @code{.comment} or
5155 @code{.rdata} section, depending on the target. When using ELF, comments are
5156 emitted to the @code{.comment} section.
5159 @section @code{.if @var{absolute expression}}
5161 @cindex conditional assembly
5162 @cindex @code{if} directive
5163 @code{.if} marks the beginning of a section of code which is only
5164 considered part of the source program being assembled if the argument
5165 (which must be an @var{absolute expression}) is non-zero. The end of
5166 the conditional section of code must be marked by @code{.endif}
5167 (@pxref{Endif,,@code{.endif}}); optionally, you may include code for the
5168 alternative condition, flagged by @code{.else} (@pxref{Else,,@code{.else}}).
5169 If you have several conditions to check, @code{.elseif} may be used to avoid
5170 nesting blocks if/else within each subsequent @code{.else} block.
5172 The following variants of @code{.if} are also supported:
5174 @cindex @code{ifdef} directive
5175 @item .ifdef @var{symbol}
5176 Assembles the following section of code if the specified @var{symbol}
5177 has been defined. Note a symbol which has been referenced but not yet defined
5178 is considered to be undefined.
5180 @cindex @code{ifb} directive
5181 @item .ifb @var{text}
5182 Assembles the following section of code if the operand is blank (empty).
5184 @cindex @code{ifc} directive
5185 @item .ifc @var{string1},@var{string2}
5186 Assembles the following section of code if the two strings are the same. The
5187 strings may be optionally quoted with single quotes. If they are not quoted,
5188 the first string stops at the first comma, and the second string stops at the
5189 end of the line. Strings which contain whitespace should be quoted. The
5190 string comparison is case sensitive.
5192 @cindex @code{ifeq} directive
5193 @item .ifeq @var{absolute expression}
5194 Assembles the following section of code if the argument is zero.
5196 @cindex @code{ifeqs} directive
5197 @item .ifeqs @var{string1},@var{string2}
5198 Another form of @code{.ifc}. The strings must be quoted using double quotes.
5200 @cindex @code{ifge} directive
5201 @item .ifge @var{absolute expression}
5202 Assembles the following section of code if the argument is greater than or
5205 @cindex @code{ifgt} directive
5206 @item .ifgt @var{absolute expression}
5207 Assembles the following section of code if the argument is greater than zero.
5209 @cindex @code{ifle} directive
5210 @item .ifle @var{absolute expression}
5211 Assembles the following section of code if the argument is less than or equal
5214 @cindex @code{iflt} directive
5215 @item .iflt @var{absolute expression}
5216 Assembles the following section of code if the argument is less than zero.
5218 @cindex @code{ifnb} directive
5219 @item .ifnb @var{text}
5220 Like @code{.ifb}, but the sense of the test is reversed: this assembles the
5221 following section of code if the operand is non-blank (non-empty).
5223 @cindex @code{ifnc} directive
5224 @item .ifnc @var{string1},@var{string2}.
5225 Like @code{.ifc}, but the sense of the test is reversed: this assembles the
5226 following section of code if the two strings are not the same.
5228 @cindex @code{ifndef} directive
5229 @cindex @code{ifnotdef} directive
5230 @item .ifndef @var{symbol}
5231 @itemx .ifnotdef @var{symbol}
5232 Assembles the following section of code if the specified @var{symbol}
5233 has not been defined. Both spelling variants are equivalent. Note a symbol
5234 which has been referenced but not yet defined is considered to be undefined.
5236 @cindex @code{ifne} directive
5237 @item .ifne @var{absolute expression}
5238 Assembles the following section of code if the argument is not equal to zero
5239 (in other words, this is equivalent to @code{.if}).
5241 @cindex @code{ifnes} directive
5242 @item .ifnes @var{string1},@var{string2}
5243 Like @code{.ifeqs}, but the sense of the test is reversed: this assembles the
5244 following section of code if the two strings are not the same.
5248 @section @code{.incbin "@var{file}"[,@var{skip}[,@var{count}]]}
5250 @cindex @code{incbin} directive
5251 @cindex binary files, including
5252 The @code{incbin} directive includes @var{file} verbatim at the current
5253 location. You can control the search paths used with the @samp{-I} command-line
5254 option (@pxref{Invoking,,Command-Line Options}). Quotation marks are required
5257 The @var{skip} argument skips a number of bytes from the start of the
5258 @var{file}. The @var{count} argument indicates the maximum number of bytes to
5259 read. Note that the data is not aligned in any way, so it is the user's
5260 responsibility to make sure that proper alignment is provided both before and
5261 after the @code{incbin} directive.
5264 @section @code{.include "@var{file}"}
5266 @cindex @code{include} directive
5267 @cindex supporting files, including
5268 @cindex files, including
5269 This directive provides a way to include supporting files at specified
5270 points in your source program. The code from @var{file} is assembled as
5271 if it followed the point of the @code{.include}; when the end of the
5272 included file is reached, assembly of the original file continues. You
5273 can control the search paths used with the @samp{-I} command-line option
5274 (@pxref{Invoking,,Command-Line Options}). Quotation marks are required
5278 @section @code{.int @var{expressions}}
5280 @cindex @code{int} directive
5281 @cindex integers, 32-bit
5282 Expect zero or more @var{expressions}, of any section, separated by commas.
5283 For each expression, emit a number that, at run time, is the value of that
5284 expression. The byte order and bit size of the number depends on what kind
5285 of target the assembly is for.
5289 On most forms of the H8/300, @code{.int} emits 16-bit
5290 integers. On the H8/300H and the Renesas SH, however, @code{.int} emits
5297 @section @code{.internal @var{names}}
5299 @cindex @code{internal} directive
5301 This is one of the ELF visibility directives. The other two are
5302 @code{.hidden} (@pxref{Hidden,,@code{.hidden}}) and
5303 @code{.protected} (@pxref{Protected,,@code{.protected}}).
5305 This directive overrides the named symbols default visibility (which is set by
5306 their binding: local, global or weak). The directive sets the visibility to
5307 @code{internal} which means that the symbols are considered to be @code{hidden}
5308 (i.e., not visible to other components), and that some extra, processor specific
5309 processing must also be performed upon the symbols as well.
5313 @section @code{.irp @var{symbol},@var{values}}@dots{}
5315 @cindex @code{irp} directive
5316 Evaluate a sequence of statements assigning different values to @var{symbol}.
5317 The sequence of statements starts at the @code{.irp} directive, and is
5318 terminated by an @code{.endr} directive. For each @var{value}, @var{symbol} is
5319 set to @var{value}, and the sequence of statements is assembled. If no
5320 @var{value} is listed, the sequence of statements is assembled once, with
5321 @var{symbol} set to the null string. To refer to @var{symbol} within the
5322 sequence of statements, use @var{\symbol}.
5324 For example, assembling
5332 is equivalent to assembling
5340 For some caveats with the spelling of @var{symbol}, see also @ref{Macro}.
5343 @section @code{.irpc @var{symbol},@var{values}}@dots{}
5345 @cindex @code{irpc} directive
5346 Evaluate a sequence of statements assigning different values to @var{symbol}.
5347 The sequence of statements starts at the @code{.irpc} directive, and is
5348 terminated by an @code{.endr} directive. For each character in @var{value},
5349 @var{symbol} is set to the character, and the sequence of statements is
5350 assembled. If no @var{value} is listed, the sequence of statements is
5351 assembled once, with @var{symbol} set to the null string. To refer to
5352 @var{symbol} within the sequence of statements, use @var{\symbol}.
5354 For example, assembling
5362 is equivalent to assembling
5370 For some caveats with the spelling of @var{symbol}, see also the discussion
5374 @section @code{.lcomm @var{symbol} , @var{length}}
5376 @cindex @code{lcomm} directive
5377 @cindex local common symbols
5378 @cindex symbols, local common
5379 Reserve @var{length} (an absolute expression) bytes for a local common
5380 denoted by @var{symbol}. The section and value of @var{symbol} are
5381 those of the new local common. The addresses are allocated in the bss
5382 section, so that at run-time the bytes start off zeroed. @var{Symbol}
5383 is not declared global (@pxref{Global,,@code{.global}}), so is normally
5384 not visible to @code{@value{LD}}.
5387 Some targets permit a third argument to be used with @code{.lcomm}. This
5388 argument specifies the desired alignment of the symbol in the bss section.
5392 The syntax for @code{.lcomm} differs slightly on the HPPA. The syntax is
5393 @samp{@var{symbol} .lcomm, @var{length}}; @var{symbol} is optional.
5397 @section @code{.lflags}
5399 @cindex @code{lflags} directive (ignored)
5400 @command{@value{AS}} accepts this directive, for compatibility with other
5401 assemblers, but ignores it.
5403 @ifclear no-line-dir
5405 @section @code{.line @var{line-number}}
5407 @cindex @code{line} directive
5408 @cindex logical line number
5410 Change the logical line number. @var{line-number} must be an absolute
5411 expression. The next line has that logical line number. Therefore any other
5412 statements on the current line (after a statement separator character) are
5413 reported as on logical line number @var{line-number} @minus{} 1. One day
5414 @command{@value{AS}} will no longer support this directive: it is recognized only
5415 for compatibility with existing assembler programs.
5418 Even though this is a directive associated with the @code{a.out} or
5419 @code{b.out} object-code formats, @command{@value{AS}} still recognizes it
5420 when producing COFF output, and treats @samp{.line} as though it
5421 were the COFF @samp{.ln} @emph{if} it is found outside a
5422 @code{.def}/@code{.endef} pair.
5424 Inside a @code{.def}, @samp{.line} is, instead, one of the directives
5425 used by compilers to generate auxiliary symbol information for
5430 @section @code{.linkonce [@var{type}]}
5432 @cindex @code{linkonce} directive
5433 @cindex common sections
5434 Mark the current section so that the linker only includes a single copy of it.
5435 This may be used to include the same section in several different object files,
5436 but ensure that the linker will only include it once in the final output file.
5437 The @code{.linkonce} pseudo-op must be used for each instance of the section.
5438 Duplicate sections are detected based on the section name, so it should be
5441 This directive is only supported by a few object file formats; as of this
5442 writing, the only object file format which supports it is the Portable
5443 Executable format used on Windows NT.
5445 The @var{type} argument is optional. If specified, it must be one of the
5446 following strings. For example:
5450 Not all types may be supported on all object file formats.
5454 Silently discard duplicate sections. This is the default.
5457 Warn if there are duplicate sections, but still keep only one copy.
5460 Warn if any of the duplicates have different sizes.
5463 Warn if any of the duplicates do not have exactly the same contents.
5467 @section @code{.list}
5469 @cindex @code{list} directive
5470 @cindex listing control, turning on
5471 Control (in conjunction with the @code{.nolist} directive) whether or
5472 not assembly listings are generated. These two directives maintain an
5473 internal counter (which is zero initially). @code{.list} increments the
5474 counter, and @code{.nolist} decrements it. Assembly listings are
5475 generated whenever the counter is greater than zero.
5477 By default, listings are disabled. When you enable them (with the
5478 @samp{-a} command line option; @pxref{Invoking,,Command-Line Options}),
5479 the initial value of the listing counter is one.
5482 @section @code{.ln @var{line-number}}
5484 @cindex @code{ln} directive
5485 @ifclear no-line-dir
5486 @samp{.ln} is a synonym for @samp{.line}.
5489 Tell @command{@value{AS}} to change the logical line number. @var{line-number}
5490 must be an absolute expression. The next line has that logical
5491 line number, so any other statements on the current line (after a
5492 statement separator character @code{;}) are reported as on logical
5493 line number @var{line-number} @minus{} 1.
5496 This directive is accepted, but ignored, when @command{@value{AS}} is
5497 configured for @code{b.out}; its effect is only associated with COFF
5503 @section @code{.loc @var{fileno} @var{lineno} [@var{column}] [@var{options}]}
5504 @cindex @code{loc} directive
5505 When emitting DWARF2 line number information,
5506 the @code{.loc} directive will add a row to the @code{.debug_line} line
5507 number matrix corresponding to the immediately following assembly
5508 instruction. The @var{fileno}, @var{lineno}, and optional @var{column}
5509 arguments will be applied to the @code{.debug_line} state machine before
5512 The @var{options} are a sequence of the following tokens in any order:
5516 This option will set the @code{basic_block} register in the
5517 @code{.debug_line} state machine to @code{true}.
5520 This option will set the @code{prologue_end} register in the
5521 @code{.debug_line} state machine to @code{true}.
5523 @item epilogue_begin
5524 This option will set the @code{epilogue_begin} register in the
5525 @code{.debug_line} state machine to @code{true}.
5527 @item is_stmt @var{value}
5528 This option will set the @code{is_stmt} register in the
5529 @code{.debug_line} state machine to @code{value}, which must be
5532 @item isa @var{value}
5533 This directive will set the @code{isa} register in the @code{.debug_line}
5534 state machine to @var{value}, which must be an unsigned integer.
5536 @item discriminator @var{value}
5537 This directive will set the @code{discriminator} register in the @code{.debug_line}
5538 state machine to @var{value}, which must be an unsigned integer.
5542 @node Loc_mark_labels
5543 @section @code{.loc_mark_labels @var{enable}}
5544 @cindex @code{loc_mark_labels} directive
5545 When emitting DWARF2 line number information,
5546 the @code{.loc_mark_labels} directive makes the assembler emit an entry
5547 to the @code{.debug_line} line number matrix with the @code{basic_block}
5548 register in the state machine set whenever a code label is seen.
5549 The @var{enable} argument should be either 1 or 0, to enable or disable
5550 this function respectively.
5554 @section @code{.local @var{names}}
5556 @cindex @code{local} directive
5557 This directive, which is available for ELF targets, marks each symbol in
5558 the comma-separated list of @code{names} as a local symbol so that it
5559 will not be externally visible. If the symbols do not already exist,
5560 they will be created.
5562 For targets where the @code{.lcomm} directive (@pxref{Lcomm}) does not
5563 accept an alignment argument, which is the case for most ELF targets,
5564 the @code{.local} directive can be used in combination with @code{.comm}
5565 (@pxref{Comm}) to define aligned local common data.
5569 @section @code{.long @var{expressions}}
5571 @cindex @code{long} directive
5572 @code{.long} is the same as @samp{.int}. @xref{Int,,@code{.int}}.
5575 @c no one seems to know what this is for or whether this description is
5576 @c what it really ought to do
5578 @section @code{.lsym @var{symbol}, @var{expression}}
5580 @cindex @code{lsym} directive
5581 @cindex symbol, not referenced in assembly
5582 @code{.lsym} creates a new symbol named @var{symbol}, but does not put it in
5583 the hash table, ensuring it cannot be referenced by name during the
5584 rest of the assembly. This sets the attributes of the symbol to be
5585 the same as the expression value:
5587 @var{other} = @var{descriptor} = 0
5588 @var{type} = @r{(section of @var{expression})}
5589 @var{value} = @var{expression}
5592 The new symbol is not flagged as external.
5596 @section @code{.macro}
5599 The commands @code{.macro} and @code{.endm} allow you to define macros that
5600 generate assembly output. For example, this definition specifies a macro
5601 @code{sum} that puts a sequence of numbers into memory:
5604 .macro sum from=0, to=5
5613 With that definition, @samp{SUM 0,5} is equivalent to this assembly input:
5625 @item .macro @var{macname}
5626 @itemx .macro @var{macname} @var{macargs} @dots{}
5627 @cindex @code{macro} directive
5628 Begin the definition of a macro called @var{macname}. If your macro
5629 definition requires arguments, specify their names after the macro name,
5630 separated by commas or spaces. You can qualify the macro argument to
5631 indicate whether all invocations must specify a non-blank value (through
5632 @samp{:@code{req}}), or whether it takes all of the remaining arguments
5633 (through @samp{:@code{vararg}}). You can supply a default value for any
5634 macro argument by following the name with @samp{=@var{deflt}}. You
5635 cannot define two macros with the same @var{macname} unless it has been
5636 subject to the @code{.purgem} directive (@pxref{Purgem}) between the two
5637 definitions. For example, these are all valid @code{.macro} statements:
5641 Begin the definition of a macro called @code{comm}, which takes no
5644 @item .macro plus1 p, p1
5645 @itemx .macro plus1 p p1
5646 Either statement begins the definition of a macro called @code{plus1},
5647 which takes two arguments; within the macro definition, write
5648 @samp{\p} or @samp{\p1} to evaluate the arguments.
5650 @item .macro reserve_str p1=0 p2
5651 Begin the definition of a macro called @code{reserve_str}, with two
5652 arguments. The first argument has a default value, but not the second.
5653 After the definition is complete, you can call the macro either as
5654 @samp{reserve_str @var{a},@var{b}} (with @samp{\p1} evaluating to
5655 @var{a} and @samp{\p2} evaluating to @var{b}), or as @samp{reserve_str
5656 ,@var{b}} (with @samp{\p1} evaluating as the default, in this case
5657 @samp{0}, and @samp{\p2} evaluating to @var{b}).
5659 @item .macro m p1:req, p2=0, p3:vararg
5660 Begin the definition of a macro called @code{m}, with at least three
5661 arguments. The first argument must always have a value specified, but
5662 not the second, which instead has a default value. The third formal
5663 will get assigned all remaining arguments specified at invocation time.
5665 When you call a macro, you can specify the argument values either by
5666 position, or by keyword. For example, @samp{sum 9,17} is equivalent to
5667 @samp{sum to=17, from=9}.
5671 Note that since each of the @var{macargs} can be an identifier exactly
5672 as any other one permitted by the target architecture, there may be
5673 occasional problems if the target hand-crafts special meanings to certain
5674 characters when they occur in a special position. For example, if the colon
5675 (@code{:}) is generally permitted to be part of a symbol name, but the
5676 architecture specific code special-cases it when occurring as the final
5677 character of a symbol (to denote a label), then the macro parameter
5678 replacement code will have no way of knowing that and consider the whole
5679 construct (including the colon) an identifier, and check only this
5680 identifier for being the subject to parameter substitution. So for example
5681 this macro definition:
5689 might not work as expected. Invoking @samp{label foo} might not create a label
5690 called @samp{foo} but instead just insert the text @samp{\l:} into the
5691 assembler source, probably generating an error about an unrecognised
5694 Similarly problems might occur with the period character (@samp{.})
5695 which is often allowed inside opcode names (and hence identifier names). So
5696 for example constructing a macro to build an opcode from a base name and a
5697 length specifier like this:
5700 .macro opcode base length
5705 and invoking it as @samp{opcode store l} will not create a @samp{store.l}
5706 instruction but instead generate some kind of error as the assembler tries to
5707 interpret the text @samp{\base.\length}.
5709 There are several possible ways around this problem:
5712 @item Insert white space
5713 If it is possible to use white space characters then this is the simplest
5722 @item Use @samp{\()}
5723 The string @samp{\()} can be used to separate the end of a macro argument from
5724 the following text. eg:
5727 .macro opcode base length
5732 @item Use the alternate macro syntax mode
5733 In the alternative macro syntax mode the ampersand character (@samp{&}) can be
5734 used as a separator. eg:
5744 Note: this problem of correctly identifying string parameters to pseudo ops
5745 also applies to the identifiers used in @code{.irp} (@pxref{Irp})
5746 and @code{.irpc} (@pxref{Irpc}) as well.
5749 @cindex @code{endm} directive
5750 Mark the end of a macro definition.
5753 @cindex @code{exitm} directive
5754 Exit early from the current macro definition.
5756 @cindex number of macros executed
5757 @cindex macros, count executed
5759 @command{@value{AS}} maintains a counter of how many macros it has
5760 executed in this pseudo-variable; you can copy that number to your
5761 output with @samp{\@@}, but @emph{only within a macro definition}.
5763 @item LOCAL @var{name} [ , @dots{} ]
5764 @emph{Warning: @code{LOCAL} is only available if you select ``alternate
5765 macro syntax'' with @samp{--alternate} or @code{.altmacro}.}
5766 @xref{Altmacro,,@code{.altmacro}}.
5770 @section @code{.mri @var{val}}
5772 @cindex @code{mri} directive
5773 @cindex MRI mode, temporarily
5774 If @var{val} is non-zero, this tells @command{@value{AS}} to enter MRI mode. If
5775 @var{val} is zero, this tells @command{@value{AS}} to exit MRI mode. This change
5776 affects code assembled until the next @code{.mri} directive, or until the end
5777 of the file. @xref{M, MRI mode, MRI mode}.
5780 @section @code{.noaltmacro}
5781 Disable alternate macro mode. @xref{Altmacro}.
5784 @section @code{.nolist}
5786 @cindex @code{nolist} directive
5787 @cindex listing control, turning off
5788 Control (in conjunction with the @code{.list} directive) whether or
5789 not assembly listings are generated. These two directives maintain an
5790 internal counter (which is zero initially). @code{.list} increments the
5791 counter, and @code{.nolist} decrements it. Assembly listings are
5792 generated whenever the counter is greater than zero.
5795 @section @code{.octa @var{bignums}}
5797 @c FIXME: double size emitted for "octa" on i960, others? Or warn?
5798 @cindex @code{octa} directive
5799 @cindex integer, 16-byte
5800 @cindex sixteen byte integer
5801 This directive expects zero or more bignums, separated by commas. For each
5802 bignum, it emits a 16-byte integer.
5804 The term ``octa'' comes from contexts in which a ``word'' is two bytes;
5805 hence @emph{octa}-word for 16 bytes.
5808 @section @code{.offset @var{loc}}
5810 @cindex @code{offset} directive
5811 Set the location counter to @var{loc} in the absolute section. @var{loc} must
5812 be an absolute expression. This directive may be useful for defining
5813 symbols with absolute values. Do not confuse it with the @code{.org}
5817 @section @code{.org @var{new-lc} , @var{fill}}
5819 @cindex @code{org} directive
5820 @cindex location counter, advancing
5821 @cindex advancing location counter
5822 @cindex current address, advancing
5823 Advance the location counter of the current section to
5824 @var{new-lc}. @var{new-lc} is either an absolute expression or an
5825 expression with the same section as the current subsection. That is,
5826 you can't use @code{.org} to cross sections: if @var{new-lc} has the
5827 wrong section, the @code{.org} directive is ignored. To be compatible
5828 with former assemblers, if the section of @var{new-lc} is absolute,
5829 @command{@value{AS}} issues a warning, then pretends the section of @var{new-lc}
5830 is the same as the current subsection.
5832 @code{.org} may only increase the location counter, or leave it
5833 unchanged; you cannot use @code{.org} to move the location counter
5836 @c double negative used below "not undefined" because this is a specific
5837 @c reference to "undefined" (as SEG_UNKNOWN is called in this manual)
5838 @c section. doc@cygnus.com 18feb91
5839 Because @command{@value{AS}} tries to assemble programs in one pass, @var{new-lc}
5840 may not be undefined. If you really detest this restriction we eagerly await
5841 a chance to share your improved assembler.
5843 Beware that the origin is relative to the start of the section, not
5844 to the start of the subsection. This is compatible with other
5845 people's assemblers.
5847 When the location counter (of the current subsection) is advanced, the
5848 intervening bytes are filled with @var{fill} which should be an
5849 absolute expression. If the comma and @var{fill} are omitted,
5850 @var{fill} defaults to zero.
5853 @section @code{.p2align[wl] @var{abs-expr}, @var{abs-expr}, @var{abs-expr}}
5855 @cindex padding the location counter given a power of two
5856 @cindex @code{p2align} directive
5857 Pad the location counter (in the current subsection) to a particular
5858 storage boundary. The first expression (which must be absolute) is the
5859 number of low-order zero bits the location counter must have after
5860 advancement. For example @samp{.p2align 3} advances the location
5861 counter until it a multiple of 8. If the location counter is already a
5862 multiple of 8, no change is needed.
5864 The second expression (also absolute) gives the fill value to be stored in the
5865 padding bytes. It (and the comma) may be omitted. If it is omitted, the
5866 padding bytes are normally zero. However, on some systems, if the section is
5867 marked as containing code and the fill value is omitted, the space is filled
5868 with no-op instructions.
5870 The third expression is also absolute, and is also optional. If it is present,
5871 it is the maximum number of bytes that should be skipped by this alignment
5872 directive. If doing the alignment would require skipping more bytes than the
5873 specified maximum, then the alignment is not done at all. You can omit the
5874 fill value (the second argument) entirely by simply using two commas after the
5875 required alignment; this can be useful if you want the alignment to be filled
5876 with no-op instructions when appropriate.
5878 @cindex @code{p2alignw} directive
5879 @cindex @code{p2alignl} directive
5880 The @code{.p2alignw} and @code{.p2alignl} directives are variants of the
5881 @code{.p2align} directive. The @code{.p2alignw} directive treats the fill
5882 pattern as a two byte word value. The @code{.p2alignl} directives treats the
5883 fill pattern as a four byte longword value. For example, @code{.p2alignw
5884 2,0x368d} will align to a multiple of 4. If it skips two bytes, they will be
5885 filled in with the value 0x368d (the exact placement of the bytes depends upon
5886 the endianness of the processor). If it skips 1 or 3 bytes, the fill value is
5891 @section @code{.popsection}
5893 @cindex @code{popsection} directive
5894 @cindex Section Stack
5895 This is one of the ELF section stack manipulation directives. The others are
5896 @code{.section} (@pxref{Section}), @code{.subsection} (@pxref{SubSection}),
5897 @code{.pushsection} (@pxref{PushSection}), and @code{.previous}
5900 This directive replaces the current section (and subsection) with the top
5901 section (and subsection) on the section stack. This section is popped off the
5907 @section @code{.previous}
5909 @cindex @code{previous} directive
5910 @cindex Section Stack
5911 This is one of the ELF section stack manipulation directives. The others are
5912 @code{.section} (@pxref{Section}), @code{.subsection} (@pxref{SubSection}),
5913 @code{.pushsection} (@pxref{PushSection}), and @code{.popsection}
5914 (@pxref{PopSection}).
5916 This directive swaps the current section (and subsection) with most recently
5917 referenced section/subsection pair prior to this one. Multiple
5918 @code{.previous} directives in a row will flip between two sections (and their
5919 subsections). For example:
5931 Will place 0x1234 and 0x9abc into subsection 1 and 0x5678 into subsection 2 of
5937 # Now in section A subsection 1
5941 # Now in section B subsection 0
5944 # Now in section B subsection 1
5947 # Now in section B subsection 0
5951 Will place 0x1234 into section A, 0x5678 and 0xdef0 into subsection 0 of
5952 section B and 0x9abc into subsection 1 of section B.
5954 In terms of the section stack, this directive swaps the current section with
5955 the top section on the section stack.
5959 @section @code{.print @var{string}}
5961 @cindex @code{print} directive
5962 @command{@value{AS}} will print @var{string} on the standard output during
5963 assembly. You must put @var{string} in double quotes.
5967 @section @code{.protected @var{names}}
5969 @cindex @code{protected} directive
5971 This is one of the ELF visibility directives. The other two are
5972 @code{.hidden} (@pxref{Hidden}) and @code{.internal} (@pxref{Internal}).
5974 This directive overrides the named symbols default visibility (which is set by
5975 their binding: local, global or weak). The directive sets the visibility to
5976 @code{protected} which means that any references to the symbols from within the
5977 components that defines them must be resolved to the definition in that
5978 component, even if a definition in another component would normally preempt
5983 @section @code{.psize @var{lines} , @var{columns}}
5985 @cindex @code{psize} directive
5986 @cindex listing control: paper size
5987 @cindex paper size, for listings
5988 Use this directive to declare the number of lines---and, optionally, the
5989 number of columns---to use for each page, when generating listings.
5991 If you do not use @code{.psize}, listings use a default line-count
5992 of 60. You may omit the comma and @var{columns} specification; the
5993 default width is 200 columns.
5995 @command{@value{AS}} generates formfeeds whenever the specified number of
5996 lines is exceeded (or whenever you explicitly request one, using
5999 If you specify @var{lines} as @code{0}, no formfeeds are generated save
6000 those explicitly specified with @code{.eject}.
6003 @section @code{.purgem @var{name}}
6005 @cindex @code{purgem} directive
6006 Undefine the macro @var{name}, so that later uses of the string will not be
6007 expanded. @xref{Macro}.
6011 @section @code{.pushsection @var{name} [, @var{subsection}] [, "@var{flags}"[, @@@var{type}[,@var{arguments}]]]}
6013 @cindex @code{pushsection} directive
6014 @cindex Section Stack
6015 This is one of the ELF section stack manipulation directives. The others are
6016 @code{.section} (@pxref{Section}), @code{.subsection} (@pxref{SubSection}),
6017 @code{.popsection} (@pxref{PopSection}), and @code{.previous}
6020 This directive pushes the current section (and subsection) onto the
6021 top of the section stack, and then replaces the current section and
6022 subsection with @code{name} and @code{subsection}. The optional
6023 @code{flags}, @code{type} and @code{arguments} are treated the same
6024 as in the @code{.section} (@pxref{Section}) directive.
6028 @section @code{.quad @var{bignums}}
6030 @cindex @code{quad} directive
6031 @code{.quad} expects zero or more bignums, separated by commas. For
6032 each bignum, it emits
6034 an 8-byte integer. If the bignum won't fit in 8 bytes, it prints a
6035 warning message; and just takes the lowest order 8 bytes of the bignum.
6036 @cindex eight-byte integer
6037 @cindex integer, 8-byte
6039 The term ``quad'' comes from contexts in which a ``word'' is two bytes;
6040 hence @emph{quad}-word for 8 bytes.
6043 a 16-byte integer. If the bignum won't fit in 16 bytes, it prints a
6044 warning message; and just takes the lowest order 16 bytes of the bignum.
6045 @cindex sixteen-byte integer
6046 @cindex integer, 16-byte
6050 @section @code{.reloc @var{offset}, @var{reloc_name}[, @var{expression}]}
6052 @cindex @code{reloc} directive
6053 Generate a relocation at @var{offset} of type @var{reloc_name} with value
6054 @var{expression}. If @var{offset} is a number, the relocation is generated in
6055 the current section. If @var{offset} is an expression that resolves to a
6056 symbol plus offset, the relocation is generated in the given symbol's section.
6057 @var{expression}, if present, must resolve to a symbol plus addend or to an
6058 absolute value, but note that not all targets support an addend. e.g. ELF REL
6059 targets such as i386 store an addend in the section contents rather than in the
6060 relocation. This low level interface does not support addends stored in the
6064 @section @code{.rept @var{count}}
6066 @cindex @code{rept} directive
6067 Repeat the sequence of lines between the @code{.rept} directive and the next
6068 @code{.endr} directive @var{count} times.
6070 For example, assembling
6078 is equivalent to assembling
6087 @section @code{.sbttl "@var{subheading}"}
6089 @cindex @code{sbttl} directive
6090 @cindex subtitles for listings
6091 @cindex listing control: subtitle
6092 Use @var{subheading} as the title (third line, immediately after the
6093 title line) when generating assembly listings.
6095 This directive affects subsequent pages, as well as the current page if
6096 it appears within ten lines of the top of a page.
6100 @section @code{.scl @var{class}}
6102 @cindex @code{scl} directive
6103 @cindex symbol storage class (COFF)
6104 @cindex COFF symbol storage class
6105 Set the storage-class value for a symbol. This directive may only be
6106 used inside a @code{.def}/@code{.endef} pair. Storage class may flag
6107 whether a symbol is static or external, or it may record further
6108 symbolic debugging information.
6111 The @samp{.scl} directive is primarily associated with COFF output; when
6112 configured to generate @code{b.out} output format, @command{@value{AS}}
6113 accepts this directive but ignores it.
6119 @section @code{.section @var{name}}
6121 @cindex named section
6122 Use the @code{.section} directive to assemble the following code into a section
6125 This directive is only supported for targets that actually support arbitrarily
6126 named sections; on @code{a.out} targets, for example, it is not accepted, even
6127 with a standard @code{a.out} section name.
6131 @c only print the extra heading if both COFF and ELF are set
6132 @subheading COFF Version
6135 @cindex @code{section} directive (COFF version)
6136 For COFF targets, the @code{.section} directive is used in one of the following
6140 .section @var{name}[, "@var{flags}"]
6141 .section @var{name}[, @var{subsection}]
6144 If the optional argument is quoted, it is taken as flags to use for the
6145 section. Each flag is a single character. The following flags are recognized:
6148 bss section (uninitialized data)
6150 section is not loaded
6156 exclude section from linking
6162 shared section (meaningful for PE targets)
6164 ignored. (For compatibility with the ELF version)
6166 section is not readable (meaningful for PE targets)
6168 single-digit power-of-two section alignment (GNU extension)
6171 If no flags are specified, the default flags depend upon the section name. If
6172 the section name is not recognized, the default will be for the section to be
6173 loaded and writable. Note the @code{n} and @code{w} flags remove attributes
6174 from the section, rather than adding them, so if they are used on their own it
6175 will be as if no flags had been specified at all.
6177 If the optional argument to the @code{.section} directive is not quoted, it is
6178 taken as a subsection number (@pxref{Sub-Sections}).
6183 @c only print the extra heading if both COFF and ELF are set
6184 @subheading ELF Version
6187 @cindex Section Stack
6188 This is one of the ELF section stack manipulation directives. The others are
6189 @code{.subsection} (@pxref{SubSection}), @code{.pushsection}
6190 (@pxref{PushSection}), @code{.popsection} (@pxref{PopSection}), and
6191 @code{.previous} (@pxref{Previous}).
6193 @cindex @code{section} directive (ELF version)
6194 For ELF targets, the @code{.section} directive is used like this:
6197 .section @var{name} [, "@var{flags}"[, @@@var{type}[,@var{flag_specific_arguments}]]]
6200 The optional @var{flags} argument is a quoted string which may contain any
6201 combination of the following characters:
6204 section is allocatable
6206 section is excluded from executable and shared library.
6210 section is executable
6212 section is mergeable
6214 section contains zero terminated strings
6216 section is a member of a section group
6218 section is used for thread-local-storage
6220 section is a member of the previously-current section's group, if any
6223 The optional @var{type} argument may contain one of the following constants:
6226 section contains data
6228 section does not contain data (i.e., section only occupies space)
6230 section contains data which is used by things other than the program
6232 section contains an array of pointers to init functions
6234 section contains an array of pointers to finish functions
6235 @item @@preinit_array
6236 section contains an array of pointers to pre-init functions
6239 Many targets only support the first three section types.
6241 Note on targets where the @code{@@} character is the start of a comment (eg
6242 ARM) then another character is used instead. For example the ARM port uses the
6245 If @var{flags} contains the @code{M} symbol then the @var{type} argument must
6246 be specified as well as an extra argument---@var{entsize}---like this:
6249 .section @var{name} , "@var{flags}"M, @@@var{type}, @var{entsize}
6252 Sections with the @code{M} flag but not @code{S} flag must contain fixed size
6253 constants, each @var{entsize} octets long. Sections with both @code{M} and
6254 @code{S} must contain zero terminated strings where each character is
6255 @var{entsize} bytes long. The linker may remove duplicates within sections with
6256 the same name, same entity size and same flags. @var{entsize} must be an
6257 absolute expression. For sections with both @code{M} and @code{S}, a string
6258 which is a suffix of a larger string is considered a duplicate. Thus
6259 @code{"def"} will be merged with @code{"abcdef"}; A reference to the first
6260 @code{"def"} will be changed to a reference to @code{"abcdef"+3}.
6262 If @var{flags} contains the @code{G} symbol then the @var{type} argument must
6263 be present along with an additional field like this:
6266 .section @var{name} , "@var{flags}"G, @@@var{type}, @var{GroupName}[, @var{linkage}]
6269 The @var{GroupName} field specifies the name of the section group to which this
6270 particular section belongs. The optional linkage field can contain:
6273 indicates that only one copy of this section should be retained
6278 Note: if both the @var{M} and @var{G} flags are present then the fields for
6279 the Merge flag should come first, like this:
6282 .section @var{name} , "@var{flags}"MG, @@@var{type}, @var{entsize}, @var{GroupName}[, @var{linkage}]
6285 If @var{flags} contains the @code{?} symbol then it may not also contain the
6286 @code{G} symbol and the @var{GroupName} or @var{linkage} fields should not be
6287 present. Instead, @code{?} says to consider the section that's current before
6288 this directive. If that section used @code{G}, then the new section will use
6289 @code{G} with those same @var{GroupName} and @var{linkage} fields implicitly.
6290 If not, then the @code{?} symbol has no effect.
6292 If no flags are specified, the default flags depend upon the section name. If
6293 the section name is not recognized, the default will be for the section to have
6294 none of the above flags: it will not be allocated in memory, nor writable, nor
6295 executable. The section will contain data.
6297 For ELF targets, the assembler supports another type of @code{.section}
6298 directive for compatibility with the Solaris assembler:
6301 .section "@var{name}"[, @var{flags}...]
6304 Note that the section name is quoted. There may be a sequence of comma
6308 section is allocatable
6312 section is executable
6314 section is excluded from executable and shared library.
6316 section is used for thread local storage
6319 This directive replaces the current section and subsection. See the
6320 contents of the gas testsuite directory @code{gas/testsuite/gas/elf} for
6321 some examples of how this directive and the other section stack directives
6327 @section @code{.set @var{symbol}, @var{expression}}
6329 @cindex @code{set} directive
6330 @cindex symbol value, setting
6331 Set the value of @var{symbol} to @var{expression}. This
6332 changes @var{symbol}'s value and type to conform to
6333 @var{expression}. If @var{symbol} was flagged as external, it remains
6334 flagged (@pxref{Symbol Attributes}).
6336 You may @code{.set} a symbol many times in the same assembly.
6338 If you @code{.set} a global symbol, the value stored in the object
6339 file is the last value stored into it.
6342 On Z80 @code{set} is a real instruction, use
6343 @samp{@var{symbol} defl @var{expression}} instead.
6347 @section @code{.short @var{expressions}}
6349 @cindex @code{short} directive
6351 @code{.short} is normally the same as @samp{.word}.
6352 @xref{Word,,@code{.word}}.
6354 In some configurations, however, @code{.short} and @code{.word} generate
6355 numbers of different lengths. @xref{Machine Dependencies}.
6359 @code{.short} is the same as @samp{.word}. @xref{Word,,@code{.word}}.
6362 This expects zero or more @var{expressions}, and emits
6363 a 16 bit number for each.
6368 @section @code{.single @var{flonums}}
6370 @cindex @code{single} directive
6371 @cindex floating point numbers (single)
6372 This directive assembles zero or more flonums, separated by commas. It
6373 has the same effect as @code{.float}.
6375 The exact kind of floating point numbers emitted depends on how
6376 @command{@value{AS}} is configured. @xref{Machine Dependencies}.
6380 On the @value{TARGET} family, @code{.single} emits 32-bit floating point
6381 numbers in @sc{ieee} format.
6387 @section @code{.size}
6389 This directive is used to set the size associated with a symbol.
6393 @c only print the extra heading if both COFF and ELF are set
6394 @subheading COFF Version
6397 @cindex @code{size} directive (COFF version)
6398 For COFF targets, the @code{.size} directive is only permitted inside
6399 @code{.def}/@code{.endef} pairs. It is used like this:
6402 .size @var{expression}
6406 @samp{.size} is only meaningful when generating COFF format output; when
6407 @command{@value{AS}} is generating @code{b.out}, it accepts this directive but
6414 @c only print the extra heading if both COFF and ELF are set
6415 @subheading ELF Version
6418 @cindex @code{size} directive (ELF version)
6419 For ELF targets, the @code{.size} directive is used like this:
6422 .size @var{name} , @var{expression}
6425 This directive sets the size associated with a symbol @var{name}.
6426 The size in bytes is computed from @var{expression} which can make use of label
6427 arithmetic. This directive is typically used to set the size of function
6432 @ifclear no-space-dir
6434 @section @code{.skip @var{size} , @var{fill}}
6436 @cindex @code{skip} directive
6437 @cindex filling memory
6438 This directive emits @var{size} bytes, each of value @var{fill}. Both
6439 @var{size} and @var{fill} are absolute expressions. If the comma and
6440 @var{fill} are omitted, @var{fill} is assumed to be zero. This is the same as
6445 @section @code{.sleb128 @var{expressions}}
6447 @cindex @code{sleb128} directive
6448 @var{sleb128} stands for ``signed little endian base 128.'' This is a
6449 compact, variable length representation of numbers used by the DWARF
6450 symbolic debugging format. @xref{Uleb128, ,@code{.uleb128}}.
6452 @ifclear no-space-dir
6454 @section @code{.space @var{size} , @var{fill}}
6456 @cindex @code{space} directive
6457 @cindex filling memory
6458 This directive emits @var{size} bytes, each of value @var{fill}. Both
6459 @var{size} and @var{fill} are absolute expressions. If the comma
6460 and @var{fill} are omitted, @var{fill} is assumed to be zero. This is the same
6465 @emph{Warning:} @code{.space} has a completely different meaning for HPPA
6466 targets; use @code{.block} as a substitute. See @cite{HP9000 Series 800
6467 Assembly Language Reference Manual} (HP 92432-90001) for the meaning of the
6468 @code{.space} directive. @xref{HPPA Directives,,HPPA Assembler Directives},
6476 @section @code{.stabd, .stabn, .stabs}
6478 @cindex symbolic debuggers, information for
6479 @cindex @code{stab@var{x}} directives
6480 There are three directives that begin @samp{.stab}.
6481 All emit symbols (@pxref{Symbols}), for use by symbolic debuggers.
6482 The symbols are not entered in the @command{@value{AS}} hash table: they
6483 cannot be referenced elsewhere in the source file.
6484 Up to five fields are required:
6488 This is the symbol's name. It may contain any character except
6489 @samp{\000}, so is more general than ordinary symbol names. Some
6490 debuggers used to code arbitrarily complex structures into symbol names
6494 An absolute expression. The symbol's type is set to the low 8 bits of
6495 this expression. Any bit pattern is permitted, but @code{@value{LD}}
6496 and debuggers choke on silly bit patterns.
6499 An absolute expression. The symbol's ``other'' attribute is set to the
6500 low 8 bits of this expression.
6503 An absolute expression. The symbol's descriptor is set to the low 16
6504 bits of this expression.
6507 An absolute expression which becomes the symbol's value.
6510 If a warning is detected while reading a @code{.stabd}, @code{.stabn},
6511 or @code{.stabs} statement, the symbol has probably already been created;
6512 you get a half-formed symbol in your object file. This is
6513 compatible with earlier assemblers!
6516 @cindex @code{stabd} directive
6517 @item .stabd @var{type} , @var{other} , @var{desc}
6519 The ``name'' of the symbol generated is not even an empty string.
6520 It is a null pointer, for compatibility. Older assemblers used a
6521 null pointer so they didn't waste space in object files with empty
6524 The symbol's value is set to the location counter,
6525 relocatably. When your program is linked, the value of this symbol
6526 is the address of the location counter when the @code{.stabd} was
6529 @cindex @code{stabn} directive
6530 @item .stabn @var{type} , @var{other} , @var{desc} , @var{value}
6531 The name of the symbol is set to the empty string @code{""}.
6533 @cindex @code{stabs} directive
6534 @item .stabs @var{string} , @var{type} , @var{other} , @var{desc} , @var{value}
6535 All five fields are specified.
6541 @section @code{.string} "@var{str}", @code{.string8} "@var{str}", @code{.string16}
6542 "@var{str}", @code{.string32} "@var{str}", @code{.string64} "@var{str}"
6544 @cindex string, copying to object file
6545 @cindex string8, copying to object file
6546 @cindex string16, copying to object file
6547 @cindex string32, copying to object file
6548 @cindex string64, copying to object file
6549 @cindex @code{string} directive
6550 @cindex @code{string8} directive
6551 @cindex @code{string16} directive
6552 @cindex @code{string32} directive
6553 @cindex @code{string64} directive
6555 Copy the characters in @var{str} to the object file. You may specify more than
6556 one string to copy, separated by commas. Unless otherwise specified for a
6557 particular machine, the assembler marks the end of each string with a 0 byte.
6558 You can use any of the escape sequences described in @ref{Strings,,Strings}.
6560 The variants @code{string16}, @code{string32} and @code{string64} differ from
6561 the @code{string} pseudo opcode in that each 8-bit character from @var{str} is
6562 copied and expanded to 16, 32 or 64 bits respectively. The expanded characters
6563 are stored in target endianness byte order.
6569 .string "B\0\0\0Y\0\0\0E\0\0\0" /* On little endian targets. */
6570 .string "\0\0\0B\0\0\0Y\0\0\0E" /* On big endian targets. */
6575 @section @code{.struct @var{expression}}
6577 @cindex @code{struct} directive
6578 Switch to the absolute section, and set the section offset to @var{expression},
6579 which must be an absolute expression. You might use this as follows:
6588 This would define the symbol @code{field1} to have the value 0, the symbol
6589 @code{field2} to have the value 4, and the symbol @code{field3} to have the
6590 value 8. Assembly would be left in the absolute section, and you would need to
6591 use a @code{.section} directive of some sort to change to some other section
6592 before further assembly.
6596 @section @code{.subsection @var{name}}
6598 @cindex @code{subsection} directive
6599 @cindex Section Stack
6600 This is one of the ELF section stack manipulation directives. The others are
6601 @code{.section} (@pxref{Section}), @code{.pushsection} (@pxref{PushSection}),
6602 @code{.popsection} (@pxref{PopSection}), and @code{.previous}
6605 This directive replaces the current subsection with @code{name}. The current
6606 section is not changed. The replaced subsection is put onto the section stack
6607 in place of the then current top of stack subsection.
6612 @section @code{.symver}
6613 @cindex @code{symver} directive
6614 @cindex symbol versioning
6615 @cindex versions of symbols
6616 Use the @code{.symver} directive to bind symbols to specific version nodes
6617 within a source file. This is only supported on ELF platforms, and is
6618 typically used when assembling files to be linked into a shared library.
6619 There are cases where it may make sense to use this in objects to be bound
6620 into an application itself so as to override a versioned symbol from a
6623 For ELF targets, the @code{.symver} directive can be used like this:
6625 .symver @var{name}, @var{name2@@nodename}
6627 If the symbol @var{name} is defined within the file
6628 being assembled, the @code{.symver} directive effectively creates a symbol
6629 alias with the name @var{name2@@nodename}, and in fact the main reason that we
6630 just don't try and create a regular alias is that the @var{@@} character isn't
6631 permitted in symbol names. The @var{name2} part of the name is the actual name
6632 of the symbol by which it will be externally referenced. The name @var{name}
6633 itself is merely a name of convenience that is used so that it is possible to
6634 have definitions for multiple versions of a function within a single source
6635 file, and so that the compiler can unambiguously know which version of a
6636 function is being mentioned. The @var{nodename} portion of the alias should be
6637 the name of a node specified in the version script supplied to the linker when
6638 building a shared library. If you are attempting to override a versioned
6639 symbol from a shared library, then @var{nodename} should correspond to the
6640 nodename of the symbol you are trying to override.
6642 If the symbol @var{name} is not defined within the file being assembled, all
6643 references to @var{name} will be changed to @var{name2@@nodename}. If no
6644 reference to @var{name} is made, @var{name2@@nodename} will be removed from the
6647 Another usage of the @code{.symver} directive is:
6649 .symver @var{name}, @var{name2@@@@nodename}
6651 In this case, the symbol @var{name} must exist and be defined within
6652 the file being assembled. It is similar to @var{name2@@nodename}. The
6653 difference is @var{name2@@@@nodename} will also be used to resolve
6654 references to @var{name2} by the linker.
6656 The third usage of the @code{.symver} directive is:
6658 .symver @var{name}, @var{name2@@@@@@nodename}
6660 When @var{name} is not defined within the
6661 file being assembled, it is treated as @var{name2@@nodename}. When
6662 @var{name} is defined within the file being assembled, the symbol
6663 name, @var{name}, will be changed to @var{name2@@@@nodename}.
6668 @section @code{.tag @var{structname}}
6670 @cindex COFF structure debugging
6671 @cindex structure debugging, COFF
6672 @cindex @code{tag} directive
6673 This directive is generated by compilers to include auxiliary debugging
6674 information in the symbol table. It is only permitted inside
6675 @code{.def}/@code{.endef} pairs. Tags are used to link structure
6676 definitions in the symbol table with instances of those structures.
6679 @samp{.tag} is only used when generating COFF format output; when
6680 @command{@value{AS}} is generating @code{b.out}, it accepts this directive but
6686 @section @code{.text @var{subsection}}
6688 @cindex @code{text} directive
6689 Tells @command{@value{AS}} to assemble the following statements onto the end of
6690 the text subsection numbered @var{subsection}, which is an absolute
6691 expression. If @var{subsection} is omitted, subsection number zero
6695 @section @code{.title "@var{heading}"}
6697 @cindex @code{title} directive
6698 @cindex listing control: title line
6699 Use @var{heading} as the title (second line, immediately after the
6700 source file name and pagenumber) when generating assembly listings.
6702 This directive affects subsequent pages, as well as the current page if
6703 it appears within ten lines of the top of a page.
6707 @section @code{.type}
6709 This directive is used to set the type of a symbol.
6713 @c only print the extra heading if both COFF and ELF are set
6714 @subheading COFF Version
6717 @cindex COFF symbol type
6718 @cindex symbol type, COFF
6719 @cindex @code{type} directive (COFF version)
6720 For COFF targets, this directive is permitted only within
6721 @code{.def}/@code{.endef} pairs. It is used like this:
6727 This records the integer @var{int} as the type attribute of a symbol table
6731 @samp{.type} is associated only with COFF format output; when
6732 @command{@value{AS}} is configured for @code{b.out} output, it accepts this
6733 directive but ignores it.
6739 @c only print the extra heading if both COFF and ELF are set
6740 @subheading ELF Version
6743 @cindex ELF symbol type
6744 @cindex symbol type, ELF
6745 @cindex @code{type} directive (ELF version)
6746 For ELF targets, the @code{.type} directive is used like this:
6749 .type @var{name} , @var{type description}
6752 This sets the type of symbol @var{name} to be either a
6753 function symbol or an object symbol. There are five different syntaxes
6754 supported for the @var{type description} field, in order to provide
6755 compatibility with various other assemblers.
6757 Because some of the characters used in these syntaxes (such as @samp{@@} and
6758 @samp{#}) are comment characters for some architectures, some of the syntaxes
6759 below do not work on all architectures. The first variant will be accepted by
6760 the GNU assembler on all architectures so that variant should be used for
6761 maximum portability, if you do not need to assemble your code with other
6764 The syntaxes supported are:
6767 .type <name> STT_<TYPE_IN_UPPER_CASE>
6768 .type <name>,#<type>
6769 .type <name>,@@<type>
6770 .type <name>,%<type>
6771 .type <name>,"<type>"
6774 The types supported are:
6779 Mark the symbol as being a function name.
6782 @itemx gnu_indirect_function
6783 Mark the symbol as an indirect function when evaluated during reloc
6784 processing. (This is only supported on assemblers targeting GNU systems).
6788 Mark the symbol as being a data object.
6792 Mark the symbol as being a thead-local data object.
6796 Mark the symbol as being a common data object.
6800 Does not mark the symbol in any way. It is supported just for completeness.
6802 @item gnu_unique_object
6803 Marks the symbol as being a globally unique data object. The dynamic linker
6804 will make sure that in the entire process there is just one symbol with this
6805 name and type in use. (This is only supported on assemblers targeting GNU
6810 Note: Some targets support extra types in addition to those listed above.
6816 @section @code{.uleb128 @var{expressions}}
6818 @cindex @code{uleb128} directive
6819 @var{uleb128} stands for ``unsigned little endian base 128.'' This is a
6820 compact, variable length representation of numbers used by the DWARF
6821 symbolic debugging format. @xref{Sleb128, ,@code{.sleb128}}.
6825 @section @code{.val @var{addr}}
6827 @cindex @code{val} directive
6828 @cindex COFF value attribute
6829 @cindex value attribute, COFF
6830 This directive, permitted only within @code{.def}/@code{.endef} pairs,
6831 records the address @var{addr} as the value attribute of a symbol table
6835 @samp{.val} is used only for COFF output; when @command{@value{AS}} is
6836 configured for @code{b.out}, it accepts this directive but ignores it.
6842 @section @code{.version "@var{string}"}
6844 @cindex @code{version} directive
6845 This directive creates a @code{.note} section and places into it an ELF
6846 formatted note of type NT_VERSION. The note's name is set to @code{string}.
6851 @section @code{.vtable_entry @var{table}, @var{offset}}
6853 @cindex @code{vtable_entry} directive
6854 This directive finds or creates a symbol @code{table} and creates a
6855 @code{VTABLE_ENTRY} relocation for it with an addend of @code{offset}.
6858 @section @code{.vtable_inherit @var{child}, @var{parent}}
6860 @cindex @code{vtable_inherit} directive
6861 This directive finds the symbol @code{child} and finds or creates the symbol
6862 @code{parent} and then creates a @code{VTABLE_INHERIT} relocation for the
6863 parent whose addend is the value of the child symbol. As a special case the
6864 parent name of @code{0} is treated as referring to the @code{*ABS*} section.
6868 @section @code{.warning "@var{string}"}
6869 @cindex warning directive
6870 Similar to the directive @code{.error}
6871 (@pxref{Error,,@code{.error "@var{string}"}}), but just emits a warning.
6874 @section @code{.weak @var{names}}
6876 @cindex @code{weak} directive
6877 This directive sets the weak attribute on the comma separated list of symbol
6878 @code{names}. If the symbols do not already exist, they will be created.
6880 On COFF targets other than PE, weak symbols are a GNU extension. This
6881 directive sets the weak attribute on the comma separated list of symbol
6882 @code{names}. If the symbols do not already exist, they will be created.
6884 On the PE target, weak symbols are supported natively as weak aliases.
6885 When a weak symbol is created that is not an alias, GAS creates an
6886 alternate symbol to hold the default value.
6889 @section @code{.weakref @var{alias}, @var{target}}
6891 @cindex @code{weakref} directive
6892 This directive creates an alias to the target symbol that enables the symbol to
6893 be referenced with weak-symbol semantics, but without actually making it weak.
6894 If direct references or definitions of the symbol are present, then the symbol
6895 will not be weak, but if all references to it are through weak references, the
6896 symbol will be marked as weak in the symbol table.
6898 The effect is equivalent to moving all references to the alias to a separate
6899 assembly source file, renaming the alias to the symbol in it, declaring the
6900 symbol as weak there, and running a reloadable link to merge the object files
6901 resulting from the assembly of the new source file and the old source file that
6902 had the references to the alias removed.
6904 The alias itself never makes to the symbol table, and is entirely handled
6905 within the assembler.
6908 @section @code{.word @var{expressions}}
6910 @cindex @code{word} directive
6911 This directive expects zero or more @var{expressions}, of any section,
6912 separated by commas.
6915 For each expression, @command{@value{AS}} emits a 32-bit number.
6918 For each expression, @command{@value{AS}} emits a 16-bit number.
6923 The size of the number emitted, and its byte order,
6924 depend on what target computer the assembly is for.
6927 @c on amd29k, i960, sparc the "special treatment to support compilers" doesn't
6928 @c happen---32-bit addressability, period; no long/short jumps.
6929 @ifset DIFF-TBL-KLUGE
6930 @cindex difference tables altered
6931 @cindex altered difference tables
6933 @emph{Warning: Special Treatment to support Compilers}
6937 Machines with a 32-bit address space, but that do less than 32-bit
6938 addressing, require the following special treatment. If the machine of
6939 interest to you does 32-bit addressing (or doesn't require it;
6940 @pxref{Machine Dependencies}), you can ignore this issue.
6943 In order to assemble compiler output into something that works,
6944 @command{@value{AS}} occasionally does strange things to @samp{.word} directives.
6945 Directives of the form @samp{.word sym1-sym2} are often emitted by
6946 compilers as part of jump tables. Therefore, when @command{@value{AS}} assembles a
6947 directive of the form @samp{.word sym1-sym2}, and the difference between
6948 @code{sym1} and @code{sym2} does not fit in 16 bits, @command{@value{AS}}
6949 creates a @dfn{secondary jump table}, immediately before the next label.
6950 This secondary jump table is preceded by a short-jump to the
6951 first byte after the secondary table. This short-jump prevents the flow
6952 of control from accidentally falling into the new table. Inside the
6953 table is a long-jump to @code{sym2}. The original @samp{.word}
6954 contains @code{sym1} minus the address of the long-jump to
6957 If there were several occurrences of @samp{.word sym1-sym2} before the
6958 secondary jump table, all of them are adjusted. If there was a
6959 @samp{.word sym3-sym4}, that also did not fit in sixteen bits, a
6960 long-jump to @code{sym4} is included in the secondary jump table,
6961 and the @code{.word} directives are adjusted to contain @code{sym3}
6962 minus the address of the long-jump to @code{sym4}; and so on, for as many
6963 entries in the original jump table as necessary.
6966 @emph{This feature may be disabled by compiling @command{@value{AS}} with the
6967 @samp{-DWORKING_DOT_WORD} option.} This feature is likely to confuse
6968 assembly language programmers.
6971 @c end DIFF-TBL-KLUGE
6974 @section Deprecated Directives
6976 @cindex deprecated directives
6977 @cindex obsolescent directives
6978 One day these directives won't work.
6979 They are included for compatibility with older assemblers.
6986 @node Object Attributes
6987 @chapter Object Attributes
6988 @cindex object attributes
6990 @command{@value{AS}} assembles source files written for a specific architecture
6991 into object files for that architecture. But not all object files are alike.
6992 Many architectures support incompatible variations. For instance, floating
6993 point arguments might be passed in floating point registers if the object file
6994 requires hardware floating point support---or floating point arguments might be
6995 passed in integer registers if the object file supports processors with no
6996 hardware floating point unit. Or, if two objects are built for different
6997 generations of the same architecture, the combination may require the
6998 newer generation at run-time.
7000 This information is useful during and after linking. At link time,
7001 @command{@value{LD}} can warn about incompatible object files. After link
7002 time, tools like @command{gdb} can use it to process the linked file
7005 Compatibility information is recorded as a series of object attributes. Each
7006 attribute has a @dfn{vendor}, @dfn{tag}, and @dfn{value}. The vendor is a
7007 string, and indicates who sets the meaning of the tag. The tag is an integer,
7008 and indicates what property the attribute describes. The value may be a string
7009 or an integer, and indicates how the property affects this object. Missing
7010 attributes are the same as attributes with a zero value or empty string value.
7012 Object attributes were developed as part of the ABI for the ARM Architecture.
7013 The file format is documented in @cite{ELF for the ARM Architecture}.
7016 * GNU Object Attributes:: @sc{gnu} Object Attributes
7017 * Defining New Object Attributes:: Defining New Object Attributes
7020 @node GNU Object Attributes
7021 @section @sc{gnu} Object Attributes
7023 The @code{.gnu_attribute} directive records an object attribute
7024 with vendor @samp{gnu}.
7026 Except for @samp{Tag_compatibility}, which has both an integer and a string for
7027 its value, @sc{gnu} attributes have a string value if the tag number is odd and
7028 an integer value if the tag number is even. The second bit (@code{@var{tag} &
7029 2} is set for architecture-independent attributes and clear for
7030 architecture-dependent ones.
7032 @subsection Common @sc{gnu} attributes
7034 These attributes are valid on all architectures.
7037 @item Tag_compatibility (32)
7038 The compatibility attribute takes an integer flag value and a vendor name. If
7039 the flag value is 0, the file is compatible with other toolchains. If it is 1,
7040 then the file is only compatible with the named toolchain. If it is greater
7041 than 1, the file can only be processed by other toolchains under some private
7042 arrangement indicated by the flag value and the vendor name.
7045 @subsection MIPS Attributes
7048 @item Tag_GNU_MIPS_ABI_FP (4)
7049 The floating-point ABI used by this object file. The value will be:
7053 0 for files not affected by the floating-point ABI.
7055 1 for files using the hardware floating-point ABI with a standard
7056 double-precision FPU.
7058 2 for files using the hardware floating-point ABI with a single-precision FPU.
7060 3 for files using the software floating-point ABI.
7062 4 for files using the deprecated hardware floating-point ABI which used 64-bit
7063 floating-point registers, 32-bit general-purpose registers and increased the
7064 number of callee-saved floating-point registers.
7066 5 for files using the hardware floating-point ABI with a double-precision FPU
7067 with either 32-bit or 64-bit floating-point registers and 32-bit
7068 general-purpose registers.
7070 6 for files using the hardware floating-point ABI with 64-bit floating-point
7071 registers and 32-bit general-purpose registers.
7073 7 for files using the hardware floating-point ABI with 64-bit floating-point
7074 registers, 32-bit general-purpose registers and a rule that forbids the
7075 direct use of odd-numbered single-precision floating-point registers.
7079 @subsection PowerPC Attributes
7082 @item Tag_GNU_Power_ABI_FP (4)
7083 The floating-point ABI used by this object file. The value will be:
7087 0 for files not affected by the floating-point ABI.
7089 1 for files using double-precision hardware floating-point ABI.
7091 2 for files using the software floating-point ABI.
7093 3 for files using single-precision hardware floating-point ABI.
7096 @item Tag_GNU_Power_ABI_Vector (8)
7097 The vector ABI used by this object file. The value will be:
7101 0 for files not affected by the vector ABI.
7103 1 for files using general purpose registers to pass vectors.
7105 2 for files using AltiVec registers to pass vectors.
7107 3 for files using SPE registers to pass vectors.
7111 @node Defining New Object Attributes
7112 @section Defining New Object Attributes
7114 If you want to define a new @sc{gnu} object attribute, here are the places you
7115 will need to modify. New attributes should be discussed on the @samp{binutils}
7120 This manual, which is the official register of attributes.
7122 The header for your architecture @file{include/elf}, to define the tag.
7124 The @file{bfd} support file for your architecture, to merge the attribute
7125 and issue any appropriate link warnings.
7127 Test cases in @file{ld/testsuite} for merging and link warnings.
7129 @file{binutils/readelf.c} to display your attribute.
7131 GCC, if you want the compiler to mark the attribute automatically.
7137 @node Machine Dependencies
7138 @chapter Machine Dependent Features
7140 @cindex machine dependencies
7141 The machine instruction sets are (almost by definition) different on
7142 each machine where @command{@value{AS}} runs. Floating point representations
7143 vary as well, and @command{@value{AS}} often supports a few additional
7144 directives or command-line options for compatibility with other
7145 assemblers on a particular platform. Finally, some versions of
7146 @command{@value{AS}} support special pseudo-instructions for branch
7149 This chapter discusses most of these differences, though it does not
7150 include details on any machine's instruction set. For details on that
7151 subject, see the hardware manufacturer's manual.
7155 * AArch64-Dependent:: AArch64 Dependent Features
7158 * Alpha-Dependent:: Alpha Dependent Features
7161 * ARC-Dependent:: ARC Dependent Features
7164 * ARM-Dependent:: ARM Dependent Features
7167 * AVR-Dependent:: AVR Dependent Features
7170 * Blackfin-Dependent:: Blackfin Dependent Features
7173 * CR16-Dependent:: CR16 Dependent Features
7176 * CRIS-Dependent:: CRIS Dependent Features
7179 * D10V-Dependent:: D10V Dependent Features
7182 * D30V-Dependent:: D30V Dependent Features
7185 * Epiphany-Dependent:: EPIPHANY Dependent Features
7188 * H8/300-Dependent:: Renesas H8/300 Dependent Features
7191 * HPPA-Dependent:: HPPA Dependent Features
7194 * ESA/390-Dependent:: IBM ESA/390 Dependent Features
7197 * i386-Dependent:: Intel 80386 and AMD x86-64 Dependent Features
7200 * i860-Dependent:: Intel 80860 Dependent Features
7203 * i960-Dependent:: Intel 80960 Dependent Features
7206 * IA-64-Dependent:: Intel IA-64 Dependent Features
7209 * IP2K-Dependent:: IP2K Dependent Features
7212 * LM32-Dependent:: LM32 Dependent Features
7215 * M32C-Dependent:: M32C Dependent Features
7218 * M32R-Dependent:: M32R Dependent Features
7221 * M68K-Dependent:: M680x0 Dependent Features
7224 * M68HC11-Dependent:: M68HC11 and 68HC12 Dependent Features
7227 * Meta-Dependent :: Meta Dependent Features
7230 * MicroBlaze-Dependent:: MICROBLAZE Dependent Features
7233 * MIPS-Dependent:: MIPS Dependent Features
7236 * MMIX-Dependent:: MMIX Dependent Features
7239 * MSP430-Dependent:: MSP430 Dependent Features
7242 * NDS32-Dependent:: Andes NDS32 Dependent Features
7245 * NiosII-Dependent:: Altera Nios II Dependent Features
7248 * NS32K-Dependent:: NS32K Dependent Features
7251 * SH-Dependent:: Renesas / SuperH SH Dependent Features
7252 * SH64-Dependent:: SuperH SH64 Dependent Features
7255 * PDP-11-Dependent:: PDP-11 Dependent Features
7258 * PJ-Dependent:: picoJava Dependent Features
7261 * PPC-Dependent:: PowerPC Dependent Features
7264 * RL78-Dependent:: RL78 Dependent Features
7267 * RX-Dependent:: RX Dependent Features
7270 * S/390-Dependent:: IBM S/390 Dependent Features
7273 * SCORE-Dependent:: SCORE Dependent Features
7276 * Sparc-Dependent:: SPARC Dependent Features
7279 * TIC54X-Dependent:: TI TMS320C54x Dependent Features
7282 * TIC6X-Dependent :: TI TMS320C6x Dependent Features
7285 * TILE-Gx-Dependent :: Tilera TILE-Gx Dependent Features
7288 * TILEPro-Dependent :: Tilera TILEPro Dependent Features
7291 * V850-Dependent:: V850 Dependent Features
7294 * Vax-Dependent:: VAX Dependent Features
7297 * Visium-Dependent:: Visium Dependent Features
7300 * XGATE-Dependent:: XGATE Features
7303 * XSTORMY16-Dependent:: XStormy16 Dependent Features
7306 * Xtensa-Dependent:: Xtensa Dependent Features
7309 * Z80-Dependent:: Z80 Dependent Features
7312 * Z8000-Dependent:: Z8000 Dependent Features
7319 @c The following major nodes are *sections* in the GENERIC version, *chapters*
7320 @c in single-cpu versions. This is mainly achieved by @lowersections. There is a
7321 @c peculiarity: to preserve cross-references, there must be a node called
7322 @c "Machine Dependencies". Hence the conditional nodenames in each
7323 @c major node below. Node defaulting in makeinfo requires adjacency of
7324 @c node and sectioning commands; hence the repetition of @chapter BLAH
7325 @c in both conditional blocks.
7328 @include c-aarch64.texi
7332 @include c-alpha.texi
7348 @include c-bfin.texi
7352 @include c-cr16.texi
7356 @include c-cris.texi
7361 @node Machine Dependencies
7362 @chapter Machine Dependent Features
7364 The machine instruction sets are different on each Renesas chip family,
7365 and there are also some syntax differences among the families. This
7366 chapter describes the specific @command{@value{AS}} features for each
7370 * H8/300-Dependent:: Renesas H8/300 Dependent Features
7371 * SH-Dependent:: Renesas SH Dependent Features
7378 @include c-d10v.texi
7382 @include c-d30v.texi
7386 @include c-epiphany.texi
7390 @include c-h8300.texi
7394 @include c-hppa.texi
7398 @include c-i370.texi
7402 @include c-i386.texi
7406 @include c-i860.texi
7410 @include c-i960.texi
7414 @include c-ia64.texi
7418 @include c-ip2k.texi
7422 @include c-lm32.texi
7426 @include c-m32c.texi
7430 @include c-m32r.texi
7434 @include c-m68k.texi
7438 @include c-m68hc11.texi
7442 @include c-metag.texi
7446 @include c-microblaze.texi
7450 @include c-mips.texi
7454 @include c-mmix.texi
7458 @include c-msp430.texi
7462 @include c-nds32.texi
7466 @include c-nios2.texi
7470 @include c-ns32k.texi
7474 @include c-pdp11.texi
7486 @include c-rl78.texi
7494 @include c-s390.texi
7498 @include c-score.texi
7503 @include c-sh64.texi
7507 @include c-sparc.texi
7511 @include c-tic54x.texi
7515 @include c-tic6x.texi
7519 @include c-tilegx.texi
7523 @include c-tilepro.texi
7527 @include c-v850.texi
7535 @include c-visium.texi
7539 @include c-xgate.texi
7543 @include c-xstormy16.texi
7547 @include c-xtensa.texi
7559 @c reverse effect of @down at top of generic Machine-Dep chapter
7563 @node Reporting Bugs
7564 @chapter Reporting Bugs
7565 @cindex bugs in assembler
7566 @cindex reporting bugs in assembler
7568 Your bug reports play an essential role in making @command{@value{AS}} reliable.
7570 Reporting a bug may help you by bringing a solution to your problem, or it may
7571 not. But in any case the principal function of a bug report is to help the
7572 entire community by making the next version of @command{@value{AS}} work better.
7573 Bug reports are your contribution to the maintenance of @command{@value{AS}}.
7575 In order for a bug report to serve its purpose, you must include the
7576 information that enables us to fix the bug.
7579 * Bug Criteria:: Have you found a bug?
7580 * Bug Reporting:: How to report bugs
7584 @section Have You Found a Bug?
7585 @cindex bug criteria
7587 If you are not sure whether you have found a bug, here are some guidelines:
7590 @cindex fatal signal
7591 @cindex assembler crash
7592 @cindex crash of assembler
7594 If the assembler gets a fatal signal, for any input whatever, that is a
7595 @command{@value{AS}} bug. Reliable assemblers never crash.
7597 @cindex error on valid input
7599 If @command{@value{AS}} produces an error message for valid input, that is a bug.
7601 @cindex invalid input
7603 If @command{@value{AS}} does not produce an error message for invalid input, that
7604 is a bug. However, you should note that your idea of ``invalid input'' might
7605 be our idea of ``an extension'' or ``support for traditional practice''.
7608 If you are an experienced user of assemblers, your suggestions for improvement
7609 of @command{@value{AS}} are welcome in any case.
7613 @section How to Report Bugs
7615 @cindex assembler bugs, reporting
7617 A number of companies and individuals offer support for @sc{gnu} products. If
7618 you obtained @command{@value{AS}} from a support organization, we recommend you
7619 contact that organization first.
7621 You can find contact information for many support companies and
7622 individuals in the file @file{etc/SERVICE} in the @sc{gnu} Emacs
7626 In any event, we also recommend that you send bug reports for @command{@value{AS}}
7630 The fundamental principle of reporting bugs usefully is this:
7631 @strong{report all the facts}. If you are not sure whether to state a
7632 fact or leave it out, state it!
7634 Often people omit facts because they think they know what causes the problem
7635 and assume that some details do not matter. Thus, you might assume that the
7636 name of a symbol you use in an example does not matter. Well, probably it does
7637 not, but one cannot be sure. Perhaps the bug is a stray memory reference which
7638 happens to fetch from the location where that name is stored in memory;
7639 perhaps, if the name were different, the contents of that location would fool
7640 the assembler into doing the right thing despite the bug. Play it safe and
7641 give a specific, complete example. That is the easiest thing for you to do,
7642 and the most helpful.
7644 Keep in mind that the purpose of a bug report is to enable us to fix the bug if
7645 it is new to us. Therefore, always write your bug reports on the assumption
7646 that the bug has not been reported previously.
7648 Sometimes people give a few sketchy facts and ask, ``Does this ring a
7649 bell?'' This cannot help us fix a bug, so it is basically useless. We
7650 respond by asking for enough details to enable us to investigate.
7651 You might as well expedite matters by sending them to begin with.
7653 To enable us to fix the bug, you should include all these things:
7657 The version of @command{@value{AS}}. @command{@value{AS}} announces it if you start
7658 it with the @samp{--version} argument.
7660 Without this, we will not know whether there is any point in looking for
7661 the bug in the current version of @command{@value{AS}}.
7664 Any patches you may have applied to the @command{@value{AS}} source.
7667 The type of machine you are using, and the operating system name and
7671 What compiler (and its version) was used to compile @command{@value{AS}}---e.g.
7675 The command arguments you gave the assembler to assemble your example and
7676 observe the bug. To guarantee you will not omit something important, list them
7677 all. A copy of the Makefile (or the output from make) is sufficient.
7679 If we were to try to guess the arguments, we would probably guess wrong
7680 and then we might not encounter the bug.
7683 A complete input file that will reproduce the bug. If the bug is observed when
7684 the assembler is invoked via a compiler, send the assembler source, not the
7685 high level language source. Most compilers will produce the assembler source
7686 when run with the @samp{-S} option. If you are using @code{@value{GCC}}, use
7687 the options @samp{-v --save-temps}; this will save the assembler source in a
7688 file with an extension of @file{.s}, and also show you exactly how
7689 @command{@value{AS}} is being run.
7692 A description of what behavior you observe that you believe is
7693 incorrect. For example, ``It gets a fatal signal.''
7695 Of course, if the bug is that @command{@value{AS}} gets a fatal signal, then we
7696 will certainly notice it. But if the bug is incorrect output, we might not
7697 notice unless it is glaringly wrong. You might as well not give us a chance to
7700 Even if the problem you experience is a fatal signal, you should still say so
7701 explicitly. Suppose something strange is going on, such as, your copy of
7702 @command{@value{AS}} is out of sync, or you have encountered a bug in the C
7703 library on your system. (This has happened!) Your copy might crash and ours
7704 would not. If you told us to expect a crash, then when ours fails to crash, we
7705 would know that the bug was not happening for us. If you had not told us to
7706 expect a crash, then we would not be able to draw any conclusion from our
7710 If you wish to suggest changes to the @command{@value{AS}} source, send us context
7711 diffs, as generated by @code{diff} with the @samp{-u}, @samp{-c}, or @samp{-p}
7712 option. Always send diffs from the old file to the new file. If you even
7713 discuss something in the @command{@value{AS}} source, refer to it by context, not
7716 The line numbers in our development sources will not match those in your
7717 sources. Your line numbers would convey no useful information to us.
7720 Here are some things that are not necessary:
7724 A description of the envelope of the bug.
7726 Often people who encounter a bug spend a lot of time investigating
7727 which changes to the input file will make the bug go away and which
7728 changes will not affect it.
7730 This is often time consuming and not very useful, because the way we
7731 will find the bug is by running a single example under the debugger
7732 with breakpoints, not by pure deduction from a series of examples.
7733 We recommend that you save your time for something else.
7735 Of course, if you can find a simpler example to report @emph{instead}
7736 of the original one, that is a convenience for us. Errors in the
7737 output will be easier to spot, running under the debugger will take
7738 less time, and so on.
7740 However, simplification is not vital; if you do not want to do this,
7741 report the bug anyway and send us the entire test case you used.
7744 A patch for the bug.
7746 A patch for the bug does help us if it is a good one. But do not omit
7747 the necessary information, such as the test case, on the assumption that
7748 a patch is all we need. We might see problems with your patch and decide
7749 to fix the problem another way, or we might not understand it at all.
7751 Sometimes with a program as complicated as @command{@value{AS}} it is very hard to
7752 construct an example that will make the program follow a certain path through
7753 the code. If you do not send us the example, we will not be able to construct
7754 one, so we will not be able to verify that the bug is fixed.
7756 And if we cannot understand what bug you are trying to fix, or why your
7757 patch should be an improvement, we will not install it. A test case will
7758 help us to understand.
7761 A guess about what the bug is or what it depends on.
7763 Such guesses are usually wrong. Even we cannot guess right about such
7764 things without first using the debugger to find the facts.
7767 @node Acknowledgements
7768 @chapter Acknowledgements
7770 If you have contributed to GAS and your name isn't listed here,
7771 it is not meant as a slight. We just don't know about it. Send mail to the
7772 maintainer, and we'll correct the situation. Currently
7774 the maintainer is Nick Clifton (email address @code{nickc@@redhat.com}).
7776 Dean Elsner wrote the original @sc{gnu} assembler for the VAX.@footnote{Any
7779 Jay Fenlason maintained GAS for a while, adding support for GDB-specific debug
7780 information and the 68k series machines, most of the preprocessing pass, and
7781 extensive changes in @file{messages.c}, @file{input-file.c}, @file{write.c}.
7783 K. Richard Pixley maintained GAS for a while, adding various enhancements and
7784 many bug fixes, including merging support for several processors, breaking GAS
7785 up to handle multiple object file format back ends (including heavy rewrite,
7786 testing, an integration of the coff and b.out back ends), adding configuration
7787 including heavy testing and verification of cross assemblers and file splits
7788 and renaming, converted GAS to strictly ANSI C including full prototypes, added
7789 support for m680[34]0 and cpu32, did considerable work on i960 including a COFF
7790 port (including considerable amounts of reverse engineering), a SPARC opcode
7791 file rewrite, DECstation, rs6000, and hp300hpux host ports, updated ``know''
7792 assertions and made them work, much other reorganization, cleanup, and lint.
7794 Ken Raeburn wrote the high-level BFD interface code to replace most of the code
7795 in format-specific I/O modules.
7797 The original VMS support was contributed by David L. Kashtan. Eric Youngdale
7798 has done much work with it since.
7800 The Intel 80386 machine description was written by Eliot Dresselhaus.
7802 Minh Tran-Le at IntelliCorp contributed some AIX 386 support.
7804 The Motorola 88k machine description was contributed by Devon Bowen of Buffalo
7805 University and Torbjorn Granlund of the Swedish Institute of Computer Science.
7807 Keith Knowles at the Open Software Foundation wrote the original MIPS back end
7808 (@file{tc-mips.c}, @file{tc-mips.h}), and contributed Rose format support
7809 (which hasn't been merged in yet). Ralph Campbell worked with the MIPS code to
7810 support a.out format.
7812 Support for the Zilog Z8k and Renesas H8/300 processors (tc-z8k,
7813 tc-h8300), and IEEE 695 object file format (obj-ieee), was written by
7814 Steve Chamberlain of Cygnus Support. Steve also modified the COFF back end to
7815 use BFD for some low-level operations, for use with the H8/300 and AMD 29k
7818 John Gilmore built the AMD 29000 support, added @code{.include} support, and
7819 simplified the configuration of which versions accept which directives. He
7820 updated the 68k machine description so that Motorola's opcodes always produced
7821 fixed-size instructions (e.g., @code{jsr}), while synthetic instructions
7822 remained shrinkable (@code{jbsr}). John fixed many bugs, including true tested
7823 cross-compilation support, and one bug in relaxation that took a week and
7824 required the proverbial one-bit fix.
7826 Ian Lance Taylor of Cygnus Support merged the Motorola and MIT syntax for the
7827 68k, completed support for some COFF targets (68k, i386 SVR3, and SCO Unix),
7828 added support for MIPS ECOFF and ELF targets, wrote the initial RS/6000 and
7829 PowerPC assembler, and made a few other minor patches.
7831 Steve Chamberlain made GAS able to generate listings.
7833 Hewlett-Packard contributed support for the HP9000/300.
7835 Jeff Law wrote GAS and BFD support for the native HPPA object format (SOM)
7836 along with a fairly extensive HPPA testsuite (for both SOM and ELF object
7837 formats). This work was supported by both the Center for Software Science at
7838 the University of Utah and Cygnus Support.
7840 Support for ELF format files has been worked on by Mark Eichin of Cygnus
7841 Support (original, incomplete implementation for SPARC), Pete Hoogenboom and
7842 Jeff Law at the University of Utah (HPPA mainly), Michael Meissner of the Open
7843 Software Foundation (i386 mainly), and Ken Raeburn of Cygnus Support (sparc,
7844 and some initial 64-bit support).
7846 Linas Vepstas added GAS support for the ESA/390 ``IBM 370'' architecture.
7848 Richard Henderson rewrote the Alpha assembler. Klaus Kaempf wrote GAS and BFD
7849 support for openVMS/Alpha.
7851 Timothy Wall, Michael Hayes, and Greg Smart contributed to the various tic*
7854 David Heine, Sterling Augustine, Bob Wilson and John Ruttenberg from Tensilica,
7855 Inc.@: added support for Xtensa processors.
7857 Several engineers at Cygnus Support have also provided many small bug fixes and
7858 configuration enhancements.
7860 Jon Beniston added support for the Lattice Mico32 architecture.
7862 Many others have contributed large or small bugfixes and enhancements. If
7863 you have contributed significant work and are not mentioned on this list, and
7864 want to be, let us know. Some of the history has been lost; we are not
7865 intentionally leaving anyone out.
7867 @node GNU Free Documentation License
7868 @appendix GNU Free Documentation License
7872 @unnumbered AS Index