Blame - docs/BytecodeFormat.html - platform/external/llvm

blob: 9b85a378fd06189a4040d93d8e49b2d25244f48a [file] [log] [blame]

Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	1	<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
				2	"http://www.w3.org/TR/html4/strict.dtd">
				3	<html>
				4	<head>
				5	<title>LLVM Bytecode File Format</title>
				6	<link rel="stylesheet" href="llvm.css" type="text/css">
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	7	<style type="text/css">
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	8	TR, TD { border: 2px solid gray; padding-left: 4pt; padding-right: 4pt; padding-top: 2pt; padding-bottom: 2pt; }
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	9	TH { border: 2px solid gray; font-weight: bold; font-size: 105%; }
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	10	TABLE { text-align: center; border: 2px solid black;
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	11	border-collapse: collapse; margin-top: 1em; margin-left: 1em; margin-right: 1em; margin-bottom: 1em; }
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	12	.td_left { border: 2px solid gray; text-align: left; }
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	13	</style>
				14	</head>
				15	<body>
				16	<div class="doc_title"> LLVM Bytecode File Format </div>
				17	<ol>
				18	<li><a href="#abstract">Abstract</a></li>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	19	<li><a href="#concepts">Concepts</a>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	20	<ol>
				21	<li><a href="#blocks">Blocks</a></li>
				22	<li><a href="#lists">Lists</a></li>
				23	<li><a href="#fields">Fields</a></li>
				24	<li><a href="#align">Alignment</a></li>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	25	<li><a href="#encoding">Encoding Primitives</a></li>
				26	<li><a href="#slots">Slots</a></li>
				27	</ol>
				28	</li>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	29	<li><a href="#general">General Structure</a> </li>
				30	<li><a href="#blockdefs">Block Definitions</a>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	31	<ol>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	32	<li><a href="#signature">Signature Block</a></li>
				33	<li><a href="#module">Module Block</a></li>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	34	<li><a href="#globaltypes">Global Type Pool</a></li>
				35	<li><a href="#globalinfo">Module Info Block</a></li>
				36	<li><a href="#constantpool">Global Constant Pool</a></li>
				37	<li><a href="#functiondefs">Function Definition</a></li>
				38	<li><a href="#compactiontable">Compaction Table</a></li>
				39	<li><a href="#instructionlist">Instruction List</a></li>
				40	<li><a href="#symtab">Symbol Table</a></li>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	41	</ol>
				42	</li>
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	43	<li><a href="#versiondiffs">Version Differences</a>
				44	<ol>
				45	<li><a href="#vers12">Version 1.2 Differences From 1.3</a></li>
				46	<li><a href="#vers11">Version 1.1 Differences From 1.2</a></li>
				47	<li><a href="#vers10">Version 1.0 Differences From 1.1</a></li>
				48	</ol>
				49	</li>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	50	</ol>
Chris Lattner	8dabb50	2004-05-25 17:44:58 +0000	[diff] [blame]	51	<div class="doc_author">
				52	<p>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a>
				53	</p>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	54	</div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	55
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	56	<!-- *********************************************************************** -->
				57	<div class="doc_section"> <a name="abstract">Abstract </a></div>
				58	<!-- *********************************************************************** -->
				59	<div class="doc_text">
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	60	<p>This document describes the LLVM bytecode file format as of version 1.3.
				61	It specifies the binary encoding rules of the bytecode file format
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	62	so that equivalent systems can encode bytecode files correctly. The LLVM
				63	bytecode representation is used to store the intermediate representation on
				64	disk in compacted form.
				65	</p>
				66	</div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	67
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	68	<!-- *********************************************************************** -->
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	69	<div class="doc_section"> <a name="concepts">Concepts</a> </div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	70	<!-- *********************************************************************** -->
				71	<div class="doc_text">
				72	<p>This section describes the general concepts of the bytecode file format
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	73	without getting into bit and byte level specifics. Note that the LLVM bytecode
				74	format may change in the future, but will always be backwards compatible with
				75	older formats. This document only describes the most current version of the
				76	bytecode format.</p>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	77	</div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	78
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	79	<!-- _______________________________________________________________________ -->
				80	<div class="doc_subsection"><a name="blocks">Blocks</a> </div>
				81	<div class="doc_text">
				82	<p>LLVM bytecode files consist simply of a sequence of blocks of bytes.
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	83	Each block begins with an header of two unsigned integers. The first value
				84	identifies the type of block and the second value provides the size of the
				85	block in bytes. The block identifier is used because it is possible for entire
				86	blocks to be omitted from the file if they are empty. The block identifier helps
				87	the reader determine which kind of block is next in the file. Note that blocks
				88	can be nested within other blocks.</p>
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	89	<p> All blocks are variable length, and the block header specifies the size of
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	90	the block. All blocks begin on a byte index that is aligned to an even 32-bit
				91	boundary. That is, the first block is 32-bit aligned because it starts at offset
				92	0. Each block is padded with zero fill bytes to ensure that the next block also
				93	starts on a 32-bit boundary.</p>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	94	</div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	95
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	96	<!-- _______________________________________________________________________ -->
				97	<div class="doc_subsection"><a name="lists">Lists</a> </div>
				98	<div class="doc_text">
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	99	<p>LLVM Bytecode blocks often contain lists of things of a similar type. For
				100	example, a function contains a list of instructions and a function type
				101	contains a list of argument types. There are two basic types of lists:
				102	length lists, and null terminated lists, as described here:</p>
				103	<ul>
				104	<li><b>Length Lists</b>. Length lists are simply preceded by the number
				105	of items in the list. The bytecode reader will read the count first and
				106	then iterate that many times to read in the list contents.</li>
				107	<li><b>Null Terminated Lists</b>. For some lists, the number of elements
				108	in the list is not readily available at the time of writing the bytecode.
				109	In these cases, the list is terminated by some null value. What constitutes
				110	a null value differs, but it almost always boils down to a zero value.</li>
				111	</ul>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	112	</div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	113
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	114	<!-- _______________________________________________________________________ -->
				115	<div class="doc_subsection"><a name="fields">Fields</a> </div>
				116	<div class="doc_text">
				117	<p>Fields are units of information that LLVM knows how to write atomically.
				118	Most fields have a uniform length or some kind of length indication built into
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	119	their encoding. For example, a constant string (array of bytes) is
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	120	written simply as the length followed by the characters. Although this is
				121	similar to a list, constant strings are treated atomically and are thus
				122	fields.</p>
				123	<p>Fields use a condensed bit format specific to the type of information
				124	they must contain. As few bits as possible are written for each field. The
				125	sections that follow will provide the details on how these fields are
				126	written and how the bits are to be interpreted.</p>
				127	</div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	128
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	129	<!-- _______________________________________________________________________ -->
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	130	<div class="doc_subsection"><a name="align">Alignment</a> </div>
Reid Spencer	7aa940d	2004-05-25 15:47:57 +0000	[diff] [blame]	131	<div class="doc_text">
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	132	<p>To support cross-platform differences, the bytecode file is aligned on
				133	certain boundaries. This means that a small amount of padding (at most 3
				134	bytes) will be added to ensure that the next entry is aligned to a 32-bit
				135	boundary.</p>
Chris Lattner	8dabb50	2004-05-25 17:44:58 +0000	[diff] [blame]	136	</div>
				137
Reid Spencer	7aa940d	2004-05-25 15:47:57 +0000	[diff] [blame]	138	<!-- _______________________________________________________________________ -->
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	139	<div class="doc_subsection"><a name="encoding">Encoding Primitives</a> </div>
				140	<div class="doc_text">
				141	<p>Each field that can be put out is encoded into the file using a small set
				142	of primitives. The rules for these primitives are described below.</p>
				143	<h3>Variable Bit Rate Encoding</h3>
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	144	<p>Most of the values written to LLVM bytecode files are small integers. To
				145	minimize the number of bytes written for these quantities, an encoding
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	146	scheme similar to UTF-8 is used to write integer data. The scheme is known as
				147	variable bit rate (vbr) encoding. In this encoding, the high bit of each
				148	byte is used to indicate if more bytes follow. If (byte & 0x80) is non-zero
				149	in any given byte, it means there is another byte immediately following that
				150	also contributes to the value. For the final byte (byte & 0x80) is false
				151	(the high bit is not set). In each byte only the low seven bits contribute to
				152	the value. Consequently 32-bit quantities can take from one to <em>five</em>
				153	bytes to encode. In general, smaller quantities will encode in fewer bytes,
				154	as follows:</p>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	155	<table>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	156	<tr>
				157	<th>Byte #</th>
				158	<th>Significant Bits</th>
				159	<th>Maximum Value</th>
				160	</tr>
				161	<tr><td>1</td><td>0-6</td><td>127</td></tr>
				162	<tr><td>2</td><td>7-13</td><td>16,383</td></tr>
				163	<tr><td>3</td><td>14-20</td><td>2,097,151</td></tr>
				164	<tr><td>4</td><td>21-27</td><td>268,435,455</td></tr>
				165	<tr><td>5</td><td>28-34</td><td>34,359,738,367</td></tr>
				166	<tr><td>6</td><td>35-41</td><td>4,398,046,511,103</td></tr>
				167	<tr><td>7</td><td>42-48</td><td>562,949,953,421,311</td></tr>
				168	<tr><td>8</td><td>49-55</td><td>72,057,594,037,927,935</td></tr>
				169	<tr><td>9</td><td>56-62</td><td>9,223,372,036,854,775,807</td></tr>
				170	<tr><td>10</td><td>63-69</td><td>1,180,591,620,717,411,303,423</td></tr>
				171	</table>
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	172	<p>Note that in practice, the tenth byte could only encode bit 63
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	173	since the maximum quantity to use this encoding is a 64-bit integer.</p>
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	174
				175	<p><em>Signed</em> VBR values are encoded with the standard vbr encoding, but
				176	with the sign bit as the low order bit instead of the high order bit. This
				177	allows small negative quantities to be encoded efficiently. For example, -3
				178	is encoded as "((3 << 1) \| 1)" and 3 is encoded as "(3 << 1) \|
				179	0)", emitted with the standard vbr encoding above.</p>
				180
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	181	<p>The table below defines the encoding rules for type names used in the
				182	descriptions of blocks and fields in the next section. Any type name with
				183	the suffix <em>_vbr</em> indicate a quantity that is encoded using
				184	variable bit rate encoding as described above.</p>
				185	<table class="doc_table" >
				186	<tr>
				187	<th><b>Type</b></th>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	188	<th class="td_left"><b>Rule</b></th>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	189	</tr>
				190	<tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	191	<td><a name="unsigned">unsigned</a></td>
				192	<td class="td_left">A 32-bit unsigned integer that always occupies four
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	193	consecutive bytes. The unsigned integer is encoded using LSB first
				194	ordering. That is bits 2<sup>0</sup> through 2<sup>7</sup> are in the
				195	byte with the lowest file offset (little endian).</td>
				196	</tr><tr>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	197	<td><a name="uint32_vbr">uint32_vbr</a></td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	198	<td class="td_left">A 32-bit unsigned integer that occupies from one to five
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	199	bytes using variable bit rate encoding.</td>
				200	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	201	<td><a name="uint64_vbr">uint64_vbr</a></td>
				202	<td class="td_left">A 64-bit unsigned integer that occupies from one to ten
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	203	bytes using variable bit rate encoding.</td>
				204	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	205	<td><a name="int64_vbr">int64_vbr</a></td>
				206	<td class="td_left">A 64-bit signed integer that occupies from one to ten
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	207	bytes using the signed variable bit rate encoding.</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	208	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	209	<td><a name="char">char</a></td>
				210	<td class="td_left">A single unsigned character encoded into one byte</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	211	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	212	<td><a name="bit">bit</a></td>
				213	<td class="td_left">A single bit within some larger integer field.</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	214	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	215	<td><a name="string">string</a></td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	216	<td class="td_left">A uint32_vbr indicating the type of the constant string
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	217	which also includes its length, immediately followed by the characters of
				218	the string. There is no terminating null byte in the string.</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	219	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	220	<td><a name="data">data</a></td>
				221	<td class="td_left">An arbitrarily long segment of data to which no
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	222	interpretation is implied. This is used for float, double, and constant
				223	initializers.</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	224	</tr><tr>
				225	<td><a name="block">block</a></td>
				226	<td class="td_left">A block of data that is logically related. A block
				227	begins with an <a href="#unsigned">unsigned</a> that provides the block
				228	identifier (constant value) and an <a href="#unsigned">unsigned</a> that
				229	provides the length of the block. Blocks may compose other blocks.
				230	</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	231	</tr>
				232	</table>
				233	</div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	234
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	235	<!-- _______________________________________________________________________ -->
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	236	<div class="doc_subsection"><a name="slots">Slots</a> </div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	237	<div class="doc_text">
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	238	<p>The bytecode format uses the notion of a "slot" to reference Types and
				239	Values. Since the bytecode file is a <em>direct</em> representation of LLVM's
				240	intermediate representation, there is a need to represent pointers in the file.
				241	Slots are used for this purpose. For example, if one has the following assembly:
				242	</p>
				243	<div class="doc_code">
				244	%MyType = type { int, sbyte }<br>
				245	%MyVar = external global %MyType
				246	</div>
				247	<p>there are two definitions. The definition of <tt>%MyVar</tt> uses
				248	<tt>%MyType</tt>. In the C++ IR this linkage between <tt>%MyVar</tt> and
				249	<tt>%MyType</tt> is
				250	explicit through the use of C++ pointers. In bytecode, however, there's no
				251	ability to store memory addresses. Instead, we compute and write out slot
				252	numbers for every Type and Value written to the file.</p>
				253	<p>A slot number is simply an unsigned 32-bit integer encoded in the variable
				254	bit rate scheme (see <a href="#encoding">encoding</a>). This ensures that
				255	low slot numbers are encoded in one byte. Through various bits of magic LLVM
				256	attempts to always keep the slot numbers low. The first attempt is to associate
				257	slot numbers with their "type plane". That is, Values of the same type are
				258	written to the bytecode file in a list (sequentially). Their order in that list
				259	determines their slot number. This means that slot #1 doesn't mean anything
				260	unless you also specify for which type you want slot #1. Types are handled
				261	specially and are always written to the file first (in the
				262	<a href="#globaltypes">Global Type Pool</a>) and
				263	in such a way that both forward and backward references of the types can often be
				264	resolved with a single pass through the type pool. </p>
				265	<p>Slot numbers are also kept small by rearranging their order. Because of the
				266	structure of LLVM, certain values are much more likely to be used frequently
				267	in the body of a function. For this reason, a compaction table is provided in
				268	the body of a function if its use would make the function body smaller.
				269	Suppose you have a function body that uses just the types "int*" and "{double}"
				270	but uses them thousands of time. Its worthwhile to ensure that the slot number
				271	for these types are low so they can be encoded in a single byte (via vbr).
				272	This is exactly what the compaction table does.</p>
				273	</div>
				274
				275	<!-- *********************************************************************** -->
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	276	<div class="doc_section"> <a name="general">General Structure</a> </div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	277	<!-- *********************************************************************** -->
				278	<div class="doc_text">
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	279	<p>This section provides the general structur of the LLVM bytecode file
				280	format. The bytecode file format requires blocks to be in a certain order and
				281	nested in a particular way so that an LLVM module can be constructed
				282	efficiently from the contents of the file. This ordering defines a general
				283	structure for bytecode files as shown below. The table below shows the order
				284	in which all block types may appear. Please note that some of the blocks are
				285	optional and some may be repeated. The structure is fairly loose because
				286	optional blocks, if empty, are completely omitted from the file.</p>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	287	<table>
				288	<tr>
				289	<th>ID</th>
				290	<th>Parent</th>
				291	<th>Optional?</th>
				292	<th>Repeated?</th>
				293	<th>Level</th>
				294	<th>Block Type</th>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	295	<th>Description</th>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	296	</tr>
				297	<tr><td>N/A</td><td>File</td><td>No</td><td>No</td><td>0</td>
				298	<td class="td_left"><a href="#signature">Signature</a></td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	299	<td class="td_left">This contains the file signature (magic number)
				300	that identifies the file as LLVM bytecode.</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	301	</tr>
				302	<tr><td>0x01</td><td>File</td><td>No</td><td>No</td><td>0</td>
				303	<td class="td_left"><a href="#module">Module</a></td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	304	<td class="td_left">This is the top level block in a bytecode file. It
				305	contains all the other blocks.</li>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	306	</tr>
				307	<tr><td>0x15</td><td>Module</td><td>No</td><td>No</td><td>1</td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	308	<td class="td_left">   <a href="#globaltypes">Global Type Pool</a></td>
				309	<td class="td_left">This block contains all the global (module) level
				310	types.</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	311	</tr>
				312	<tr><td>0x14</td><td>Module</td><td>No</td><td>No</td><td>1</td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	313	<td class="td_left">   <a href="#globalinfo">Module Globals Info</a></td>
				314	<td class="td_left">This block contains the type, constness, and linkage
				315	for each of the global variables in the module. It also contains the
				316	type of the functions and the constant initializers.</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	317	</tr>
				318	<tr><td>0x12</td><td>Module</td><td>Yes</td><td>No</td><td>1</td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	319	<td class="td_left">   <a href="#constantpool">Module Constant Pool</a></td>
				320	<td class="td_left">This block contains all the global constants
				321	except function arguments, global values and constant strings.</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	322	</tr>
				323	<tr><td>0x11</td><td>Module</td><td>Yes</td><td>Yes</td><td>1</td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	324	<td class="td_left">   <a href="#functiondefs">Function Definitions</a></td>
				325	<td class="td_left">One function block is written for each function in
				326	the module. The function block contains the instructions, compaction
				327	table, type constant pool, and symbol table for the function.</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	328	</tr>
				329	<tr><td>0x12</td><td>Function</td><td>Yes</td><td>No</td><td>2</td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	330	<td class="td_left">      <a href="#constantpool">Function Constant Pool</a></td>
				331	<td class="td_left">Any constants (including types) used solely
				332	within the function are emitted here in the function constant pool.
				333	</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	334	</tr>
				335	<tr><td>0x33</td><td>Function</td><td>Yes</td><td>No</td><td>2</td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	336	<td class="td_left">      <a href="#compactiontable">Compaction Table</a></td>
				337	<td class="td_left">This table reduces bytecode size by providing a
				338	funtion-local mapping of type and value slot numbers to their
				339	global slot numbers</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	340	</tr>
				341	<tr><td>0x32</td><td>Function</td><td>No</td><td>No</td><td>2</td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	342	<td class="td_left">      <a href="#instructionlist">Instruction List</a></td>
				343	<td class="td_left">This block contains all the instructions of the
				344	function. The basic blocks are inferred by terminating instructions.
				345	</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	346	</tr>
				347	<tr><td>0x13</td><td>Function</td><td>Yes</td><td>No</td><td>2</td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	348	<td class="td_left">      <a href="#symtab">Function Symbol Table</a></td>
				349	<td class="td_left">This symbol table provides the names for the
				350	function specific values used (basic block labels mostly).</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	351	</tr>
				352	<tr><td>0x13</td><td>Module</td><td>Yes</td><td>No</td><td>1</td>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	353	<td class="td_left">   <a href="#symtab">Module Symbol Table</a></td>
				354	<td class="td_left">This symbol table provides the names for the various
				355	entries in the file that are not function specific (global vars, and
				356	functions mostly).</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	357	</tr>
				358	</table>
				359	<p>Use the links in the table or see <a href="#blocktypes">Block Types</a> for
				360	details about the contents of each of the block types.</p>
				361	</div>
				362
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	363	<!-- *********************************************************************** -->
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	364	<div class="doc_section"> <a name="blockdefs">Block Definitions</a> </div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	365	<!-- *********************************************************************** -->
				366	<div class="doc_text">
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	367	<p>This section provides the detailed layout of the individual block types
				368	in the LLVM bytecode file format. </p>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	369	</div>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	370
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	371	<!-- _______________________________________________________________________ -->
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	372	<div class="doc_subsection"><a name="signature">Signature Block</a> </div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	373	<div class="doc_text">
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	374	<p>The signature occurs in every LLVM bytecode file and is always first.
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	375	It simply provides a few bytes of data to identify the file as being an LLVM
				376	bytecode file. This block is always four bytes in length and differs from the
				377	other blocks because there is no identifier and no block length at the start
				378	of the block. Essentially, this block is just the "magic number" for the file.
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	379	<table>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	380	<tr>
Reid Spencer	939290f	2004-05-22 05:56:41 +0000	[diff] [blame]	381	<th><b>Type</b></th>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	382	<th class="td_left"><b>Field Description</b></th>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	383	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	384	<td><a href="#char">char</a></td>
				385	<td class="td_left">Constant "l" (0x6C)</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	386	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	387	<td><a href="#char">char</a></td>
				388	<td class="td_left">Constant "l" (0x6C)</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	389	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	390	<td><a href="#char">char</a></td>
				391	<td class="td_left">Constant "v" (0x76)</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	392	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	393	<td><a href="#char">char</a></td>
				394	<td class="td_left">Constant "m" (0x6D)</td>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	395	</tr>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	396	</table>
				397	</div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	398
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	399	<!-- _______________________________________________________________________ -->
				400	<div class="doc_subsection"><a name="module">Module Block</a> </div>
				401	<div class="doc_text">
				402	<p>The module block contains a small pre-amble and all the other blocks in
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	403	the file. The table below shows the structure of the module block. Note that it
				404	only provides the module identifier, size of the module block, and the format
				405	information. Everything else is contained in other blocks, described in other
				406	sections.</p>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	407	<table>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	408	<tr>
				409	<th><b>Type</b></th>
				410	<th class="td_left"><b>Field Description</b></th>
				411	</tr><tr>
				412	<td><a href="#unsigned">unsigned</a></td>
				413	<td class="td_left">Module Identifier (0x01)</td>
				414	</tr><tr>
				415	<td><a href="#unsigned">unsigned</a></td>
				416	<td class="td_left">Size of the module block in bytes</td>
				417	</tr><tr>
				418	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				419	<td class="td_left"><a href="#format">Format Information</a></td>
				420	</tr><tr>
				421	<td><a href="#block">block</a></td>
				422	<td class="td_left"><a href="#globaltypes">Global Type Pool</a></td>
				423	</tr><tr>
				424	<td><a href="#block">block</a></td>
				425	<td class="td_left"><a href="#globalinfo">Module Globals Info</a></td>
				426	</tr><tr>
				427	<td><a href="#block">block</a></td>
				428	<td class="td_left"><a href="#constantpool">Module Constant Pool</a></td>
				429	</tr><tr>
				430	<td><a href="#block">block</a></td>
				431	<td class="td_left"><a href="#functiondefs">Function Definitions</a></td>
				432	</tr><tr>
				433	<td><a href="#block">block</a></td>
				434	<td class="td_left"><a href="#symboltable">Module Symbol Table</a></td>
				435	</tr>
				436	</table>
				437	</div>
				438
				439	<!-- _______________________________________________________________________ -->
				440	<div class="doc_subsubsection"><a name="format">Format Information</a></div>
				441	<div class="doc_text">
				442	<p>The format information field is encoded into a 32-bit vbr-encoded unsigned
				443	integer as shown in the following table.</p>
				444	<table>
				445	<tr>
				446	<th><b>Bit(s)</b></th>
				447	<th><b>Type</b></th>
				448	<th class="td_left"><b>Description</b></th>
				449	</tr><tr>
				450	<td>0</td><td>bit</td>
				451	<td class="td_left">Big Endian?</td>
				452	</tr><tr>
				453	<td>1</td><td>bit</td>
				454	<td class="td_left">Pointers Are 64-bit?</td>
				455	</tr><tr>
				456	<td>2</td><td>bit</td>
				457	<td class="td_left">Has No Endianess?</td>
				458	</tr><tr>
				459	<td>3</td><td>bit</td>
				460	<td class="td_left">Has No Pointer Size?</td>
				461	</tr><tr>
				462	<td>4-31</td><td>bit</td>
				463	<td class="td_left">Bytecode Format Version</td>
				464	</tr>
				465	</table>
				466	<p>
				467	Of particular note, the bytecode format number is simply a 28-bit
				468	monotonically increase integer that identifies the version of the bytecode
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	469	format (which is not directly related to the LLVM release number). The
				470	bytecode versions defined so far are (note that this document only describes
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	471	the latest version, 1.3):</p>
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	472	<ul>
				473	<li>#0: LLVM 1.0 & 1.1</li>
				474	<li>#1: LLVM 1.2</li>
				475	<li>#2: LLVM 1.3</li>
				476	</ul>
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	477	<p>Note that we plan to eventually expand the target description capabilities
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	478	of bytecode files to <a href="http://llvm.cs.uiuc.edu/PR263">target triples</a>.
				479	</p>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	480	</div>
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	481
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	482	<!-- _______________________________________________________________________ -->
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	483	<div class="doc_subsection"><a name="globaltypes">Global Type Pool</a> </div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	484	<div class="doc_text">
Chris Lattner	2b90565	2004-05-24 05:35:17 +0000	[diff] [blame]	485	<p>The global type pool consists of type definitions. Their order of appearance
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	486	in the file determines their slot number (0 based). Slot numbers are used to
				487	replace pointers in the intermediate representation. Each slot number uniquely
				488	identifies one entry in a type plane (a collection of values of the same type).
				489	Since all values have types and are associated with the order in which the type
				490	pool is written, the global type pool <em>must</em> be written as the first
				491	block of a module. If it is not, attempts to read the file will fail because
				492	both forward and backward type resolution will not be possible.</p>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	493	<p>The type pool is simply a list of type definitions, as shown in the table
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	494	below.</p>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	495	<table>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	496	<tr>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	497	<th><b>Type</b></th>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	498	<th class="td_left"><b>Field Description</b></th>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	499	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	500	<td><a href="#unsigned">unsigned</a></td>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	501	<td class="td_left">Type Pool Identifier (0x15)</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	502	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	503	<td><a href="#unsigned">unsigned</a></td>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	504	<td class="td_left">Size in bytes of the type pool block.</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	505	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	506	<td><a href="#uint32_vbr">uint32_vbr</a></td>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	507	<td class="td_left">Number of type definitions that follow in the next
				508	field.</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	509	</tr><tr>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	510	<td><a href="#type">type</a></td>
				511	<td class="td_left">Each of the type definitions (see below)<sup>1</sup></td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	512	</tr>
				513	</table>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	514	Notes:
				515	<ol>
				516	<li>Repeated field.</li>
				517	</ol>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	518	</div>
				519	<!-- _______________________________________________________________________ -->
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	520	<div class="doc_subsubsection"><a name="type">Type Definitions</a></div>
				521	<div class="doc_text">
				522	<p>Types in the type pool are defined using a different format for each
				523	basic type of type as given in the following sections.</p>
				524	<h3>Primitive Types</h3>
				525	<p>The primitive types encompass the basic integer and floating point types</p>
				526	<table>
				527	<tr>
				528	<th><b>Type</b></th>
				529	<th class="td_left"><b>Description</b></th>
				530	</tr><tr>
				531	<td><a href="#uint32_vbr">uint32_vbr</td>
				532	<td class="td_left">Type ID For The Primitive (1-11)<sup>1</sup></td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	533	</tr>
				534	</table>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	535	Notes:
				536	<ol>
				537	<li>See the definition of Type::TypeID in Type.h for the numeric equivalents
				538	of the primitive type ids.</li>
				539	</ol>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	540	<h3>Function Types</h3>
				541	<table>
				542	<tr>
				543	<th><b>Type</b></th>
				544	<th class="td_left"><b>Description</b></th>
				545	</tr><tr>
				546	<td><a href="#uint32_vbr">uint32_vbr</td>
				547	<td class="td_left">Type ID for function types (13)</td>
				548	</tr><tr>
				549	<td><a href="#uint32_vbr">uint32_vbr</td>
				550	<td class="td_left">Slot number of function's return type.</td>
				551	</tr><tr>
				552	<td><a href="#uint32_vbr">uint32_vbr</td>
				553	<td class="td_left">The number of arguments in the function.</td>
				554	</tr><tr>
				555	<td><a href="#uint32_vbr">uint32_vbr</td>
				556	<td class="td_left">Slot number of each argument's type.<sup>1</sup></td>
				557	</tr><tr>
				558	<td><a href="#uint32_vbr">uint32_vbr</td>
				559	<td class="td_left">Value 0 if this is a varargs function.<sup>2</sup></td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	560	</tr>
				561	</table>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	562	Notes:
				563	<ol>
				564	<li>Repeated field.</li>
				565	<li>Optional field.</li>
				566	</ol>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	567	<h3>Structure Types</h3>
				568	<table>
				569	<tr>
				570	<th><b>Type</b></th>
				571	<th class="td_left"><b>Description</b></th>
				572	</tr><tr>
				573	<td><a href="#uint32_vbr">uint32_vbr</td>
				574	<td class="td_left">Type ID for structure types (14)</td>
				575	</tr><tr>
				576	<td><a href="#uint32_vbr">uint32_vbr</td>
				577	<td class="td_left">Slot number of each of the element's fields.<sup>1</sup></td>
				578	</tr><tr>
				579	<td><a href="#uint32_vbr">uint32_vbr</td>
				580	<td class="td_left">Null Terminator (VoidTy type id)</td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	581	</tr>
				582	</table>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	583	Notes:
				584	<ol>
				585	<li>Repeatable field.</li>
				586	</ol>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	587	<h3>Array Types</h3>
				588	<table>
				589	<tr>
				590	<th><b>Type</b></th>
				591	<th class="td_left"><b>Description</b></th>
				592	</tr><tr>
				593	<td><a href="#uint32_vbr">uint32_vbr</td>
				594	<td class="td_left">Type ID for Array Types (15)</td>
				595	</tr><tr>
				596	<td><a href="#uint32_vbr">uint32_vbr</td>
				597	<td class="td_left">Slot number of array's element type.</td>
				598	</tr><tr>
				599	<td><a href="#uint32_vbr">uint32_vbr</td>
				600	<td class="td_left">The number of elements in the array.</td>
				601	</tr>
				602	</table>
				603	<h3>Pointer Types</h3>
				604	<table>
				605	<tr>
				606	<th><b>Type</b></th>
				607	<th class="td_left"><b>Description</b></th>
				608	</tr><tr>
				609	<td><a href="#uint32_vbr">uint32_vbr</td>
				610	<td class="td_left">Type ID For Pointer Types (16)</td>
				611	</tr><tr>
				612	<td><a href="#uint32_vbr">uint32_vbr</td>
				613	<td class="td_left">Slot number of pointer's element type.</td>
				614	</tr>
				615	</table>
				616	<h3>Opaque Types</h3>
				617	<table>
				618	<tr>
				619	<th><b>Type</b></th>
				620	<th class="td_left"><b>Description</b></th>
				621	</tr><tr>
				622	<td><a href="#uint32_vbr">uint32_vbr</td>
				623	<td class="td_left">Type ID For Opaque Types (17)</td>
				624	</tr>
				625	</table>
				626	</div>
				627	<!-- _______________________________________________________________________ -->
				628	<div class="doc_subsection"><a name="globalinfo">Module Global Info</a> </div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	629	<div class="doc_text">
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	630	<p>The module global info block contains the definitions of all global
				631	variables including their initializers and the <em>declaration</em> of all
				632	functions. The format is shown in the table below</p>
				633	<table>
				634	<tr>
				635	<th><b>Type</b></th>
				636	<th class="td_left"><b>Field Description</b></th>
				637	</tr><tr>
				638	<td><a href="#unsigned">unsigned</a></td>
				639	<td class="td_left">Module global info identifier (0x14)</td>
				640	</tr><tr>
				641	<td><a href="#unsigned">unsigned</a></td>
				642	<td class="td_left">Size in bytes of the module global info block.</td>
				643	</tr><tr>
				644	<td><a href="#globalvar">globalvar</a></td>
				645	<td class="td_left">Definition of the global variable (see below).
				646	<sup>1</sup>
				647	</td>
				648	</tr><tr>
				649	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				650	<td class="td_left">Slot number of the global variable's constant
				651	initializer.<sup>1,2</sup>
				652	</td>
				653	</tr><tr>
				654	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				655	<td class="td_left">Zero. This terminates the list of global variables.
				656	</td>
				657	</tr><tr>
				658	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				659	<td class="td_left">Type slot number of a function defined in this
				660	bytecode file.<sup>3</sup>
				661	</td>
				662	</tr><tr>
				663	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				664	<td class="td_left">Zero. This terminates the list of function
				665	declarations.
				666	</tr>
				667	</table>
				668	Notes:<ol>
				669	<li>Both these fields are repeatable but in pairs.</li>
				670	<li>Optional field.</li>
				671	<li>Repeatable field.</li>
				672	</ol>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	673	</div>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	674
				675	<!-- _______________________________________________________________________ -->
				676	<div class="doc_subsubsection"><a name="globalvar">Global Variable Field</a>
				677	</div>
				678	<div class="doc_text">
				679	<p>Global variables are written using a single
				680	<a href="#uint32_vbr">uint32_vbr</a> that encodes information about the global
				681	variable. The table below provides the bit layout of the value written for
				682	each global variable.</p>
				683	<table>
				684	<tr>
				685	<th><b>Bit(s)</b></th>
				686	<th><b>Type</b></th>
				687	<th class="td_left"><b>Description</b></th>
				688	</tr><tr>
				689	<td>0</td><td>bit</td>
				690	<td class="td_left">Is constant?</td>
				691	</tr><tr>
				692	<td>1</td><td>bit</td>
				693	<td class="td_left">Has initializer?<sup>1</sup></td>
				694	</tr><tr>
				695	<td>2-4</td><td>enumeration</td>
				696	<td class="td_left">Linkage type: 0=External, 1=Weak, 2=Appending,
				697	3=Internal, 4=LinkOnce</td>
				698	</tr><tr>
				699	<td>5-31</td><td>type slot</td>
				700	<td class="td_left">Slot number of type for the global variable.</td>
				701	</tr>
				702	</table>
				703	Notes:
				704	<ol>
				705	<li>This bit determines whether the constant initializer field follows
				706	immediately after this field</li>
				707	</ol>
				708	</div>
				709
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	710	<!-- _______________________________________________________________________ -->
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	711	<div class="doc_subsection"><a name="constantpool">Constant Pool</a> </div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	712	<div class="doc_text">
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	713	<p>A constant pool defines as set of constant values. There are actually two
				714	types of constant pool blocks: one for modules and one for functions. For
				715	modules, the block begins with the constant strings encountered anywhere in
				716	the module. For functions, the block begins with types only encountered in
				717	the function. In both cases the header is identical. The tables the follow,
				718	show the header, module constant pool preamble, function constant pool
				719	preamble, and the part common to both function and module constant pools.</p>
				720	<p><b>Common Block Header</b></p>
				721	<table>
				722	<tr>
				723	<th><b>Type</b></th>
				724	<th class="td_left"><b>Field Description</b></th>
				725	</tr><tr>
				726	<td><a href="#unsigned">unsigned</a></td>
				727	<td class="td_left">Constant pool identifier (0x12)</td>
				728	</tr>
				729	</table>
				730	<p><b>Module Constant Pool Preamble (constant strings)</b></p>
				731	<table>
				732	<tr>
				733	<th><b>Type</b></th>
				734	<th class="td_left"><b>Field Description</b></th>
				735	</tr><tr>
				736	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				737	<td class="td_left">The number of constant strings that follow.</td>
				738	</tr><tr>
				739	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				740	<td class="td_left">Zero. This identifies the following "plane" as
				741	containing the constant strings.
				742	</td>
				743	</tr><tr>
				744	<td><a href="#string">string</a></td>
				745	<td class="td_left">Slot number of the constant string's type which
				746	includes the length of the string.<sup>1</sup>
				747	</td>
				748	</tr>
				749	</table>
				750	Notes:
				751	<ol>
				752	<li>Repeated field.</li>
				753	</ol>
				754	<p><b>Function Constant Pool Preamble (function types)</b></p>
				755	<p>The structure of the types for functions is identical to the
				756	<a href="#globaltypes">Global Type Pool</a>. Please refer to that section
				757	for the details.
				758	<p><b>Common Part (other constants)</b></p>
				759	<table>
				760	<tr>
				761	<th><b>Type</b></th>
				762	<th class="td_left"><b>Field Description</b></th>
				763	</tr><tr>
				764	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				765	<td class="td_left">Number of entries in this type plane.</td>
				766	</tr><tr>
				767	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				768	<td class="td_left">Type slot number of this plane.</td>
				769	</tr><tr>
				770	<td><a href="#constant">constant</a></td>
				771	<td class="td_left">The definition of a constant (see below).</td>
				772	</tr>
				773	</table>
				774	</div>
				775	<!-- _______________________________________________________________________ -->
				776	<div class="doc_subsubsection"><a name="constant">Constant Field</a></div>
				777	<div class="doc_text">
				778	<p>Constants come in many shapes and flavors. The sections that followe define
				779	the format for each of them. All constants start with a
				780	<a href="#uint32_vbr">uint32_vbr</a> encoded integer that provides the number
				781	of operands for the constant. For primitive, structure, and array constants,
				782	this will always be zero since those types of constants have no operands.
				783	In this case, we have the following field definitions:</p>
				784	<ul>
				785	<li><b>Bool</b>. This is written as an <a href="#uint32_vbr">uint32_vbr</a>
				786	of value 1U or 0U.</li>
				787	<li><b>Signed Integers (sbyte,short,int,long)</b>. These are written as
				788	an <a href="#int64_vbr">int64_vbr</a> with the corresponding value.</li>
				789	<li><b>Unsigned Integers (ubyte,ushort,uint,ulong)</b>. These are written
				790	as an <a href="#uint64_vbr">uint64_vbr</a> with the corresponding value.
				791	</li>
				792	<li><b>Floating Point</b>. Both the float and double types are written
				793	literally in binary format.</li>
				794	<li><b>Arrays</b>. Arrays are written simply as a list of
				795	<a href="#uint32_vbr">uint32_vbr</a> encoded slot numbers to the constant
				796	element values.</li>
				797	<li><b>Structures</b>. Structures are written simply as a list of
				798	<a href="#uint32_vbr">uint32_vbr</a> encoded slot numbers to the constant
				799	field values of the structure.</li>
				800	</ul>
				801	<p>When the number of operands to the constant is non-zero, we have a
				802	constant expression and its field format is provided in the table below.</p>
				803	<table>
				804	<tr>
				805	<th><b>Type</b></th>
				806	<th class="td_left"><b>Field Description</b></th>
				807	</tr><tr>
				808	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				809	<td class="td_left">Op code of the instruction for the constant
				810	expression.</td>
				811	</tr><tr>
				812	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				813	<td class="td_left">The slot number of the constant value for an
				814	operand.<sup>1</sup></td>
				815	</tr><tr>
				816	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				817	<td class="td_left">The slot number for the type of the constant value
				818	for an operand.<sup>1</sup></td>
				819	</tr>
				820	</table>
				821	Notes:<ol>
				822	<li>Both these fields are repeatable but only in pairs.</li>
				823	</ol>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	824	</div>
				825	<!-- _______________________________________________________________________ -->
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	826	<div class="doc_subsection"><a name="functiondefs">Function Definition</a></div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	827	<div class="doc_text">
				828	<p>To be determined.</p>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	829	<table>
				830	<tr>
				831	<th><b>Type</b></th>
				832	<th class="td_left"><b>Field Description</b></th>
				833	</tr><tr>
				834	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				835	<td class="td_left">The linkage type of the function: 0=External, 1=Weak,
				836	2=Appending, 3=Internal, 4=LinkOnce<sup>1</sup></td>
				837	</tr><tr>
				838	<td><a href="#constantpool">constant pool</a></td>
				839	<td class="td_left">The constant pool block for this function.
				840	<sup>2</sup>
				841	</td>
				842	</tr><tr>
				843	<td><a href="#compactiontable">compaction table</a></td>
				844	<td class="td_left">The compaction table block for the function.
				845	<sup>2</sup>
				846	</td>
				847	</tr><tr>
				848	<td><a href="#instructionlist">instruction list</a></td>
				849	<td class="td_left">The list of instructions in the function.</td>
				850	</tr><tr>
				851	<td><a href="#symboltable">symbol table</a></td>
				852	<td class="td_left">The function's slot table containing only those
				853	symbols pertinent to the function (mostly block labels).
				854	</td>
				855	</tr>
				856	</table>
				857	Notes:<ol>
				858	<li>Note that if the linkage type is "External" then none of the other
				859	fields will be present as the function is defined elsewhere.</li>
				860	<li>Note that only one of the constant pool or compaction table will be
				861	written. Compaction tables are only written if they will actually save
				862	bytecode space. If not, then a regular constant pool is written.</li>
				863	</ol>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	864	</div>
				865	<!-- _______________________________________________________________________ -->
				866	<div class="doc_subsection"><a name="compactiontable">Compaction Table</a> </div>
				867	<div class="doc_text">
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	868	<p>Compaction tables are part of a function definition. They are merely a
				869	device for reducing the size of bytecode files. The size of a bytecode
				870	file is dependent on the <em>value</em> of the slot numbers used because
				871	larger values use more bytes in the variable bit rate encoding scheme.
				872	Furthermore, the compresses instruction format reserves only six bits for
				873	the type of the instruction. In large modules, declaring hundreds or thousands
				874	of types, the values of the slot numbers can be quite large. However,
				875	functions may use only a small fraction of the global types. In such cases
				876	a compaction table is created that maps the global type and value slot
				877	numbers to smaller values used by a function. Compaction tables have the
				878	format shown in the table below.</p>
				879	<table>
				880	<tr>
				881	<th><b>Type</b></th>
				882	<th class="td_left"><b>Field Description</b></th>
				883	</tr><tr>
				884	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				885	<td class="td_left">The number of types that follow</td>
				886	</tr><tr>
				887	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				888	<td class="td_left">The slot number in the global type plane of the
				889	type that will be referenced in the function with the index of
				890	this entry in the compaction table.<sup>1</sup></td>
				891	</tr><tr>
				892	<td><a href="#type_len">type_len</a></td>
				893	<td class="td_left">An encoding of the type and number of values that
				894	follow.<sup>2</sup></td>
				895	</tr><tr>
				896	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				897	<td class="td_left">The slot number in the globals of the value that
				898	will be referenced in the function with the index of this entry in
				899	the compaction table<sup>1</sup></td>
				900	</tr>
				901	</table>
				902	Notes:<ol>
				903	<li>Repeated field.</li>
				904	<li>This field's encoding varies depending on the size of the type plane.
				905	See <a href="#type_len">Type and Length</a> for further details.
				906	</ol>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	907	</div>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	908
				909	<!-- _______________________________________________________________________ -->
				910	<div class="doc_subsubsection"><a name="type_len">Type and Length</a></div>
				911	<div class="doc_text">
				912	<p>The type and length of a compaction table type plane is encoded differently
				913	depending on the length of the plane. For planes of length 1 or 2, the length
				914	is encoded into bits 0 and 1 of a <a href="#uint32_vbr">uint32_vbr</a> and the
				915	type is encoded into bits 2-31. Because type numbers are often small, this
				916	often saves an extra byte per plane. If the length of the plane is greater
				917	than 2 then the encoding uses a <a href="#uint32_vbr">uint32_vbr</a> for each
				918	of the length and type, in that order.</p>
				919	</div>
				920
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	921	<!-- _______________________________________________________________________ -->
				922	<div class="doc_subsection"><a name="instructionlist">Instruction List</a> </div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	923	<div class="doc_text">
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	924	<p>The instructions in a function are written as a simple list. Basic blocks
				925	are inferred by the terminating instruction types. The format of the block
				926	is given in the following table.</p>
				927	<table>
				928	<tr>
				929	<th><b>Type</b></th>
				930	<th class="td_left"><b>Field Description</b></th>
				931	</tr><tr>
				932	<td><a href="#unsigned">unsigned</a></td>
				933	<td class="td_left">Instruction list identifier (0x33).</td>
				934	</tr><tr>
				935	<td><a href="#unsigned">unsigned</a></td>
				936	<td class="td_left">Size in bytes of the instruction list.</td>
				937	</tr><tr>
				938	<td><a href="#instruction">instruction</a></td>
				939	<td class="td_left">An instruction.<sup>1</sup></td>
				940	</tr>
				941	</table>
				942	Notes:
				943	<ol>
				944	<li>A repeated field with a variety of formats. See
				945	<a href="#instruction">Instructions</a> for details.</li>
				946	</ol>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	947	</div>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	948
				949	<!-- _______________________________________________________________________ -->
				950	<div class="doc_subsubsection"><a name="instruction">Instructions</a></div>
				951	<div class="doc_text">
				952	<p>For brevity, instructions are written in one of four formats, depending on
				953	the number of operands to the instruction. Each instruction begins with a
				954	<a href="#uint32_vbr">uint32_vbr</a> that encodes the type of the instruction
				955	as well as other things. The tables that follow describe the format of this
				956	first word of each instruction.</p>
				957	<p><b>Instruction Format 0</b></p>
				958	<p>This format is used for a few instructions that can't easily be optimized
				959	because they have large numbers of operands (e.g. PHI Node or getelementptr).
				960	Each of the opcode, type, and operand fields is as successive fields.</p>
				961	<table>
				962	<tr>
				963	<th><b>Type</b></th>
				964	<th class="td_left"><b>Field Description</b></th>
				965	</tr><tr>
				966	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				967	<td class="td_left">Specifies the opcode of the instruction. Note that for
				968	compatibility with the other instruction formats, the opcode is shifted
				969	left by 2 bits. Bits 0 and 1 must have value zero for this format.</td>
				970	</tr><tr>
				971	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				972	<td class="td_left">Provides the slot number of the result type of the
				973	instruction</td>
				974	</tr><tr>
				975	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				976	<td class="td_left">The number of operands that follow.</td>
				977	</tr><tr>
				978	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				979	<td class="td_left">The slot number of the value for the operand(s).
				980	<sup>1,2</sup></td>
				981	</tr>
				982	</table>
				983	Notes:<ol>
				984	<li>Repeatable field (limit given by previous field).</li>
				985	<li>Note that if the instruction is a getelementptr and the type of the
				986	operand is a sequential type (array or pointer) then the slot number is
				987	shifted up two bits and the low order bits will encode the type of index
				988	used, as follows: 0=uint, 1=int, 2=ulong, 3=long.</li>
				989	</ol>
				990	<p><b>Instruction Format 1</b></p>
				991	<p>This format encodes the opcode, type and a single operand into a single
				992	<a href="#uint32_vbr">uint32_vbr</a> as follows:</p>
				993	<table>
				994	<tr>
				995	<th><b>Bits</b></th>
				996	<th><b>Type</b></th>
				997	<th class="td_left"><b>Field Description</b></th>
				998	</tr><tr>
				999	<td>0-1</td><td>constant "1"</td>
				1000	<td class="td_left">These two bits must be the value 1 which identifies
				1001	this as an instruction of format 1.</td>
				1002	</td>
				1003	</tr><tr>
				1004	<td>2-7</td><td><a href="#opcodes">opcode</a></td>
				1005	<td class="td_left">Specifies the opcode of the instruction. Note that
				1006	the maximum opcode value si 63.</td>
				1007	</tr><tr>
				1008	<td>8-19</td><td><a href="#unsigned">unsigned</a></td>
				1009	<td class="td_left">Specifies the slot number of the type for this
				1010	instruction. Maximum slot number is 2<sup>12</sup>-1=4095.</td>
				1011	</tr><tr>
				1012	<td>20-31</td><td><a href="#unsigned">unsigned</a></td>
				1013	<td class="td_left">Specifies the slot number of the value for the
				1014	first operand. Maximum slot number is 2<sup>12</sup>-1=4095. Note
				1015	that the value 2<sup>12</sup>-1 denotes zero operands.</td>
				1016	</tr>
				1017	</table>
				1018	<p><b>Instruction Format 2</b></p>
				1019	<p>This format encodes the opcode, type and two operands into a single
				1020	<a href="#uint32_vbr">uint32_vbr</a> as follows:</p>
				1021	<table>
				1022	<tr>
				1023	<th><b>Bits</b></th>
				1024	<th><b>Type</b></th>
				1025	<th class="td_left"><b>Field Description</b></th>
				1026	</tr><tr>
				1027	<td>0-1</td><td>constant "2"</td>
				1028	<td class="td_left">These two bits must be the value 2 which identifies
				1029	this as an instruction of format 2.</td>
				1030	</td>
				1031	</tr><tr>
				1032	<td>2-7</td><td><a href="#opcodes">opcode</a></td>
				1033	<td class="td_left">Specifies the opcode of the instruction. Note that
				1034	the maximum opcode value si 63.</td>
				1035	</tr><tr>
				1036	<td>8-15</td><td><a href="#unsigned">unsigned</a></td>
				1037	<td class="td_left">Specifies the slot number of the type for this
				1038	instruction. Maximum slot number is 2<sup>8</sup>-1=255.</td>
				1039	</tr><tr>
				1040	<td>16-23</td><td><a href="#unsigned">unsigned</a></td>
				1041	<td class="td_left">Specifies the slot number of the value for the
				1042	first operand. Maximum slot number is 2<sup>8</sup>-1=255.</td>
				1043	</tr><tr>
				1044	<td>24-31</td><td><a href="#unsigned">unsigned</a></td>
				1045	<td class="td_left">Specifies the slot number of the value for the
				1046	second operand. Maximum slot number is 2<sup>8</sup>-1=255.</td>
				1047	</tr>
				1048	</table>
				1049	<p><b>Instruction Format 3</b></p>
				1050	<p>This format encodes the opcode, type and three operands into a single
				1051	<a href="#uint32_vbr">uint32_vbr</a> as follows:</p>
				1052	<table>
				1053	<tr>
				1054	<th><b>Bits</b></th>
				1055	<th><b>Type</b></th>
				1056	<th class="td_left"><b>Field Description</b></th>
				1057	</tr><tr>
				1058	<td>0-1</td><td>constant "3"</td>
				1059	<td class="td_left">These two bits must be the value 3 which identifies
				1060	this as an instruction of format 3.</td>
				1061	</td>
				1062	</tr><tr>
				1063	<td>2-7</td><td><a href="#opcodes">opcode</a></td>
				1064	<td class="td_left">Specifies the opcode of the instruction. Note that
				1065	the maximum opcode value si 63.</td>
				1066	</tr><tr>
				1067	<td>8-13</td><td><a href="#unsigned">unsigned</a></td>
				1068	<td class="td_left">Specifies the slot number of the type for this
				1069	instruction. Maximum slot number is 2<sup>6</sup>-1=63.</td>
				1070	</tr><tr>
				1071	<td>14-19</td><td><a href="#unsigned">unsigned</a></td>
				1072	<td class="td_left">Specifies the slot number of the value for the
				1073	first operand. Maximum slot number is 2<sup>6</sup>-1=63.</td>
				1074	</tr><tr>
				1075	<td>20-25</td><td><a href="#unsigned">unsigned</a></td>
				1076	<td class="td_left">Specifies the slot number of the value for the
				1077	second operand. Maximum slot number is 2<sup>6</sup>-1=63.</td>
				1078	</tr><tr>
				1079	<td>26-31</td><td><a href="#unsigned">unsigned</a></td>
				1080	<td class="td_left">Specifies the slot number of the value for the
				1081	third operand. Maximum slot number is 2<sup>6</sup>-1=63.</td>
				1082	</tr>
				1083	</table>
				1084	</div>
				1085
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	1086	<!-- _______________________________________________________________________ -->
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1087	<div class="doc_subsection"><a name="symtab">Symbol Table</a> </div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	1088	<div class="doc_text">
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1089	<p>A symbol table can be put out in conjunction with a module or a function.
				1090	A symbol table is a list of type planes. Each type plane starts with the number
				1091	of entries in the plane and the type plane's slot number (so the type can be
				1092	looked up in the global type pool). For each entry in a type plane, the slot
				1093	number of the value and the name associated with that value are written. The
				1094	format is given in the table below. </p>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	1095	<table>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1096	<tr>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1097	<th><b>Type</b></th>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1098	<th class="td_left"><b>Field Description</b></th>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1099	</tr><tr>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	1100	<td><a href="#unsigned">unsigned</a></td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1101	<td class="td_left">Symbol Table Identifier (0x13)</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1102	</tr><tr>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	1103	<td><a href="#unsigned">unsigned</a></td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1104	<td class="td_left">Size in bytes of the symbol table block.</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1105	</tr><tr>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	1106	<td><a href="#uint32_vbr">uint32_vbr</a></td>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1107	<td class="td_left">Number of entries in type plane</td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1108	</tr><tr>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	1109	<td><a href="#symtab_entry">symtab_entry</a></td>
				1110	<td class="td_left">Provides the slot number of the type and its name.
				1111	<sup>1</sup></td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1112	</tr><tr>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	1113	<td><a href="#symtab_plane">symtab_plane</a></td>
				1114	<td class="td_left">A type plane containing value slot number and name
				1115	for all values of the same type.<sup>1</sup></td>
Reid Spencer	b39021b	2004-05-23 17:05:09 +0000	[diff] [blame]	1116	</tr>
				1117	</table>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	1118	Notes:
				1119	<ol>
Reid Spencer	2cc3615	2004-07-05 19:04:27 +0000	[diff] [blame]	1120	<li>Repeated field.</li>
				1121	</ol>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	1122	</div>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	1123
				1124	<!-- _______________________________________________________________________ -->
				1125	<div class="doc_subsubsection"> <a name="symtab_plane">Symbol Table Plane</a>
				1126	</div>
				1127	<div class="doc_text">
				1128	<p>A symbol table plane provides the symbol table entries for all values of
				1129	a common type. The encoding is given in the following table:</p>
				1130	<table>
				1131	<tr>
				1132	<th><b>Type</b></th>
				1133	<th class="td_left"><b>Field Description</b></th>
				1134	</tr><tr>
				1135	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				1136	<td class="td_left">Number of entries in this plane.</td>
				1137	</tr><tr>
				1138	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				1139	<td class="td_left">Slot number of type for this plane.</td>
				1140	</tr><tr>
				1141	<td><a href="#symtab_entry">symtab_entry</a></td>
				1142	<td class="td_left">The symbol table entries for this plane (repeated).</td>
				1143	</tr>
				1144	</table>
				1145	</div>
				1146
				1147	<!-- _______________________________________________________________________ -->
				1148	<div class="doc_subsubsection"> <a name="symtab_entry">Symbol Table Entry</a>
				1149	</div>
				1150	<div class="doc_text">
				1151	<p>A symbol table entry provides the assocation between a type or value's
				1152	slot number and the name given to that type or value. The format is given
				1153	in the following table:</p>
				1154	<table>
				1155	<tr>
				1156	<th><b>Type</b></th>
				1157	<th class="td_left"><b>Field Description</b></th>
				1158	</tr><tr>
				1159	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				1160	<td class="td_left">Slot number of the type or value being given a name.
				1161	</td>
				1162	</tr><tr>
				1163	<td><a href="#uint32_vbr">uint32_vbr</a></td>
				1164	<td class="td_left">Length of the character array that follows.</td>
				1165	</tr><tr>
				1166	<td><a href="#char">char</a></td>
				1167	<td class="td_left">The characters of the name (repeated).</td>
				1168	</tr>
				1169	</table>
				1170	</div>
				1171
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	1172	<!-- *********************************************************************** -->
				1173	<div class="doc_section"> <a name="versiondiffs">Version Differences</a> </div>
				1174	<!-- *********************************************************************** -->
				1175	<div class="doc_text">
				1176	<p>This section describes the differences in the Bytecode Format across LLVM
				1177	versions. The versions are listed in reverse order because it assumes the
				1178	current version is as documented in the previous sections. Each section here
Chris Lattner	1cc070c	2004-07-05 18:05:48 +0000	[diff] [blame]	1179	describes the differences between that version and the one that <i>follows</i>.
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	1180	</p>
				1181	</div>
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	1182
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	1183	<!-- _______________________________________________________________________ -->
				1184	<div class="doc_subsection">
				1185	<a name="vers12">Version 1.2 Differences From 1.3</a></div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1186	<!-- _______________________________________________________________________ -->
				1187	<div class="doc_subsubsection">Type Derives From Value</div>
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	1188	<div class="doc_text">
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1189	<p>In version 1.2, the Type class in the LLVM IR derives from the Value class.
				1190	This is not the case in version 1.3. Consequently, in version 1.2 the notion
				1191	of a "Type Type" was used to write out values that were Types. The types
				1192	always occuped plane 12 (corresponding to the TypeTyID) of any type planed
				1193	set of values. In 1.3 this representation is not convenient because the
				1194	TypeTyID (12) is not present and its value is now used for LabelTyID.
				1195	Consequently, the data structures written that involve types do so by writing
				1196	all the types first and then each of the value planes according to those
				1197	types. In version 1.2, the types would have been written intermingled with
				1198	the values.</p>
				1199	</div>
				1200
				1201	<!-- _______________________________________________________________________ -->
				1202	<div class="doc_subsubsection">Restricted getelementptr Types</a></div>
				1203	<div class="doc_text">
				1204	<p>In version 1.2, the getelementptr instruction required a ubyte type index
				1205	for accessing a structure field and a long type index for accessing an array
				1206	element. Consequently, it was only possible to access structures of 255 or
				1207	fewer elements. Starting in version 1.3, this restriction was lifted.
Chris Lattner	7c66ab3	2004-07-05 17:55:28 +0000	[diff] [blame]	1208	Structures must now be indexed with uint constants. Arrays may now be
				1209	indexed with int, uint, long, or ulong typed values.
				1210	The consequence of this was that the bytecode format had to
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1211	change in order to accommodate the larger range of structure indices.</p>
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	1212	</div>
				1213
				1214	<!-- _______________________________________________________________________ -->
				1215	<div class="doc_subsection">
				1216	<a name="vers11">Version 1.1 Differences From 1.2 </a></div>
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1217	<!-- _______________________________________________________________________ -->
				1218	<div class="doc_subsubsection">Explicit Primitive Zeros</div>
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	1219	<div class="doc_text">
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1220	<p>In version 1.1, the zero value for primitives was explicitly encoded into
				1221	the bytecode format. Since these zero values are constant values in the
				1222	LLVM IR and never change, there is no reason to explicitly encode them. This
				1223	explicit encoding was removed in version 1.2.</p>
				1224	</div>
				1225
				1226	<!-- _______________________________________________________________________ -->
				1227	<div class="doc_subsubsection">Inconsistent Module Global Info</div>
				1228	<div class="doc_text">
				1229	<p>In version 1.1, the Module Global Info block was not aligned causing the
				1230	next block to be read in on an unaligned boundary. This problem was corrected
				1231	in version 1.2.</p>
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	1232	</div>
				1233
				1234	<!-- _______________________________________________________________________ -->
				1235	<div class="doc_subsection">
Reid Spencer	51f31e0	2004-07-05 22:28:02 +0000	[diff] [blame^]	1236	<a name="vers10">Version 1.0 Differences From 1.1</a></div>
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	1237	<div class="doc_text">
Reid Spencer	1ab929c	2004-07-05 08:18:07 +0000	[diff] [blame]	1238	<p>None. Version 1.0 and 1.1 bytecode formats are identical.</p>
Reid Spencer	7c76d33	2004-06-08 07:41:41 +0000	[diff] [blame]	1239	</div>
Reid Spencer	5002661	2004-05-22 02:28:36 +0000	[diff] [blame]	1240
				1241	<!-- *********************************************************************** -->
				1242	<hr>
				1243	<address>
				1244	<a href="http://jigsaw.w3.org/css-validator/check/referer"><img
				1245	src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
				1246	<a href="http://validator.w3.org/check/referer"><img
				1247	src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
				1248
				1249	<a href="mailto:rspencer@x10sys.com">Reid Spencer</a> and
				1250	<a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
				1251	<a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
				1252	Last modified: $Date$
				1253	</address>
				1254	</body>
				1255	</html>
				1256	<!-- vim: sw=2
				1257	-->