14 This document describes extensions to tools and formats LLVM seeks compatibility
17 General Assembly Syntax
18 ===========================
20 C99-style Hexadecimal Floating-point Constants
21 ----------------------------------------------
23 LLVM's assemblers allow floating-point constants to be written in C99's
24 hexadecimal format instead of decimal if desired.
31 Machine-specific Assembly Syntax
32 ================================
40 The following additional relocation types are supported:
42 **@IMGREL** (AT&T syntax only) generates an image-relative relocation that
43 corresponds to the COFF relocation types ``IMAGE_REL_I386_DIR32NB`` (32-bit) or
44 ``IMAGE_REL_AMD64_ADDR32NB`` (64-bit).
50 mov foo@IMGREL(%ebx, %ecx, 4), %eax
54 .long (fun@imgrel + 0x3F)
55 .long $unwind$fun@imgrel
57 **.secrel32** generates a relocation that corresponds to the COFF relocation
58 types ``IMAGE_REL_I386_SECREL`` (32-bit) or ``IMAGE_REL_AMD64_SECREL`` (64-bit).
60 **.secidx** relocation generates an index of the section that contains
61 the target. It corresponds to the COFF relocation types
62 ``IMAGE_REL_I386_SECTION`` (32-bit) or ``IMAGE_REL_AMD64_SECTION`` (64-bit).
66 .section .debug$S,"rn"
70 .secrel32 _function_name + 0
71 .secidx _function_name
74 ``.linkonce`` Directive
75 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
79 ``.linkonce [ comdat type ]``
81 Supported COMDAT types:
84 Discards duplicate sections with the same COMDAT symbol. This is the default
85 if no type is specified.
88 If the symbol is defined multiple times, the linker issues an error.
91 Duplicates are discarded, but the linker issues an error if any have
95 Duplicates are discarded, but the linker issues an error if any duplicates
96 do not have exactly the same content.
99 Links the largest section from among the duplicates.
102 Links the newest section from among the duplicates.
111 ``.section`` Directive
112 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
114 MC supports passing the information in ``.linkonce`` at the end of
115 ``.section``. For example, these two codes are equivalent
119 .section secName, "dr", discard, "Symbol1"
126 .section secName, "dr"
132 Note that in the combined form the COMDAT symbol is explicit. This
133 extension exists to support multiple sections with the same name in
139 .section secName, "dr", discard, "Symbol1"
144 .section secName, "dr", discard, "Symbol2"
149 In addition to the types allowed with ``.linkonce``, ``.section`` also accepts
150 ``associative``. The meaning is that the section is linked if a certain other
151 COMDAT section is linked. This other section is indicated by the comdat symbol
152 in this directive. It can be any symbol defined in the associated section, but
153 is usually the associated section's comdat.
155 The following restrictions apply to the associated section:
157 1. It must be a COMDAT section.
158 2. It cannot be another associative COMDAT section.
160 In the following example the symbol ``sym`` is the comdat symbol of ``.foo``
161 and ``.bar`` is associated to ``.foo``.
165 .section .foo,"bw",discard, "sym"
166 .section .bar,"rd",associative, "sym"
168 MC supports these flags in the COFF ``.section`` directive:
170 - ``b``: BSS section (``IMAGE_SCN_CNT_INITIALIZED_DATA``)
171 - ``d``: Data section (``IMAGE_SCN_CNT_UNINITIALIZED_DATA``)
172 - ``n``: Section is not loaded (``IMAGE_SCN_LNK_REMOVE``)
174 - ``s``: Shared section
176 - ``x``: Executable section
177 - ``y``: Not readable
178 - ``D``: Discardable (``IMAGE_SCN_MEM_DISCARDABLE``)
180 These flags are all compatible with gas, with the exception of the ``D`` flag,
181 which gnu as does not support. For gas compatibility, sections with a name
182 starting with ".debug" are implicitly discardable.
191 The following additional symbol variants are supported:
193 **:secrel_lo12:** generates a relocation that corresponds to the COFF relocation
194 types ``IMAGE_REL_ARM64_SECREL_LOW12A`` or ``IMAGE_REL_ARM64_SECREL_LOW12L``.
196 **:secrel_hi12:** generates a relocation that corresponds to the COFF relocation
197 type ``IMAGE_REL_ARM64_SECREL_HIGH12A``.
201 add x0, x0, :secrel_hi12:symbol
202 ldr x0, [x0, :secrel_lo12:symbol]
204 add x1, x1, :secrel_hi12:symbol
205 add x1, x1, :secrel_lo12:symbol
212 ``.section`` Directive
213 ^^^^^^^^^^^^^^^^^^^^^^
215 In order to support creating multiple sections with the same name and comdat,
216 it is possible to add an unique number at the end of the ``.section`` directive.
217 For example, the following code creates two sections named ``.text``.
221 .section .text,"ax",@progbits,unique,1
224 .section .text,"ax",@progbits,unique,2
228 The unique number is not present in the resulting object at all. It is just used
229 in the assembler to differentiate the sections.
231 The 'o' flag is mapped to SHF_LINK_ORDER. If it is present, a symbol
232 must be given that identifies the section to be placed is the
237 .section .foo,"a",@progbits
239 .section .bar,"ao",@progbits,.Ltmp
241 which is equivalent to just
245 .section .foo,"a",@progbits
246 .section .bar,"ao",@progbits,.foo
248 ``.linker-options`` Section (linker options)
249 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
251 In order to support passing linker options from the frontend to the linker, a
252 special section of type ``SHT_LLVM_LINKER_OPTIONS`` (usually named
253 ``.linker-options`` though the name is not significant as it is identified by
254 the type). The contents of this section is a simple pair-wise encoding of
255 directives for consideration by the linker. The strings are encoded as standard
256 null-terminated UTF-8 strings. They are emitted inline to avoid having the
257 linker traverse the object file for retrieving the value. The linker is
258 permitted to not honour the option and instead provide a warning/error to the
259 user that the requested option was not honoured.
261 The section has type ``SHT_LLVM_LINKER_OPTIONS`` and has the ``SHF_EXCLUDE``
262 flag to ensure that the section is treated as opaque by linkers which do not
263 support the feature and will not be emitted into the final linked binary.
265 This would be equivalent to the follow raw assembly:
269 .section ".linker-options","e",@llvm_linker_options
275 The following directives are specified:
279 The parameter identifies a library to be linked against. The library will
280 be looked up in the default and any specified library search paths
281 (specified to this point).
285 The parameter identifies an additional library search path to be considered
286 when looking up libraries after the inclusion of this option.
288 ``SHT_LLVM_DEPENDENT_LIBRARIES`` Section (Dependent Libraries)
289 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
291 This section contains strings specifying libraries to be added to the link by
294 The section should be consumed by the linker and not written to the output.
296 The strings are encoded as standard null-terminated UTF-8 strings.
302 .section ".deplibs","MS",@llvm_dependent_libraries,1
303 .asciz "library specifier 1"
304 .asciz "library specifier 2"
306 The interpretation of the library specifiers is defined by the consuming linker.
308 ``SHT_LLVM_CALL_GRAPH_PROFILE`` Section (Call Graph Profile)
309 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
311 This section is used to pass a call graph profile to the linker which can be
312 used to optimize the placement of sections. It contains a sequence of
313 (from symbol, to symbol, weight) tuples.
315 It shall have a type of ``SHT_LLVM_CALL_GRAPH_PROFILE`` (0x6fff4c02), shall
316 have the ``SHF_EXCLUDE`` flag set, the ``sh_link`` member shall hold the section
317 header index of the associated symbol table, and shall have a ``sh_entsize`` of
318 16. It should be named ``.llvm.call-graph-profile``.
320 The contents of the section shall be a sequence of ``Elf_CGProfile`` entries.
327 Elf_Xword cgp_weight;
331 The symbol index of the source of the edge.
334 The symbol index of the destination of the edge.
337 The weight of the edge.
339 This is represented in assembly as:
343 .cg_profile from, to, 42
345 ``.cg_profile`` directives are processed at the end of the file. It is an error
346 if either ``from`` or ``to`` are undefined temporary symbols. If either symbol
347 is a temporary symbol, then the section symbol is used instead. If either
348 symbol is undefined, then that symbol is defined as if ``.weak symbol`` has been
349 written at the end of the file. This forces the symbol to show up in the symbol
352 ``SHT_LLVM_ADDRSIG`` Section (address-significance table)
353 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
355 This section is used to mark symbols as address-significant, i.e. the address
356 of the symbol is used in a comparison or leaks outside the translation unit. It
357 has the same meaning as the absence of the LLVM attributes ``unnamed_addr``
358 and ``local_unnamed_addr``.
360 Any sections referred to by symbols that are not marked as address-significant
361 in any object file may be safely merged by a linker without breaking the
362 address uniqueness guarantee provided by the C and C++ language standards.
364 The contents of the section are a sequence of ULEB128-encoded integers
365 referring to the symbol table indexes of the address-significant symbols.
367 There are two associated assembly directives:
373 This instructs the assembler to emit an address-significance table. Without
374 this directive, all symbols are considered address-significant.
380 If ``sym`` is not otherwise referenced or defined anywhere else in the file,
381 this directive is a no-op. Otherwise, mark ``sym`` as address-significant.
383 ``SHT_LLVM_SYMPART`` Section (symbol partition specification)
384 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
386 This section is used to mark symbols with the `partition`_ that they
387 belong to. An ``.llvm_sympart`` section consists of a null-terminated string
388 specifying the name of the partition followed by a relocation referring to
389 the symbol that belongs to the partition. It may be constructed as follows:
393 .section ".llvm_sympart","",@llvm_sympart
394 .asciz "libpartition.so"
395 .word symbol_in_partition
397 .. _partition: https://lld.llvm.org/Partitions.html
399 ``SHT_LLVM_BB_ADDR_MAP`` Section (basic block address map)
400 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
401 This section stores the binary address of basic blocks along with other related
402 metadata. This information can be used to map binary profiles (like perf
403 profiles) directly to machine basic blocks.
404 This section is emitted with ``-basic-block-sections=labels`` and will contain
405 a BB address map table for every function.
407 The ``SHT_LLVM_BB_ADDR_MAP`` type provides backward compatibility to allow
408 reading older versions of the BB address map generated by older compilers. Each
409 function entry starts with a version byte which specifies the encoding version
410 to use. The following versioning schemes are currently supported.
412 Version 1 (newest): basic block address offsets are computed relative to the end
419 .section ".llvm_bb_addr_map","",@llvm_bb_addr_map
420 .byte 1 # version number
421 .byte 0 # feature byte (reserved for future use)
422 .quad .Lfunc_begin0 # address of the function
423 .byte 2 # number of basic blocks
425 .uleb128 .Lfunc_beign0-.Lfunc_begin0 # BB_0 offset relative to function entry (always zero)
426 .uleb128 .LBB_END0_0-.Lfunc_begin0 # BB_0 size
427 .byte x # BB_0 metadata
429 .uleb128 .LBB0_1-.LBB_END0_0 # BB_1 offset relative to the end of last block (BB_0).
430 .uleb128 .LBB_END0_1-.LBB0_1 # BB_1 size
431 .byte y # BB_1 metadata
433 Version 0: basic block address offsets are computed relative to the function
434 address. This uses the unversioned ``SHT_LLVM_BB_ADDR_MAP_V0`` section type and
435 is semantically equivalent to using ``SHT_LLVM_BB_ADDR_MAP`` with a zero
442 .section ".llvm_bb_addr_map","",@llvm_bb_addr_map_v0
443 .quad .Lfunc_begin0 # address of the function
444 .byte 2 # number of basic blocks
446 .uleb128 .Lfunc_beign0-.Lfunc_begin0 # BB_0 offset relative to the function entry (always zero)
447 .uleb128 .LBB_END0_0-.Lfunc_begin0 # BB_0 size
448 .byte x # BB_0 metadata
450 .uleb128 .LBB0_1-.Lfunc_begin0 # BB_1 offset relative to the function entry
451 .uleb128 .LBB_END0_1-.LBB0_1 # BB_1 size
452 .byte y # BB_1 metadata
457 PGO related analysis data can be emitted after each function within the
458 ``SHT_LLVM_BB_ADDR_MAP`` through the optional ``pgo-analysis-map`` flag.
459 Supported analyses currently are Function Entry Count, Basic Block Frequencies,
460 and Branch Probabilities.
462 Each analysis is enabled or disabled via a bit in the feature byte. Currently
465 #. Function Entry Count - Number of times the function was called as taken
466 from a PGO profile. This will always be zero if PGO was not used or the
467 function was not encountered in the profile.
469 #. Basic Block Frequencies - Encoded as raw block frequency value taken from
470 MBFI analysis. This value is an integer that encodes the relative frequency
471 compared to the entry block. More information can be found in
472 'llvm/Support/BlockFrequency.h'.
474 #. Branch Probabilities - Encoded as raw numerator for branch probability
475 taken from MBPI analysis. This value is the numerator for a fixed point ratio
476 defined in 'llvm/Support/BranchProbability.h'. It indicates the probability
477 that the block is followed by a given successor block during execution.
479 This extra data requires version 2 or above. This is necessary since successors
480 of basic blocks won't know their index but will know their BB ID.
482 Example of BBAddrMap with PGO data:
486 .section ".llvm_bb_addr_map","",@llvm_bb_addr_map
487 .byte 2 # version number
488 .byte 7 # feature byte - PGO analyses enabled mask
489 .quad .Lfunc_begin0 # address of the function
490 .uleb128 4 # number of basic blocks
492 .uleb128 0 # BB_0 BB ID
493 .uleb128 .Lfunc_begin0-.Lfunc_begin0 # BB_0 offset relative to function entry (always zero)
494 .uleb128 .LBB_END0_0-.Lfunc_begin0 # BB_0 size
495 .byte 0x18 # BB_0 metadata (multiple successors)
497 .uleb128 1 # BB_1 BB ID
498 .uleb128 .LBB0_1-.LBB_END0_0 # BB_1 offset relative to the end of last block (BB_0).
499 .uleb128 .LBB_END0_1-.LBB0_1 # BB_1 size
500 .byte 0x0 # BB_1 metadata (two successors)
502 .uleb128 2 # BB_2 BB ID
503 .uleb128 .LBB0_2-.LBB_END1_0 # BB_2 offset relative to the end of last block (BB_1).
504 .uleb128 .LBB_END0_2-.LBB0_2 # BB_2 size
505 .byte 0x0 # BB_2 metadata (one successor)
507 .uleb128 3 # BB_3 BB ID
508 .uleb128 .LBB0_3-.LBB_END0_2 # BB_3 offset relative to the end of last block (BB_2).
509 .uleb128 .LBB_END0_3-.LBB0_3 # BB_3 size
510 .byte 0x0 # BB_3 metadata (zero successors)
512 .uleb128 1000 # function entry count (only when enabled)
513 # PGO data record for BB_0
514 .uleb128 1000 # BB_0 basic block frequency (only when enabled)
515 .uleb128 3 # BB_0 successors count (only enabled with branch probabilities)
516 .uleb128 1 # BB_0 successor 1 BB ID (only enabled with branch probabilities)
517 .uleb128 0x22222222 # BB_0 successor 1 branch probability (only enabled with branch probabilities)
518 .uleb128 2 # BB_0 successor 2 BB ID (only enabled with branch probabilities)
519 .uleb128 0x33333333 # BB_0 successor 2 branch probability (only enabled with branch probabilities)
520 .uleb128 3 # BB_0 successor 3 BB ID (only enabled with branch probabilities)
521 .uleb128 0xaaaaaaaa # BB_0 successor 3 branch probability (only enabled with branch probabilities)
522 # PGO data record for BB_1
523 .uleb128 133 # BB_1 basic block frequency (only when enabled)
524 .uleb128 2 # BB_1 successors count (only enabled with branch probabilities)
525 .uleb128 2 # BB_1 successor 1 BB ID (only enabled with branch probabilities)
526 .uleb128 0x11111111 # BB_1 successor 1 branch probability (only enabled with branch probabilities)
527 .uleb128 3 # BB_1 successor 2 BB ID (only enabled with branch probabilities)
528 .uleb128 0x11111111 # BB_1 successor 2 branch probability (only enabled with branch probabilities)
529 # PGO data record for BB_2
530 .uleb128 18 # BB_2 basic block frequency (only when enabled)
531 .uleb128 1 # BB_2 successors count (only enabled with branch probabilities)
532 .uleb128 3 # BB_2 successor 1 BB ID (only enabled with branch probabilities)
533 .uleb128 0xffffffff # BB_2 successor 1 branch probability (only enabled with branch probabilities)
534 # PGO data record for BB_3
535 .uleb128 1000 # BB_3 basic block frequency (only when enabled)
536 .uleb128 0 # BB_3 successors count (only enabled with branch probabilities)
538 ``SHT_LLVM_OFFLOADING`` Section (offloading data)
539 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
540 This section stores the binary data used to perform offloading device linking
541 and execution, creating a fat binary. This section is emitted during compilation
542 of offloading languages such as OpenMP or CUDA. If the data is intended to be
543 used by the device linker only, it should use the ``SHF_EXCLUDE`` flag so it is
544 automatically stripped from the final executable or shared library.
546 The binary data stored in this section conforms to a custom binary format used
547 for storing offloading metadata. This format is effectively a string table
548 containing metadata accompanied by a device image.
550 ``SHT_LLVM_LTO`` Section (LLVM bitcode for fat LTO)
551 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
552 This section stores LLVM bitcode used to perform regular LTO or ThinLTO at link
553 time. This section is generated when the compiler enables fat LTO. This section
554 has the ``SHF_EXCLUDE`` flag so that it is stripped from the final executable
560 ``.cv_file`` Directive
561 ^^^^^^^^^^^^^^^^^^^^^^
563 ``.cv_file`` *FileNumber FileName* [ *checksum* ] [ *checksumkind* ]
565 ``.cv_func_id`` Directive
566 ^^^^^^^^^^^^^^^^^^^^^^^^^
567 Introduces a function ID that can be used with ``.cv_loc``.
570 ``.cv_func_id`` *FunctionId*
572 ``.cv_inline_site_id`` Directive
573 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
574 Introduces a function ID that can be used with ``.cv_loc``. Includes
575 ``inlined at`` source location information for use in the line table of the
576 caller, whether the caller is a real function or another inlined call site.
579 ``.cv_inline_site_id`` *FunctionId* ``within`` *Function* ``inlined_at`` *FileNumber Line* [ *Column* ]
581 ``.cv_loc`` Directive
582 ^^^^^^^^^^^^^^^^^^^^^
583 The first number is a file number, must have been previously assigned with a
584 ``.file`` directive, the second number is the line number and optionally the
585 third number is a column position (zero if not specified). The remaining
586 optional items are ``.loc`` sub-directives.
589 ``.cv_loc`` *FunctionId FileNumber* [ *Line* ] [ *Column* ] [ *prologue_end* ] [ ``is_stmt`` *value* ]
591 ``.cv_linetable`` Directive
592 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
594 ``.cv_linetable`` *FunctionId* ``,`` *FunctionStart* ``,`` *FunctionEnd*
596 ``.cv_inline_linetable`` Directive
597 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
599 ``.cv_inline_linetable`` *PrimaryFunctionId* ``,`` *FileNumber Line FunctionStart FunctionEnd*
601 ``.cv_def_range`` Directive
602 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
603 The *GapStart* and *GapEnd* options may be repeated as needed.
606 ``.cv_def_range`` *RangeStart RangeEnd* [ *GapStart GapEnd* ] ``,`` *bytes*
608 ``.cv_stringtable`` Directive
609 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
611 ``.cv_filechecksums`` Directive
612 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
614 ``.cv_filechecksumoffset`` Directive
615 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
617 ``.cv_filechecksumoffset`` *FileNumber*
619 ``.cv_fpo_data`` Directive
620 ^^^^^^^^^^^^^^^^^^^^^^^^^^
622 ``.cv_fpo_data`` *procsym*
624 Target Specific Behaviour
625 =========================
633 **@ABS8** can be applied to symbols which appear as immediate operands to
634 instructions that have an 8-bit immediate form for that operand. It causes
635 the assembler to use the 8-bit form and an 8-bit relocation (e.g. ``R_386_8``
636 or ``R_X86_64_8``) for the symbol.
644 This causes the assembler to select the form of the 64-bit ``cmpq`` instruction
645 that takes an 8-bit immediate operand that is sign extended to 64 bits, as
646 opposed to ``cmpq $foo, %rdi`` which takes a 32-bit immediate operand. This
647 is also not the same as ``cmpb $foo, %dil``, which is an 8-bit comparison.
650 **@GOTPCREL_NORELAX** can be used in place of ``@GOTPCREL`` to guarantee that
651 the assembler emits an ``R_X86_64_GOTPCREL`` relocation instead of a relaxable
652 ``R_X86_64[_REX]_GOTPCRELX`` relocation.
660 The reference implementation (Microsoft Visual Studio 2012) emits stack probes
661 in the following fashion:
669 However, this has the limitation of 32 MiB (±16MiB). In order to accommodate
670 larger binaries, LLVM supports the use of ``-mcmodel=large`` to allow a 4GiB
671 range via a slight deviation. It will generate an indirect jump as follows:
676 movw r12, :lower16:__chkstk
677 movt r12, :upper16:__chkstk
681 Variable Length Arrays
682 ^^^^^^^^^^^^^^^^^^^^^^
684 The reference implementation (Microsoft Visual Studio 2012) does not permit the
685 emission of Variable Length Arrays (VLAs).
687 The Windows ARM Itanium ABI extends the base ABI by adding support for emitting
688 a dynamic stack allocation. When emitting a variable stack allocation, a call
689 to ``__chkstk`` is emitted unconditionally to ensure that guard pages are setup
690 properly. The emission of this stack probe emission is handled similar to the
691 standard stack probe emission.
693 The MSVC environment does not emit code for VLAs currently.
701 The reference implementation (Microsoft Visual Studio 2017) emits stack probes
702 in the following fashion:
708 sub sp, sp, x15, lsl #4
710 However, this has the limitation of 256 MiB (±128MiB). In order to accommodate
711 larger binaries, LLVM supports the use of ``-mcmodel=large`` to allow a 8GiB
712 (±4GiB) range via a slight deviation. It will generate an indirect jump as
719 add x16, x16, :lo12:__chkstk
721 sub sp, sp, x15, lsl #4