blob: ed9bd082b56ad351a320c4d5831ef05ec8e98a97 [file] [log] [blame]
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
"http://www.w3.org/TR/html4/strict.dtd">
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<title>LLVM Bitcode File Format</title>
<link rel="stylesheet" href="llvm.css" type="text/css">
</head>
<body>
<div class="doc_title"> LLVM Bitcode File Format </div>
<ol>
<li><a href="#abstract">Abstract</a></li>
<li><a href="#overview">Overview</a></li>
<li><a href="#bitstream">Bitstream Format</a>
<ol>
<li><a href="#magic">Magic Numbers</a></li>
<li><a href="#primitives">Primitives</a></li>
<li><a href="#abbrevid">Abbreviation IDs</a></li>
<li><a href="#blocks">Blocks</a></li>
<li><a href="#datarecord">Data Records</a></li>
<li><a href="#abbreviations">Abbreviations</a></li>
<li><a href="#stdblocks">Standard Blocks</a></li>
</ol>
</li>
<li><a href="#wrapper">Bitcode Wrapper Format</a>
</li>
<li><a href="#llvmir">LLVM IR Encoding</a>
<ol>
<li><a href="#basics">Basics</a></li>
</ol>
</li>
</ol>
<div class="doc_author">
<p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>
and <a href="http://www.reverberate.org">Joshua Haberman</a>.
</p>
</div>
<!-- *********************************************************************** -->
<div class="doc_section"> <a name="abstract">Abstract</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
<p>This document describes the LLVM bitstream file format and the encoding of
the LLVM IR into it.</p>
</div>
<!-- *********************************************************************** -->
<div class="doc_section"> <a name="overview">Overview</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
<p>
What is commonly known as the LLVM bitcode file format (also, sometimes
anachronistically known as bytecode) is actually two things: a <a
href="#bitstream">bitstream container format</a>
and an <a href="#llvmir">encoding of LLVM IR</a> into the container format.</p>
<p>
The bitstream format is an abstract encoding of structured data, very
similar to XML in some ways. Like XML, bitstream files contain tags, and nested
structures, and you can parse the file without having to understand the tags.
Unlike XML, the bitstream format is a binary encoding, and unlike XML it
provides a mechanism for the file to self-describe "abbreviations", which are
effectively size optimizations for the content.</p>
<p>LLVM IR files may be optionally embedded into a <a
href="#wrapper">wrapper</a> structure that makes it easy to embed extra data
along with LLVM IR files.</p>
<p>This document first describes the LLVM bitstream format, describes the
wrapper format, then describes the record structure used by LLVM IR files.
</p>
</div>
<!-- *********************************************************************** -->
<div class="doc_section"> <a name="bitstream">Bitstream Format</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
<p>
The bitstream format is literally a stream of bits, with a very simple
structure. This structure consists of the following concepts:
</p>
<ul>
<li>A "<a href="#magic">magic number</a>" that identifies the contents of
the stream.</li>
<li>Encoding <a href="#primitives">primitives</a> like variable bit-rate
integers.</li>
<li><a href="#blocks">Blocks</a>, which define nested content.</li>
<li><a href="#datarecord">Data Records</a>, which describe entities within the
file.</li>
<li>Abbreviations, which specify compression optimizations for the file.</li>
</ul>
<p>Note that the <a
href="CommandGuide/html/llvm-bcanalyzer.html">llvm-bcanalyzer</a> tool can be
used to dump and inspect arbitrary bitstreams, which is very useful for
understanding the encoding.</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection"><a name="magic">Magic Numbers</a>
</div>
<div class="doc_text">
<p>The first two bytes of a bitcode file are 'BC' (0x42, 0x43).
The second two bytes are an application-specific magic number. Generic
bitcode tools can look at only the first two bytes to verify the file is
bitcode, while application-specific programs will want to look at all four.</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection"><a name="primitives">Primitives</a>
</div>
<div class="doc_text">
<p>
A bitstream literally consists of a stream of bits, which are read in order
starting with the least significant bit of each byte. The stream is made up of a
number of primitive values that encode a stream of unsigned integer values.
These
integers are are encoded in two ways: either as <a href="#fixedwidth">Fixed
Width Integers</a> or as <a href="#variablewidth">Variable Width
Integers</a>.
</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"> <a name="fixedwidth">Fixed Width Integers</a>
</div>
<div class="doc_text">
<p>Fixed-width integer values have their low bits emitted directly to the file.
For example, a 3-bit integer value encodes 1 as 001. Fixed width integers
are used when there are a well-known number of options for a field. For
example, boolean values are usually encoded with a 1-bit wide integer.
</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"> <a name="variablewidth">Variable Width
Integers</a></div>
<div class="doc_text">
<p>Variable-width integer (VBR) values encode values of arbitrary size,
optimizing for the case where the values are small. Given a 4-bit VBR field,
any 3-bit value (0 through 7) is encoded directly, with the high bit set to
zero. Values larger than N-1 bits emit their bits in a series of N-1 bit
chunks, where all but the last set the high bit.</p>
<p>For example, the value 27 (0x1B) is encoded as 1011 0011 when emitted as a
vbr4 value. The first set of four bits indicates the value 3 (011) with a
continuation piece (indicated by a high bit of 1). The next word indicates a
value of 24 (011 << 3) with no continuation. The sum (3+24) yields the value
27.
</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"> <a name="char6">6-bit characters</a></div>
<div class="doc_text">
<p>6-bit characters encode common characters into a fixed 6-bit field. They
represent the following characters with the following 6-bit values:</p>
<ul>
<li>'a' .. 'z' - 0 .. 25</li>
<li>'A' .. 'Z' - 26 .. 51</li>
<li>'0' .. '9' - 52 .. 61</li>
<li>'.' - 62</li>
<li>'_' - 63</li>
</ul>
<p>This encoding is only suitable for encoding characters and strings that
consist only of the above characters. It is completely incapable of encoding
characters not in the set.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"> <a name="wordalign">Word Alignment</a></div>
<div class="doc_text">
<p>Occasionally, it is useful to emit zero bits until the bitstream is a
multiple of 32 bits. This ensures that the bit position in the stream can be
represented as a multiple of 32-bit words.</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection"><a name="abbrevid">Abbreviation IDs</a>
</div>
<div class="doc_text">
<p>
A bitstream is a sequential series of <a href="#blocks">Blocks</a> and
<a href="#datarecord">Data Records</a>. Both of these start with an
abbreviation ID encoded as a fixed-bitwidth field. The width is specified by
the current block, as described below. The value of the abbreviation ID
specifies either a builtin ID (which have special meanings, defined below) or
one of the abbreviation IDs defined by the stream itself.
</p>
<p>
The set of builtin abbrev IDs is:
</p>
<ul>
<li>0 - <a href="#END_BLOCK">END_BLOCK</a> - This abbrev ID marks the end of the
current block.</li>
<li>1 - <a href="#ENTER_SUBBLOCK">ENTER_SUBBLOCK</a> - This abbrev ID marks the
beginning of a new block.</li>
<li>2 - <a href="#DEFINE_ABBREV">DEFINE_ABBREV</a> - This defines a new
abbreviation.</li>
<li>3 - <a href="#UNABBREV_RECORD">UNABBREV_RECORD</a> - This ID specifies the
definition of an unabbreviated record.</li>
</ul>
<p>Abbreviation IDs 4 and above are defined by the stream itself, and specify
an <a href="#abbrev_records">abbreviated record encoding</a>.</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection"><a name="blocks">Blocks</a>
</div>
<div class="doc_text">
<p>
Blocks in a bitstream denote nested regions of the stream, and are identified by
a content-specific id number (for example, LLVM IR uses an ID of 12 to represent
function bodies). Block IDs 0-7 are reserved for <a href="#stdblocks">standard blocks</a>
whose meaning is defined by Bitcode; block IDs 8 and greater are
application specific. Nested blocks capture the hierachical structure of the data
encoded in it, and various properties are associated with blocks as the file is
parsed. Block definitions allow the reader to efficiently skip blocks
in constant time if the reader wants a summary of blocks, or if it wants to
efficiently skip data they do not understand. The LLVM IR reader uses this
mechanism to skip function bodies, lazily reading them on demand.
</p>
<p>
When reading and encoding the stream, several properties are maintained for the
block. In particular, each block maintains:
</p>
<ol>
<li>A current abbrev id width. This value starts at 2, and is set every time a
block record is entered. The block entry specifies the abbrev id width for
the body of the block.</li>
<li>A set of abbreviations. Abbreviations may be defined within a block, in
which case they are only defined in that block (neither subblocks nor
enclosing blocks see the abbreviation). Abbreviations can also be defined
inside a <a href="#BLOCKINFO">BLOCKINFO</a> block, in which case they are
defined in all blocks that match the ID that the BLOCKINFO block is describing.
</li>
</ol>
<p>As sub blocks are entered, these properties are saved and the new sub-block
has its own set of abbreviations, and its own abbrev id width. When a sub-block
is popped, the saved values are restored.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"> <a name="ENTER_SUBBLOCK">ENTER_SUBBLOCK
Encoding</a></div>
<div class="doc_text">
<p><tt>[ENTER_SUBBLOCK, blockid<sub>vbr8</sub>, newabbrevlen<sub>vbr4</sub>,
&lt;align32bits&gt;, blocklen<sub>32</sub>]</tt></p>
<p>
The ENTER_SUBBLOCK abbreviation ID specifies the start of a new block record.
The <tt>blockid</tt> value is encoded as a 8-bit VBR identifier, and indicates
the type of block being entered (which can be a <a href="#stdblocks">standard
block</a> or an application-specific block). The
<tt>newabbrevlen</tt> value is a 4-bit VBR which specifies the
abbrev id width for the sub-block. The <tt>blocklen</tt> is a 32-bit aligned
value that specifies the size of the subblock, in 32-bit words. This value
allows the reader to skip over the entire block in one jump.
</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"> <a name="END_BLOCK">END_BLOCK
Encoding</a></div>
<div class="doc_text">
<p><tt>[END_BLOCK, &lt;align32bits&gt;]</tt></p>
<p>
The END_BLOCK abbreviation ID specifies the end of the current block record.
Its end is aligned to 32-bits to ensure that the size of the block is an even
multiple of 32-bits.</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection"><a name="datarecord">Data Records</a>
</div>
<div class="doc_text">
<p>
Data records consist of a record code and a number of (up to) 64-bit integer
values. The interpretation of the code and values is application specific and
there are multiple different ways to encode a record (with an unabbrev record
or with an abbreviation). In the LLVM IR format, for example, there is a record
which encodes the target triple of a module. The code is MODULE_CODE_TRIPLE,
and the values of the record are the ascii codes for the characters in the
string.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"> <a name="UNABBREV_RECORD">UNABBREV_RECORD
Encoding</a></div>
<div class="doc_text">
<p><tt>[UNABBREV_RECORD, code<sub>vbr6</sub>, numops<sub>vbr6</sub>,
op0<sub>vbr6</sub>, op1<sub>vbr6</sub>, ...]</tt></p>
<p>An UNABBREV_RECORD provides a default fallback encoding, which is both
completely general and also extremely inefficient. It can describe an arbitrary
record, by emitting the code and operands as vbrs.</p>
<p>For example, emitting an LLVM IR target triple as an unabbreviated record
requires emitting the UNABBREV_RECORD abbrevid, a vbr6 for the
MODULE_CODE_TRIPLE code, a vbr6 for the length of the string (which is equal to
the number of operands), and a vbr6 for each character. Since there are no
letters with value less than 32, each letter would need to be emitted as at
least a two-part VBR, which means that each letter would require at least 12
bits. This is not an efficient encoding, but it is fully general.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"> <a name="abbrev_records">Abbreviated Record
Encoding</a></div>
<div class="doc_text">
<p><tt>[&lt;abbrevid&gt;, fields...]</tt></p>
<p>An abbreviated record is a abbreviation id followed by a set of fields that
are encoded according to the <a href="#abbreviations">abbreviation
definition</a>. This allows records to be encoded significantly more densely
than records encoded with the <a href="#UNABBREV_RECORD">UNABBREV_RECORD</a>
type, and allows the abbreviation types to be specified in the stream itself,
which allows the files to be completely self describing. The actual encoding
of abbreviations is defined below.
</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection"><a name="abbreviations">Abbreviations</a>
</div>
<div class="doc_text">
<p>
Abbreviations are an important form of compression for bitstreams. The idea is
to specify a dense encoding for a class of records once, then use that encoding
to emit many records. It takes space to emit the encoding into the file, but
the space is recouped (hopefully plus some) when the records that use it are
emitted.
</p>
<p>
Abbreviations can be determined dynamically per client, per file. Since the
abbreviations are stored in the bitstream itself, different streams of the same
format can contain different sets of abbreviations if the specific stream does
not need it. As a concrete example, LLVM IR files usually emit an abbreviation
for binary operators. If a specific LLVM module contained no or few binary
operators, the abbreviation does not need to be emitted.
</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"><a name="DEFINE_ABBREV">DEFINE_ABBREV
Encoding</a></div>
<div class="doc_text">
<p><tt>[DEFINE_ABBREV, numabbrevops<sub>vbr5</sub>, abbrevop0, abbrevop1,
...]</tt></p>
<p>A DEFINE_ABBREV record adds an abbreviation to the list of currently
defined abbreviations in the scope of this block. This definition only
exists inside this immediate block -- it is not visible in subblocks or
enclosing blocks.
Abbreviations are implicitly assigned IDs
sequentially starting from 4 (the first application-defined abbreviation ID).
Any abbreviations defined in a BLOCKINFO record receive IDs first, in order,
followed by any abbreviations defined within the block itself.
Abbreviated data records reference this ID to indicate what abbreviation
they are invoking.</p>
<p>An abbreviation definition consists of the DEFINE_ABBREV abbrevid followed
by a VBR that specifies the number of abbrev operands, then the abbrev
operands themselves. Abbreviation operands come in three forms. They all start
with a single bit that indicates whether the abbrev operand is a literal operand
(when the bit is 1) or an encoding operand (when the bit is 0).</p>
<ol>
<li>Literal operands - <tt>[1<sub>1</sub>, litvalue<sub>vbr8</sub>]</tt> -
Literal operands specify that the value in the result
is always a single specific value. This specific value is emitted as a vbr8
after the bit indicating that it is a literal operand.</li>
<li>Encoding info without data - <tt>[0<sub>1</sub>, encoding<sub>3</sub>]</tt>
- Operand encodings that do not have extra data are just emitted as their code.
</li>
<li>Encoding info with data - <tt>[0<sub>1</sub>, encoding<sub>3</sub>,
value<sub>vbr5</sub>]</tt> - Operand encodings that do have extra data are
emitted as their code, followed by the extra data.
</li>
</ol>
<p>The possible operand encodings are:</p>
<ul>
<li>1 - Fixed - The field should be emitted as a <a
href="#fixedwidth">fixed-width value</a>, whose width
is specified by the operand's extra data.</li>
<li>2 - VBR - The field should be emitted as a <a
href="#variablewidth">variable-width value</a>, whose width
is specified by the operand's extra data.</li>
<li>3 - Array - This field is an array of values. The array operand has no
extra data, but expects another operand to follow it which indicates the
element type of the array. When reading an array in an abbreviated record,
the first integer is a vbr6 that indicates the array length, followed by
the encoded elements of the array. An array may only occur as the last
operand of an abbreviation (except for the one final operand that gives
the array's type).</li>
<li>4 - Char6 - This field should be emitted as a <a href="#char6">char6-encoded
value</a>. This operand type takes no extra data.</li>
</ul>
<p>For example, target triples in LLVM modules are encoded as a record of the
form <tt>[TRIPLE, 'a', 'b', 'c', 'd']</tt>. Consider if the bitstream emitted
the following abbrev entry:</p>
<ul>
<li><tt>[0, Fixed, 4]</tt></li>
<li><tt>[0, Array]</tt></li>
<li><tt>[0, Char6]</tt></li>
</ul>
<p>When emitting a record with this abbreviation, the above entry would be
emitted as:</p>
<p><tt>[4<sub>abbrevwidth</sub>, 2<sub>4</sub>, 4<sub>vbr6</sub>,
0<sub>6</sub>, 1<sub>6</sub>, 2<sub>6</sub>, 3<sub>6</sub>]</tt></p>
<p>These values are:</p>
<ol>
<li>The first value, 4, is the abbreviation ID for this abbreviation.</li>
<li>The second value, 2, is the code for TRIPLE in LLVM IR files.</li>
<li>The third value, 4, is the length of the array.</li>
<li>The rest of the values are the char6 encoded values for "abcd".</li>
</ol>
<p>With this abbreviation, the triple is emitted with only 37 bits (assuming a
abbrev id width of 3). Without the abbreviation, significantly more space would
be required to emit the target triple. Also, since the TRIPLE value is not
emitted as a literal in the abbreviation, the abbreviation can also be used for
any other string value.
</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection"><a name="stdblocks">Standard Blocks</a>
</div>
<div class="doc_text">
<p>
In addition to the basic block structure and record encodings, the bitstream
also defines specific builtin block types. These block types specify how the
stream is to be decoded or other metadata. In the future, new standard blocks
may be added. Block IDs 0-7 are reserved for standard blocks.
</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"><a name="BLOCKINFO">#0 - BLOCKINFO
Block</a></div>
<div class="doc_text">
<p>The BLOCKINFO block allows the description of metadata for other blocks. The
currently specified records are:</p>
<ul>
<li><tt>[SETBID (#1), blockid]</tt></li>
<li><tt>[DEFINE_ABBREV, ...]</tt></li>
</ul>
<p>
The SETBID record indicates which block ID is being described. SETBID
records can occur multiple times throughout the block to change which
block ID is being described. There must be a SETBID record prior to
any other records.
</p>
<p>
Standard DEFINE_ABBREV records can occur inside BLOCKINFO blocks, but unlike
their occurrence in normal blocks, the abbreviation is defined for blocks
matching the block ID we are describing, <i>not</i> the BLOCKINFO block itself.
The abbreviations defined in BLOCKINFO blocks receive abbreviation ids
as described in <a href="#DEFINE_ABBREV">DEFINE_ABBREV</a>.
</p>
<p>
Note that although the data in BLOCKINFO blocks is described as "metadata," the
abbreviations they contain are essential for parsing records from the
corresponding blocks. It is not safe to skip them.
</p>
</div>
<!-- *********************************************************************** -->
<div class="doc_section"> <a name="wrapper">Bitcode Wrapper Format</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
<p>Bitcode files for LLVM IR may optionally be wrapped in a simple wrapper
structure. This structure contains a simple header that indicates the offset
and size of the embedded BC file. This allows additional information to be
stored alongside the BC file. The structure of this file header is:
</p>
<p>
<pre>
[Magic<sub>32</sub>,
Version<sub>32</sub>,
Offset<sub>32</sub>,
Size<sub>32</sub>,
CPUType<sub>32</sub>]
</pre></p>
<p>Each of the fields are 32-bit fields stored in little endian form (as with
the rest of the bitcode file fields). The Magic number is always
<tt>0x0B17C0DE</tt> and the version is currently always <tt>0</tt>. The Offset
field is the offset in bytes to the start of the bitcode stream in the file, and
the Size field is a size in bytes of the stream. CPUType is a target-specific
value that can be used to encode the CPU of the target.
</div>
<!-- *********************************************************************** -->
<div class="doc_section"> <a name="llvmir">LLVM IR Encoding</a></div>
<!-- *********************************************************************** -->
<div class="doc_text">
<p>LLVM IR is encoded into a bitstream by defining blocks and records. It uses
blocks for things like constant pools, functions, symbol tables, etc. It uses
records for things like instructions, global variable descriptors, type
descriptions, etc. This document does not describe the set of abbreviations
that the writer uses, as these are fully self-described in the file, and the
reader is not allowed to build in any knowledge of this.</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection"><a name="basics">Basics</a>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"><a name="ir_magic">LLVM IR Magic Number</a></div>
<div class="doc_text">
<p>
The magic number for LLVM IR files is:
</p>
<p><tt>[0x0<sub>4</sub>, 0xC<sub>4</sub>, 0xE<sub>4</sub>, 0xD<sub>4</sub>]</tt></p>
<p>When combined with the bitcode magic number and viewed as bytes, this is "BC 0xC0DE".</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"><a name="ir_signed_vbr">Signed VBRs</a></div>
<div class="doc_text">
<p>
<a href="#variablewidth">Variable Width Integers</a> are an efficient way to
encode arbitrary sized unsigned values, but is an extremely inefficient way to
encode signed values (as signed values are otherwise treated as maximally large
unsigned values).</p>
<p>As such, signed vbr values of a specific width are emitted as follows:</p>
<ul>
<li>Positive values are emitted as vbrs of the specified width, but with their
value shifted left by one.</li>
<li>Negative values are emitted as vbrs of the specified width, but the negated
value is shifted left by one, and the low bit is set.</li>
</ul>
<p>With this encoding, small positive and small negative values can both be
emitted efficiently.</p>
</div>
<!-- _______________________________________________________________________ -->
<div class="doc_subsubsection"><a name="ir_blocks">LLVM IR Blocks</a></div>
<div class="doc_text">
<p>
LLVM IR is defined with the following blocks:
</p>
<ul>
<li>8 - MODULE_BLOCK - This is the top-level block that contains the
entire module, and describes a variety of per-module information.</li>
<li>9 - PARAMATTR_BLOCK - This enumerates the parameter attributes.</li>
<li>10 - TYPE_BLOCK - This describes all of the types in the module.</li>
<li>11 - CONSTANTS_BLOCK - This describes constants for a module or
function.</li>
<li>12 - FUNCTION_BLOCK - This describes a function body.</li>
<li>13 - TYPE_SYMTAB_BLOCK - This describes the type symbol table.</li>
<li>14 - VALUE_SYMTAB_BLOCK - This describes a value symbol table.</li>
</ul>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection"><a name="MODULE_BLOCK">MODULE_BLOCK Contents</a>
</div>
<div class="doc_text">
<p>
</p>
</div>
<!-- *********************************************************************** -->
<hr>
<address> <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
<a href="http://validator.w3.org/check/referer"><img
src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
<a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
<a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
Last modified: $Date$
</address>
</body>
</html>