• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1.. role:: raw-html(raw)
2   :format: html
3
4========================
5LLVM Bitcode File Format
6========================
7
8.. contents::
9   :local:
10
11Abstract
12========
13
14This document describes the LLVM bitstream file format and the encoding of the
15LLVM IR into it.
16
17Overview
18========
19
20What is commonly known as the LLVM bitcode file format (also, sometimes
21anachronistically known as bytecode) is actually two things: a `bitstream
22container format`_ and an `encoding of LLVM IR`_ into the container format.
23
24The bitstream format is an abstract encoding of structured data, very similar to
25XML in some ways.  Like XML, bitstream files contain tags, and nested
26structures, and you can parse the file without having to understand the tags.
27Unlike XML, the bitstream format is a binary encoding, and unlike XML it
28provides a mechanism for the file to self-describe "abbreviations", which are
29effectively size optimizations for the content.
30
31LLVM IR files may be optionally embedded into a `wrapper`_ structure, or in a
32`native object file`_. Both of these mechanisms make it easy to embed extra
33data along with LLVM IR files.
34
35This document first describes the LLVM bitstream format, describes the wrapper
36format, then describes the record structure used by LLVM IR files.
37
38.. _bitstream container format:
39
40Bitstream Format
41================
42
43The bitstream format is literally a stream of bits, with a very simple
44structure.  This structure consists of the following concepts:
45
46* A "`magic number`_" that identifies the contents of the stream.
47
48* Encoding `primitives`_ like variable bit-rate integers.
49
50* `Blocks`_, which define nested content.
51
52* `Data Records`_, which describe entities within the file.
53
54* Abbreviations, which specify compression optimizations for the file.
55
56Note that the :doc:`llvm-bcanalyzer <CommandGuide/llvm-bcanalyzer>` tool can be
57used to dump and inspect arbitrary bitstreams, which is very useful for
58understanding the encoding.
59
60.. _magic number:
61
62Magic Numbers
63-------------
64
65The first two bytes of a bitcode file are 'BC' (``0x42``, ``0x43``).  The second
66two bytes are an application-specific magic number.  Generic bitcode tools can
67look at only the first two bytes to verify the file is bitcode, while
68application-specific programs will want to look at all four.
69
70.. _primitives:
71
72Primitives
73----------
74
75A bitstream literally consists of a stream of bits, which are read in order
76starting with the least significant bit of each byte.  The stream is made up of
77a number of primitive values that encode a stream of unsigned integer values.
78These integers are encoded in two ways: either as `Fixed Width Integers`_ or as
79`Variable Width Integers`_.
80
81.. _Fixed Width Integers:
82.. _fixed-width value:
83
84Fixed Width Integers
85^^^^^^^^^^^^^^^^^^^^
86
87Fixed-width integer values have their low bits emitted directly to the file.
88For example, a 3-bit integer value encodes 1 as 001.  Fixed width integers are
89used when there are a well-known number of options for a field.  For example,
90boolean values are usually encoded with a 1-bit wide integer.
91
92.. _Variable Width Integers:
93.. _Variable Width Integer:
94.. _variable-width value:
95
96Variable Width Integers
97^^^^^^^^^^^^^^^^^^^^^^^
98
99Variable-width integer (VBR) values encode values of arbitrary size, optimizing
100for the case where the values are small.  Given a 4-bit VBR field, any 3-bit
101value (0 through 7) is encoded directly, with the high bit set to zero.  Values
102larger than N-1 bits emit their bits in a series of N-1 bit chunks, where all
103but the last set the high bit.
104
105For example, the value 27 (0x1B) is encoded as 1011 0011 when emitted as a vbr4
106value.  The first set of four bits indicates the value 3 (011) with a
107continuation piece (indicated by a high bit of 1).  The next word indicates a
108value of 24 (011 << 3) with no continuation.  The sum (3+24) yields the value
10927.
110
111.. _char6-encoded value:
112
1136-bit characters
114^^^^^^^^^^^^^^^^
115
1166-bit characters encode common characters into a fixed 6-bit field.  They
117represent the following characters with the following 6-bit values:
118
119::
120
121  'a' .. 'z' ---  0 .. 25
122  'A' .. 'Z' --- 26 .. 51
123  '0' .. '9' --- 52 .. 61
124         '.' --- 62
125         '_' --- 63
126
127This encoding is only suitable for encoding characters and strings that consist
128only of the above characters.  It is completely incapable of encoding characters
129not in the set.
130
131Word Alignment
132^^^^^^^^^^^^^^
133
134Occasionally, it is useful to emit zero bits until the bitstream is a multiple
135of 32 bits.  This ensures that the bit position in the stream can be represented
136as a multiple of 32-bit words.
137
138Abbreviation IDs
139----------------
140
141A bitstream is a sequential series of `Blocks`_ and `Data Records`_.  Both of
142these start with an abbreviation ID encoded as a fixed-bitwidth field.  The
143width is specified by the current block, as described below.  The value of the
144abbreviation ID specifies either a builtin ID (which have special meanings,
145defined below) or one of the abbreviation IDs defined for the current block by
146the stream itself.
147
148The set of builtin abbrev IDs is:
149
150* 0 - `END_BLOCK`_ --- This abbrev ID marks the end of the current block.
151
152* 1 - `ENTER_SUBBLOCK`_ --- This abbrev ID marks the beginning of a new
153  block.
154
155* 2 - `DEFINE_ABBREV`_ --- This defines a new abbreviation.
156
157* 3 - `UNABBREV_RECORD`_ --- This ID specifies the definition of an
158  unabbreviated record.
159
160Abbreviation IDs 4 and above are defined by the stream itself, and specify an
161`abbreviated record encoding`_.
162
163.. _Blocks:
164
165Blocks
166------
167
168Blocks in a bitstream denote nested regions of the stream, and are identified by
169a content-specific id number (for example, LLVM IR uses an ID of 12 to represent
170function bodies).  Block IDs 0-7 are reserved for `standard blocks`_ whose
171meaning is defined by Bitcode; block IDs 8 and greater are application
172specific. Nested blocks capture the hierarchical structure of the data encoded
173in it, and various properties are associated with blocks as the file is parsed.
174Block definitions allow the reader to efficiently skip blocks in constant time
175if the reader wants a summary of blocks, or if it wants to efficiently skip data
176it does not understand.  The LLVM IR reader uses this mechanism to skip function
177bodies, lazily reading them on demand.
178
179When reading and encoding the stream, several properties are maintained for the
180block.  In particular, each block maintains:
181
182#. A current abbrev id width.  This value starts at 2 at the beginning of the
183   stream, and is set every time a block record is entered.  The block entry
184   specifies the abbrev id width for the body of the block.
185
186#. A set of abbreviations.  Abbreviations may be defined within a block, in
187   which case they are only defined in that block (neither subblocks nor
188   enclosing blocks see the abbreviation).  Abbreviations can also be defined
189   inside a `BLOCKINFO`_ block, in which case they are defined in all blocks
190   that match the ID that the ``BLOCKINFO`` block is describing.
191
192As sub blocks are entered, these properties are saved and the new sub-block has
193its own set of abbreviations, and its own abbrev id width.  When a sub-block is
194popped, the saved values are restored.
195
196.. _ENTER_SUBBLOCK:
197
198ENTER_SUBBLOCK Encoding
199^^^^^^^^^^^^^^^^^^^^^^^
200
201:raw-html:`<tt>`
202[ENTER_SUBBLOCK, blockid\ :sub:`vbr8`, newabbrevlen\ :sub:`vbr4`, <align32bits>, blocklen_32]
203:raw-html:`</tt>`
204
205The ``ENTER_SUBBLOCK`` abbreviation ID specifies the start of a new block
206record.  The ``blockid`` value is encoded as an 8-bit VBR identifier, and
207indicates the type of block being entered, which can be a `standard block`_ or
208an application-specific block.  The ``newabbrevlen`` value is a 4-bit VBR, which
209specifies the abbrev id width for the sub-block.  The ``blocklen`` value is a
21032-bit aligned value that specifies the size of the subblock in 32-bit
211words. This value allows the reader to skip over the entire block in one jump.
212
213.. _END_BLOCK:
214
215END_BLOCK Encoding
216^^^^^^^^^^^^^^^^^^
217
218``[END_BLOCK, <align32bits>]``
219
220The ``END_BLOCK`` abbreviation ID specifies the end of the current block record.
221Its end is aligned to 32-bits to ensure that the size of the block is an even
222multiple of 32-bits.
223
224.. _Data Records:
225
226Data Records
227------------
228
229Data records consist of a record code and a number of (up to) 64-bit integer
230values.  The interpretation of the code and values is application specific and
231may vary between different block types.  Records can be encoded either using an
232unabbrev record, or with an abbreviation.  In the LLVM IR format, for example,
233there is a record which encodes the target triple of a module.  The code is
234``MODULE_CODE_TRIPLE``, and the values of the record are the ASCII codes for the
235characters in the string.
236
237.. _UNABBREV_RECORD:
238
239UNABBREV_RECORD Encoding
240^^^^^^^^^^^^^^^^^^^^^^^^
241
242:raw-html:`<tt>`
243[UNABBREV_RECORD, code\ :sub:`vbr6`, numops\ :sub:`vbr6`, op0\ :sub:`vbr6`, op1\ :sub:`vbr6`, ...]
244:raw-html:`</tt>`
245
246An ``UNABBREV_RECORD`` provides a default fallback encoding, which is both
247completely general and extremely inefficient.  It can describe an arbitrary
248record by emitting the code and operands as VBRs.
249
250For example, emitting an LLVM IR target triple as an unabbreviated record
251requires emitting the ``UNABBREV_RECORD`` abbrevid, a vbr6 for the
252``MODULE_CODE_TRIPLE`` code, a vbr6 for the length of the string, which is equal
253to the number of operands, and a vbr6 for each character.  Because there are no
254letters with values less than 32, each letter would need to be emitted as at
255least a two-part VBR, which means that each letter would require at least 12
256bits.  This is not an efficient encoding, but it is fully general.
257
258.. _abbreviated record encoding:
259
260Abbreviated Record Encoding
261^^^^^^^^^^^^^^^^^^^^^^^^^^^
262
263``[<abbrevid>, fields...]``
264
265An abbreviated record is a abbreviation id followed by a set of fields that are
266encoded according to the `abbreviation definition`_.  This allows records to be
267encoded significantly more densely than records encoded with the
268`UNABBREV_RECORD`_ type, and allows the abbreviation types to be specified in
269the stream itself, which allows the files to be completely self describing.  The
270actual encoding of abbreviations is defined below.
271
272The record code, which is the first field of an abbreviated record, may be
273encoded in the abbreviation definition (as a literal operand) or supplied in the
274abbreviated record (as a Fixed or VBR operand value).
275
276.. _abbreviation definition:
277
278Abbreviations
279-------------
280
281Abbreviations are an important form of compression for bitstreams.  The idea is
282to specify a dense encoding for a class of records once, then use that encoding
283to emit many records.  It takes space to emit the encoding into the file, but
284the space is recouped (hopefully plus some) when the records that use it are
285emitted.
286
287Abbreviations can be determined dynamically per client, per file. Because the
288abbreviations are stored in the bitstream itself, different streams of the same
289format can contain different sets of abbreviations according to the needs of the
290specific stream.  As a concrete example, LLVM IR files usually emit an
291abbreviation for binary operators.  If a specific LLVM module contained no or
292few binary operators, the abbreviation does not need to be emitted.
293
294.. _DEFINE_ABBREV:
295
296DEFINE_ABBREV Encoding
297^^^^^^^^^^^^^^^^^^^^^^
298
299:raw-html:`<tt>`
300[DEFINE_ABBREV, numabbrevops\ :sub:`vbr5`, abbrevop0, abbrevop1, ...]
301:raw-html:`</tt>`
302
303A ``DEFINE_ABBREV`` record adds an abbreviation to the list of currently defined
304abbreviations in the scope of this block.  This definition only exists inside
305this immediate block --- it is not visible in subblocks or enclosing blocks.
306Abbreviations are implicitly assigned IDs sequentially starting from 4 (the
307first application-defined abbreviation ID).  Any abbreviations defined in a
308``BLOCKINFO`` record for the particular block type receive IDs first, in order,
309followed by any abbreviations defined within the block itself.  Abbreviated data
310records reference this ID to indicate what abbreviation they are invoking.
311
312An abbreviation definition consists of the ``DEFINE_ABBREV`` abbrevid followed
313by a VBR that specifies the number of abbrev operands, then the abbrev operands
314themselves.  Abbreviation operands come in three forms.  They all start with a
315single bit that indicates whether the abbrev operand is a literal operand (when
316the bit is 1) or an encoding operand (when the bit is 0).
317
318#. Literal operands --- :raw-html:`<tt>` [1\ :sub:`1`, litvalue\
319   :sub:`vbr8`] :raw-html:`</tt>` --- Literal operands specify that the value in
320   the result is always a single specific value.  This specific value is emitted
321   as a vbr8 after the bit indicating that it is a literal operand.
322
323#. Encoding info without data --- :raw-html:`<tt>` [0\ :sub:`1`, encoding\
324   :sub:`3`] :raw-html:`</tt>` --- Operand encodings that do not have extra data
325   are just emitted as their code.
326
327#. Encoding info with data --- :raw-html:`<tt>` [0\ :sub:`1`, encoding\
328   :sub:`3`, value\ :sub:`vbr5`] :raw-html:`</tt>` --- Operand encodings that do
329   have extra data are emitted as their code, followed by the extra data.
330
331The possible operand encodings are:
332
333* Fixed (code 1): The field should be emitted as a `fixed-width value`_, whose
334  width is specified by the operand's extra data.
335
336* VBR (code 2): The field should be emitted as a `variable-width value`_, whose
337  width is specified by the operand's extra data.
338
339* Array (code 3): This field is an array of values.  The array operand has no
340  extra data, but expects another operand to follow it, indicating the element
341  type of the array.  When reading an array in an abbreviated record, the first
342  integer is a vbr6 that indicates the array length, followed by the encoded
343  elements of the array.  An array may only occur as the last operand of an
344  abbreviation (except for the one final operand that gives the array's
345  type).
346
347* Char6 (code 4): This field should be emitted as a `char6-encoded value`_.
348  This operand type takes no extra data. Char6 encoding is normally used as an
349  array element type.
350
351* Blob (code 5): This field is emitted as a vbr6, followed by padding to a
352  32-bit boundary (for alignment) and an array of 8-bit objects.  The array of
353  bytes is further followed by tail padding to ensure that its total length is a
354  multiple of 4 bytes.  This makes it very efficient for the reader to decode
355  the data without having to make a copy of it: it can use a pointer to the data
356  in the mapped in file and poke directly at it.  A blob may only occur as the
357  last operand of an abbreviation.
358
359For example, target triples in LLVM modules are encoded as a record of the form
360``[TRIPLE, 'a', 'b', 'c', 'd']``.  Consider if the bitstream emitted the
361following abbrev entry:
362
363::
364
365  [0, Fixed, 4]
366  [0, Array]
367  [0, Char6]
368
369When emitting a record with this abbreviation, the above entry would be emitted
370as:
371
372:raw-html:`<tt><blockquote>`
373[4\ :sub:`abbrevwidth`, 2\ :sub:`4`, 4\ :sub:`vbr6`, 0\ :sub:`6`, 1\ :sub:`6`, 2\ :sub:`6`, 3\ :sub:`6`]
374:raw-html:`</blockquote></tt>`
375
376These values are:
377
378#. The first value, 4, is the abbreviation ID for this abbreviation.
379
380#. The second value, 2, is the record code for ``TRIPLE`` records within LLVM IR
381   file ``MODULE_BLOCK`` blocks.
382
383#. The third value, 4, is the length of the array.
384
385#. The rest of the values are the char6 encoded values for ``"abcd"``.
386
387With this abbreviation, the triple is emitted with only 37 bits (assuming a
388abbrev id width of 3).  Without the abbreviation, significantly more space would
389be required to emit the target triple.  Also, because the ``TRIPLE`` value is
390not emitted as a literal in the abbreviation, the abbreviation can also be used
391for any other string value.
392
393.. _standard blocks:
394.. _standard block:
395
396Standard Blocks
397---------------
398
399In addition to the basic block structure and record encodings, the bitstream
400also defines specific built-in block types.  These block types specify how the
401stream is to be decoded or other metadata.  In the future, new standard blocks
402may be added.  Block IDs 0-7 are reserved for standard blocks.
403
404.. _BLOCKINFO:
405
406#0 - BLOCKINFO Block
407^^^^^^^^^^^^^^^^^^^^
408
409The ``BLOCKINFO`` block allows the description of metadata for other blocks.
410The currently specified records are:
411
412::
413
414  [SETBID (#1), blockid]
415  [DEFINE_ABBREV, ...]
416  [BLOCKNAME, ...name...]
417  [SETRECORDNAME, RecordID, ...name...]
418
419The ``SETBID`` record (code 1) indicates which block ID is being described.
420``SETBID`` records can occur multiple times throughout the block to change which
421block ID is being described.  There must be a ``SETBID`` record prior to any
422other records.
423
424Standard ``DEFINE_ABBREV`` records can occur inside ``BLOCKINFO`` blocks, but
425unlike their occurrence in normal blocks, the abbreviation is defined for blocks
426matching the block ID we are describing, *not* the ``BLOCKINFO`` block
427itself.  The abbreviations defined in ``BLOCKINFO`` blocks receive abbreviation
428IDs as described in `DEFINE_ABBREV`_.
429
430The ``BLOCKNAME`` record (code 2) can optionally occur in this block.  The
431elements of the record are the bytes of the string name of the block.
432llvm-bcanalyzer can use this to dump out bitcode files symbolically.
433
434The ``SETRECORDNAME`` record (code 3) can also optionally occur in this block.
435The first operand value is a record ID number, and the rest of the elements of
436the record are the bytes for the string name of the record.  llvm-bcanalyzer can
437use this to dump out bitcode files symbolically.
438
439Note that although the data in ``BLOCKINFO`` blocks is described as "metadata,"
440the abbreviations they contain are essential for parsing records from the
441corresponding blocks.  It is not safe to skip them.
442
443.. _wrapper:
444
445Bitcode Wrapper Format
446======================
447
448Bitcode files for LLVM IR may optionally be wrapped in a simple wrapper
449structure.  This structure contains a simple header that indicates the offset
450and size of the embedded BC file.  This allows additional information to be
451stored alongside the BC file.  The structure of this file header is:
452
453:raw-html:`<tt><blockquote>`
454[Magic\ :sub:`32`, Version\ :sub:`32`, Offset\ :sub:`32`, Size\ :sub:`32`, CPUType\ :sub:`32`]
455:raw-html:`</blockquote></tt>`
456
457Each of the fields are 32-bit fields stored in little endian form (as with the
458rest of the bitcode file fields).  The Magic number is always ``0x0B17C0DE`` and
459the version is currently always ``0``.  The Offset field is the offset in bytes
460to the start of the bitcode stream in the file, and the Size field is the size
461in bytes of the stream. CPUType is a target-specific value that can be used to
462encode the CPU of the target.
463
464.. _native object file:
465
466Native Object File Wrapper Format
467=================================
468
469Bitcode files for LLVM IR may also be wrapped in a native object file
470(i.e. ELF, COFF, Mach-O).  The bitcode must be stored in a section of the object
471file named ``__LLVM,__bitcode`` for MachO and ``.llvmbc`` for the other object
472formats.  This wrapper format is useful for accommodating LTO in compilation
473pipelines where intermediate objects must be native object files which contain
474metadata in other sections.
475
476Not all tools support this format.
477
478.. _encoding of LLVM IR:
479
480LLVM IR Encoding
481================
482
483LLVM IR is encoded into a bitstream by defining blocks and records.  It uses
484blocks for things like constant pools, functions, symbol tables, etc.  It uses
485records for things like instructions, global variable descriptors, type
486descriptions, etc.  This document does not describe the set of abbreviations
487that the writer uses, as these are fully self-described in the file, and the
488reader is not allowed to build in any knowledge of this.
489
490Basics
491------
492
493LLVM IR Magic Number
494^^^^^^^^^^^^^^^^^^^^
495
496The magic number for LLVM IR files is:
497
498:raw-html:`<tt><blockquote>`
499[0x0\ :sub:`4`, 0xC\ :sub:`4`, 0xE\ :sub:`4`, 0xD\ :sub:`4`]
500:raw-html:`</blockquote></tt>`
501
502When combined with the bitcode magic number and viewed as bytes, this is
503``"BC 0xC0DE"``.
504
505.. _Signed VBRs:
506
507Signed VBRs
508^^^^^^^^^^^
509
510`Variable Width Integer`_ encoding is an efficient way to encode arbitrary sized
511unsigned values, but is an extremely inefficient for encoding signed values, as
512signed values are otherwise treated as maximally large unsigned values.
513
514As such, signed VBR values of a specific width are emitted as follows:
515
516* Positive values are emitted as VBRs of the specified width, but with their
517  value shifted left by one.
518
519* Negative values are emitted as VBRs of the specified width, but the negated
520  value is shifted left by one, and the low bit is set.
521
522With this encoding, small positive and small negative values can both be emitted
523efficiently. Signed VBR encoding is used in ``CST_CODE_INTEGER`` and
524``CST_CODE_WIDE_INTEGER`` records within ``CONSTANTS_BLOCK`` blocks.
525It is also used for phi instruction operands in `MODULE_CODE_VERSION`_ 1.
526
527LLVM IR Blocks
528^^^^^^^^^^^^^^
529
530LLVM IR is defined with the following blocks:
531
532* 8 --- `MODULE_BLOCK`_ --- This is the top-level block that contains the entire
533  module, and describes a variety of per-module information.
534
535* 9 --- `PARAMATTR_BLOCK`_ --- This enumerates the parameter attributes.
536
537* 10 --- `TYPE_BLOCK`_ --- This describes all of the types in the module.
538
539* 11 --- `CONSTANTS_BLOCK`_ --- This describes constants for a module or
540  function.
541
542* 12 --- `FUNCTION_BLOCK`_ --- This describes a function body.
543
544* 13 --- `TYPE_SYMTAB_BLOCK`_ --- This describes the type symbol table.
545
546* 14 --- `VALUE_SYMTAB_BLOCK`_ --- This describes a value symbol table.
547
548* 15 --- `METADATA_BLOCK`_ --- This describes metadata items.
549
550* 16 --- `METADATA_ATTACHMENT`_ --- This contains records associating metadata
551  with function instruction values.
552
553.. _MODULE_BLOCK:
554
555MODULE_BLOCK Contents
556---------------------
557
558The ``MODULE_BLOCK`` block (id 8) is the top-level block for LLVM bitcode files,
559and each bitcode file must contain exactly one. In addition to records
560(described below) containing information about the module, a ``MODULE_BLOCK``
561block may contain the following sub-blocks:
562
563* `BLOCKINFO`_
564* `PARAMATTR_BLOCK`_
565* `TYPE_BLOCK`_
566* `TYPE_SYMTAB_BLOCK`_
567* `VALUE_SYMTAB_BLOCK`_
568* `CONSTANTS_BLOCK`_
569* `FUNCTION_BLOCK`_
570* `METADATA_BLOCK`_
571
572.. _MODULE_CODE_VERSION:
573
574MODULE_CODE_VERSION Record
575^^^^^^^^^^^^^^^^^^^^^^^^^^
576
577``[VERSION, version#]``
578
579The ``VERSION`` record (code 1) contains a single value indicating the format
580version. Versions 0 and 1 are supported at this time. The difference between
581version 0 and 1 is in the encoding of instruction operands in
582each `FUNCTION_BLOCK`_.
583
584In version 0, each value defined by an instruction is assigned an ID
585unique to the function. Function-level value IDs are assigned starting from
586``NumModuleValues`` since they share the same namespace as module-level
587values. The value enumerator resets after each function. When a value is
588an operand of an instruction, the value ID is used to represent the operand.
589For large functions or large modules, these operand values can be large.
590
591The encoding in version 1 attempts to avoid large operand values
592in common cases. Instead of using the value ID directly, operands are
593encoded as relative to the current instruction. Thus, if an operand
594is the value defined by the previous instruction, the operand
595will be encoded as 1.
596
597For example, instead of
598
599.. code-block:: llvm
600
601  #n = load #n-1
602  #n+1 = icmp eq #n, #const0
603  br #n+1, label #(bb1), label #(bb2)
604
605version 1 will encode the instructions as
606
607.. code-block:: llvm
608
609  #n = load #1
610  #n+1 = icmp eq #1, (#n+1)-#const0
611  br #1, label #(bb1), label #(bb2)
612
613Note in the example that operands which are constants also use
614the relative encoding, while operands like basic block labels
615do not use the relative encoding.
616
617Forward references will result in a negative value.
618This can be inefficient, as operands are normally encoded
619as unsigned VBRs. However, forward references are rare, except in the
620case of phi instructions. For phi instructions, operands are encoded as
621`Signed VBRs`_ to deal with forward references.
622
623
624MODULE_CODE_TRIPLE Record
625^^^^^^^^^^^^^^^^^^^^^^^^^
626
627``[TRIPLE, ...string...]``
628
629The ``TRIPLE`` record (code 2) contains a variable number of values representing
630the bytes of the ``target triple`` specification string.
631
632MODULE_CODE_DATALAYOUT Record
633^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
634
635``[DATALAYOUT, ...string...]``
636
637The ``DATALAYOUT`` record (code 3) contains a variable number of values
638representing the bytes of the ``target datalayout`` specification string.
639
640MODULE_CODE_ASM Record
641^^^^^^^^^^^^^^^^^^^^^^
642
643``[ASM, ...string...]``
644
645The ``ASM`` record (code 4) contains a variable number of values representing
646the bytes of ``module asm`` strings, with individual assembly blocks separated
647by newline (ASCII 10) characters.
648
649.. _MODULE_CODE_SECTIONNAME:
650
651MODULE_CODE_SECTIONNAME Record
652^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
653
654``[SECTIONNAME, ...string...]``
655
656The ``SECTIONNAME`` record (code 5) contains a variable number of values
657representing the bytes of a single section name string. There should be one
658``SECTIONNAME`` record for each section name referenced (e.g., in global
659variable or function ``section`` attributes) within the module. These records
660can be referenced by the 1-based index in the *section* fields of ``GLOBALVAR``
661or ``FUNCTION`` records.
662
663MODULE_CODE_DEPLIB Record
664^^^^^^^^^^^^^^^^^^^^^^^^^
665
666``[DEPLIB, ...string...]``
667
668The ``DEPLIB`` record (code 6) contains a variable number of values representing
669the bytes of a single dependent library name string, one of the libraries
670mentioned in a ``deplibs`` declaration.  There should be one ``DEPLIB`` record
671for each library name referenced.
672
673MODULE_CODE_GLOBALVAR Record
674^^^^^^^^^^^^^^^^^^^^^^^^^^^^
675
676``[GLOBALVAR, pointer type, isconst, initid, linkage, alignment, section, visibility, threadlocal, unnamed_addr, externally_initialized, dllstorageclass, comdat]``
677
678The ``GLOBALVAR`` record (code 7) marks the declaration or definition of a
679global variable. The operand fields are:
680
681* *pointer type*: The type index of the pointer type used to point to this
682  global variable
683
684* *isconst*: Non-zero if the variable is treated as constant within the module,
685  or zero if it is not
686
687* *initid*: If non-zero, the value index of the initializer for this variable,
688  plus 1.
689
690.. _linkage type:
691
692* *linkage*: An encoding of the linkage type for this variable:
693
694  * ``external``: code 0
695  * ``weak``: code 1
696  * ``appending``: code 2
697  * ``internal``: code 3
698  * ``linkonce``: code 4
699  * ``dllimport``: code 5
700  * ``dllexport``: code 6
701  * ``extern_weak``: code 7
702  * ``common``: code 8
703  * ``private``: code 9
704  * ``weak_odr``: code 10
705  * ``linkonce_odr``: code 11
706  * ``available_externally``: code 12
707  * deprecated : code 13
708  * deprecated : code 14
709
710* alignment*: The logarithm base 2 of the variable's requested alignment, plus 1
711
712* *section*: If non-zero, the 1-based section index in the table of
713  `MODULE_CODE_SECTIONNAME`_ entries.
714
715.. _visibility:
716
717* *visibility*: If present, an encoding of the visibility of this variable:
718
719  * ``default``: code 0
720  * ``hidden``: code 1
721  * ``protected``: code 2
722
723.. _bcthreadlocal:
724
725* *threadlocal*: If present, an encoding of the thread local storage mode of the
726  variable:
727
728  * ``not thread local``: code 0
729  * ``thread local; default TLS model``: code 1
730  * ``localdynamic``: code 2
731  * ``initialexec``: code 3
732  * ``localexec``: code 4
733
734.. _bcunnamedaddr:
735
736* *unnamed_addr*: If present, an encoding of the ``unnamed_addr`` attribute of this
737  variable:
738
739  * not ``unnamed_addr``: code 0
740  * ``unnamed_addr``: code 1
741  * ``local_unnamed_addr``: code 2
742
743.. _bcdllstorageclass:
744
745* *dllstorageclass*: If present, an encoding of the DLL storage class of this variable:
746
747  * ``default``: code 0
748  * ``dllimport``: code 1
749  * ``dllexport``: code 2
750
751* *comdat*: An encoding of the COMDAT of this function
752
753.. _FUNCTION:
754
755MODULE_CODE_FUNCTION Record
756^^^^^^^^^^^^^^^^^^^^^^^^^^^
757
758``[FUNCTION, type, callingconv, isproto, linkage, paramattr, alignment, section, visibility, gc, prologuedata, dllstorageclass, comdat, prefixdata, personalityfn]``
759
760The ``FUNCTION`` record (code 8) marks the declaration or definition of a
761function. The operand fields are:
762
763* *type*: The type index of the function type describing this function
764
765* *callingconv*: The calling convention number:
766  * ``ccc``: code 0
767  * ``fastcc``: code 8
768  * ``coldcc``: code 9
769  * ``webkit_jscc``: code 12
770  * ``anyregcc``: code 13
771  * ``preserve_mostcc``: code 14
772  * ``preserve_allcc``: code 15
773  * ``swiftcc`` : code 16
774  * ``cxx_fast_tlscc``: code 17
775  * ``x86_stdcallcc``: code 64
776  * ``x86_fastcallcc``: code 65
777  * ``arm_apcscc``: code 66
778  * ``arm_aapcscc``: code 67
779  * ``arm_aapcs_vfpcc``: code 68
780
781* isproto*: Non-zero if this entry represents a declaration rather than a
782  definition
783
784* *linkage*: An encoding of the `linkage type`_ for this function
785
786* *paramattr*: If nonzero, the 1-based parameter attribute index into the table
787  of `PARAMATTR_CODE_ENTRY`_ entries.
788
789* *alignment*: The logarithm base 2 of the function's requested alignment, plus
790  1
791
792* *section*: If non-zero, the 1-based section index in the table of
793  `MODULE_CODE_SECTIONNAME`_ entries.
794
795* *visibility*: An encoding of the `visibility`_ of this function
796
797* *gc*: If present and nonzero, the 1-based garbage collector index in the table
798  of `MODULE_CODE_GCNAME`_ entries.
799
800* *unnamed_addr*: If present, an encoding of the
801  :ref:`unnamed_addr<bcunnamedaddr>` attribute of this function
802
803* *prologuedata*: If non-zero, the value index of the prologue data for this function,
804  plus 1.
805
806* *dllstorageclass*: An encoding of the
807  :ref:`dllstorageclass<bcdllstorageclass>` of this function
808
809* *comdat*: An encoding of the COMDAT of this function
810
811* *prefixdata*: If non-zero, the value index of the prefix data for this function,
812  plus 1.
813
814* *personalityfn*: If non-zero, the value index of the personality function for this function,
815  plus 1.
816
817MODULE_CODE_ALIAS Record
818^^^^^^^^^^^^^^^^^^^^^^^^
819
820``[ALIAS, alias type, aliasee val#, linkage, visibility, dllstorageclass, threadlocal, unnamed_addr]``
821
822The ``ALIAS`` record (code 9) marks the definition of an alias. The operand
823fields are
824
825* *alias type*: The type index of the alias
826
827* *aliasee val#*: The value index of the aliased value
828
829* *linkage*: An encoding of the `linkage type`_ for this alias
830
831* *visibility*: If present, an encoding of the `visibility`_ of the alias
832
833* *dllstorageclass*: If present, an encoding of the
834  :ref:`dllstorageclass<bcdllstorageclass>` of the alias
835
836* *threadlocal*: If present, an encoding of the
837  :ref:`thread local property<bcthreadlocal>` of the alias
838
839* *unnamed_addr*: If present, an encoding of the
840  :ref:`unnamed_addr<bcunnamedaddr>` attribute of this alias
841
842MODULE_CODE_PURGEVALS Record
843^^^^^^^^^^^^^^^^^^^^^^^^^^^^
844
845``[PURGEVALS, numvals]``
846
847The ``PURGEVALS`` record (code 10) resets the module-level value list to the
848size given by the single operand value. Module-level value list items are added
849by ``GLOBALVAR``, ``FUNCTION``, and ``ALIAS`` records.  After a ``PURGEVALS``
850record is seen, new value indices will start from the given *numvals* value.
851
852.. _MODULE_CODE_GCNAME:
853
854MODULE_CODE_GCNAME Record
855^^^^^^^^^^^^^^^^^^^^^^^^^
856
857``[GCNAME, ...string...]``
858
859The ``GCNAME`` record (code 11) contains a variable number of values
860representing the bytes of a single garbage collector name string. There should
861be one ``GCNAME`` record for each garbage collector name referenced in function
862``gc`` attributes within the module. These records can be referenced by 1-based
863index in the *gc* fields of ``FUNCTION`` records.
864
865.. _PARAMATTR_BLOCK:
866
867PARAMATTR_BLOCK Contents
868------------------------
869
870The ``PARAMATTR_BLOCK`` block (id 9) contains a table of entries describing the
871attributes of function parameters. These entries are referenced by 1-based index
872in the *paramattr* field of module block `FUNCTION`_ records, or within the
873*attr* field of function block ``INST_INVOKE`` and ``INST_CALL`` records.
874
875Entries within ``PARAMATTR_BLOCK`` are constructed to ensure that each is unique
876(i.e., no two indices represent equivalent attribute lists).
877
878.. _PARAMATTR_CODE_ENTRY:
879
880PARAMATTR_CODE_ENTRY Record
881^^^^^^^^^^^^^^^^^^^^^^^^^^^
882
883``[ENTRY, paramidx0, attr0, paramidx1, attr1...]``
884
885The ``ENTRY`` record (code 1) contains an even number of values describing a
886unique set of function parameter attributes. Each *paramidx* value indicates
887which set of attributes is represented, with 0 representing the return value
888attributes, 0xFFFFFFFF representing function attributes, and other values
889representing 1-based function parameters. Each *attr* value is a bitmap with the
890following interpretation:
891
892* bit 0: ``zeroext``
893* bit 1: ``signext``
894* bit 2: ``noreturn``
895* bit 3: ``inreg``
896* bit 4: ``sret``
897* bit 5: ``nounwind``
898* bit 6: ``noalias``
899* bit 7: ``byval``
900* bit 8: ``nest``
901* bit 9: ``readnone``
902* bit 10: ``readonly``
903* bit 11: ``noinline``
904* bit 12: ``alwaysinline``
905* bit 13: ``optsize``
906* bit 14: ``ssp``
907* bit 15: ``sspreq``
908* bits 16-31: ``align n``
909* bit 32: ``nocapture``
910* bit 33: ``noredzone``
911* bit 34: ``noimplicitfloat``
912* bit 35: ``naked``
913* bit 36: ``inlinehint``
914* bits 37-39: ``alignstack n``, represented as the logarithm
915  base 2 of the requested alignment, plus 1
916
917.. _TYPE_BLOCK:
918
919TYPE_BLOCK Contents
920-------------------
921
922The ``TYPE_BLOCK`` block (id 10) contains records which constitute a table of
923type operator entries used to represent types referenced within an LLVM
924module. Each record (with the exception of `NUMENTRY`_) generates a single type
925table entry, which may be referenced by 0-based index from instructions,
926constants, metadata, type symbol table entries, or other type operator records.
927
928Entries within ``TYPE_BLOCK`` are constructed to ensure that each entry is
929unique (i.e., no two indices represent structurally equivalent types).
930
931.. _TYPE_CODE_NUMENTRY:
932.. _NUMENTRY:
933
934TYPE_CODE_NUMENTRY Record
935^^^^^^^^^^^^^^^^^^^^^^^^^
936
937``[NUMENTRY, numentries]``
938
939The ``NUMENTRY`` record (code 1) contains a single value which indicates the
940total number of type code entries in the type table of the module. If present,
941``NUMENTRY`` should be the first record in the block.
942
943TYPE_CODE_VOID Record
944^^^^^^^^^^^^^^^^^^^^^
945
946``[VOID]``
947
948The ``VOID`` record (code 2) adds a ``void`` type to the type table.
949
950TYPE_CODE_HALF Record
951^^^^^^^^^^^^^^^^^^^^^
952
953``[HALF]``
954
955The ``HALF`` record (code 10) adds a ``half`` (16-bit floating point) type to
956the type table.
957
958TYPE_CODE_FLOAT Record
959^^^^^^^^^^^^^^^^^^^^^^
960
961``[FLOAT]``
962
963The ``FLOAT`` record (code 3) adds a ``float`` (32-bit floating point) type to
964the type table.
965
966TYPE_CODE_DOUBLE Record
967^^^^^^^^^^^^^^^^^^^^^^^
968
969``[DOUBLE]``
970
971The ``DOUBLE`` record (code 4) adds a ``double`` (64-bit floating point) type to
972the type table.
973
974TYPE_CODE_LABEL Record
975^^^^^^^^^^^^^^^^^^^^^^
976
977``[LABEL]``
978
979The ``LABEL`` record (code 5) adds a ``label`` type to the type table.
980
981TYPE_CODE_OPAQUE Record
982^^^^^^^^^^^^^^^^^^^^^^^
983
984``[OPAQUE]``
985
986The ``OPAQUE`` record (code 6) adds an ``opaque`` type to the type table. Note
987that distinct ``opaque`` types are not unified.
988
989TYPE_CODE_INTEGER Record
990^^^^^^^^^^^^^^^^^^^^^^^^
991
992``[INTEGER, width]``
993
994The ``INTEGER`` record (code 7) adds an integer type to the type table. The
995single *width* field indicates the width of the integer type.
996
997TYPE_CODE_POINTER Record
998^^^^^^^^^^^^^^^^^^^^^^^^
999
1000``[POINTER, pointee type, address space]``
1001
1002The ``POINTER`` record (code 8) adds a pointer type to the type table. The
1003operand fields are
1004
1005* *pointee type*: The type index of the pointed-to type
1006
1007* *address space*: If supplied, the target-specific numbered address space where
1008  the pointed-to object resides. Otherwise, the default address space is zero.
1009
1010TYPE_CODE_FUNCTION Record
1011^^^^^^^^^^^^^^^^^^^^^^^^^
1012
1013``[FUNCTION, vararg, ignored, retty, ...paramty... ]``
1014
1015The ``FUNCTION`` record (code 9) adds a function type to the type table. The
1016operand fields are
1017
1018* *vararg*: Non-zero if the type represents a varargs function
1019
1020* *ignored*: This value field is present for backward compatibility only, and is
1021  ignored
1022
1023* *retty*: The type index of the function's return type
1024
1025* *paramty*: Zero or more type indices representing the parameter types of the
1026  function
1027
1028TYPE_CODE_STRUCT Record
1029^^^^^^^^^^^^^^^^^^^^^^^
1030
1031``[STRUCT, ispacked, ...eltty...]``
1032
1033The ``STRUCT`` record (code 10) adds a struct type to the type table. The
1034operand fields are
1035
1036* *ispacked*: Non-zero if the type represents a packed structure
1037
1038* *eltty*: Zero or more type indices representing the element types of the
1039  structure
1040
1041TYPE_CODE_ARRAY Record
1042^^^^^^^^^^^^^^^^^^^^^^
1043
1044``[ARRAY, numelts, eltty]``
1045
1046The ``ARRAY`` record (code 11) adds an array type to the type table.  The
1047operand fields are
1048
1049* *numelts*: The number of elements in arrays of this type
1050
1051* *eltty*: The type index of the array element type
1052
1053TYPE_CODE_VECTOR Record
1054^^^^^^^^^^^^^^^^^^^^^^^
1055
1056``[VECTOR, numelts, eltty]``
1057
1058The ``VECTOR`` record (code 12) adds a vector type to the type table.  The
1059operand fields are
1060
1061* *numelts*: The number of elements in vectors of this type
1062
1063* *eltty*: The type index of the vector element type
1064
1065TYPE_CODE_X86_FP80 Record
1066^^^^^^^^^^^^^^^^^^^^^^^^^
1067
1068``[X86_FP80]``
1069
1070The ``X86_FP80`` record (code 13) adds an ``x86_fp80`` (80-bit floating point)
1071type to the type table.
1072
1073TYPE_CODE_FP128 Record
1074^^^^^^^^^^^^^^^^^^^^^^
1075
1076``[FP128]``
1077
1078The ``FP128`` record (code 14) adds an ``fp128`` (128-bit floating point) type
1079to the type table.
1080
1081TYPE_CODE_PPC_FP128 Record
1082^^^^^^^^^^^^^^^^^^^^^^^^^^
1083
1084``[PPC_FP128]``
1085
1086The ``PPC_FP128`` record (code 15) adds a ``ppc_fp128`` (128-bit floating point)
1087type to the type table.
1088
1089TYPE_CODE_METADATA Record
1090^^^^^^^^^^^^^^^^^^^^^^^^^
1091
1092``[METADATA]``
1093
1094The ``METADATA`` record (code 16) adds a ``metadata`` type to the type table.
1095
1096.. _CONSTANTS_BLOCK:
1097
1098CONSTANTS_BLOCK Contents
1099------------------------
1100
1101The ``CONSTANTS_BLOCK`` block (id 11) ...
1102
1103.. _FUNCTION_BLOCK:
1104
1105FUNCTION_BLOCK Contents
1106-----------------------
1107
1108The ``FUNCTION_BLOCK`` block (id 12) ...
1109
1110In addition to the record types described below, a ``FUNCTION_BLOCK`` block may
1111contain the following sub-blocks:
1112
1113* `CONSTANTS_BLOCK`_
1114* `VALUE_SYMTAB_BLOCK`_
1115* `METADATA_ATTACHMENT`_
1116
1117.. _TYPE_SYMTAB_BLOCK:
1118
1119TYPE_SYMTAB_BLOCK Contents
1120--------------------------
1121
1122The ``TYPE_SYMTAB_BLOCK`` block (id 13) contains entries which map between
1123module-level named types and their corresponding type indices.
1124
1125.. _TST_CODE_ENTRY:
1126
1127TST_CODE_ENTRY Record
1128^^^^^^^^^^^^^^^^^^^^^
1129
1130``[ENTRY, typeid, ...string...]``
1131
1132The ``ENTRY`` record (code 1) contains a variable number of values, with the
1133first giving the type index of the designated type, and the remaining values
1134giving the character codes of the type name. Each entry corresponds to a single
1135named type.
1136
1137.. _VALUE_SYMTAB_BLOCK:
1138
1139VALUE_SYMTAB_BLOCK Contents
1140---------------------------
1141
1142The ``VALUE_SYMTAB_BLOCK`` block (id 14) ...
1143
1144.. _METADATA_BLOCK:
1145
1146METADATA_BLOCK Contents
1147-----------------------
1148
1149The ``METADATA_BLOCK`` block (id 15) ...
1150
1151.. _METADATA_ATTACHMENT:
1152
1153METADATA_ATTACHMENT Contents
1154----------------------------
1155
1156The ``METADATA_ATTACHMENT`` block (id 16) ...
1157