Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / llvm / 2cc361555bc4f9da77eea67e9aa0728c94164b13 / . / docs / BytecodeFormat.html
blob: d9657750ceaca4f35f07f5b7465e3bba4e13f568 [file] [log] [blame]
Reid Spencer50026612004-05-22 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 Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]7 <style type="text/css">
Reid Spencer2cc36152004-07-05 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 Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]9 TH { border: 2px solid gray; font-weight: bold; font-size: 105%; }
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]10 TABLE { text-align: center; border: 2px solid black;
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]11 border-collapse: collapse; margin-top: 1em; margin-left: 1em; margin-right: 1em; margin-bottom: 1em; }
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]12 .td_left { border: 2px solid gray; text-align: left; }
Reid Spencer50026612004-05-22 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 Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]19 <li><a href="#concepts">Concepts</a>
Reid Spencer50026612004-05-22 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 Spencer1ab929c2004-07-05 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>
29 <li><a href="#general">General Layout</a>
30 <ol>
31 <li><a href="#structure">Structure</a></li>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]32 </ol>
Reid Spencer6f1d6992004-05-23 17:12:45 +0000[diff] [blame]33 </li>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]34 <li><a href="#details">Detailed Layout</a>
35 <ol>
36 <li><a href="#notation">Notation</a></li>
37 <li><a href="#blocktypes">Blocks Types</a></li>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]38 <li><a href="#signature">Signature Block</a></li>
39 <li><a href="#module">Module Block</a></li>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]40 <li><a href="#globaltypes">Global Type Pool</a></li>
41 <li><a href="#globalinfo">Module Info Block</a></li>
42 <li><a href="#constantpool">Global Constant Pool</a></li>
43 <li><a href="#functiondefs">Function Definition</a></li>
44 <li><a href="#compactiontable">Compaction Table</a></li>
45 <li><a href="#instructionlist">Instruction List</a></li>
46 <li><a href="#symtab">Symbol Table</a></li>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]47 </ol>
48 </li>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]49 <li><a href="#versiondiffs">Version Differences</a>
50 <ol>
51 <li><a href="#vers12">Version 1.2 Differences From 1.3</a></li>
52 <li><a href="#vers11">Version 1.1 Differences From 1.2</a></li>
53 <li><a href="#vers10">Version 1.0 Differences From 1.1</a></li>
54 </ol>
55 </li>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]56</ol>
Chris Lattner8dabb502004-05-25 17:44:58 +0000[diff] [blame]57<div class="doc_author">
58<p>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a>
59</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]60</div>
Reid Spencerc0a2af12004-06-05 14:18:02 +0000[diff] [blame]61<div class="doc_warning">
62 <p>Warning: This is a work in progress.</p>
63</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]64
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]65<!-- *********************************************************************** -->
66<div class="doc_section"> <a name="abstract">Abstract </a></div>
67<!-- *********************************************************************** -->
68<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]69<p>This document describes the LLVM bytecode file format as of version 1.3.
70It specifies the binary encoding rules of the bytecode file format
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]71so that equivalent systems can encode bytecode files correctly. The LLVM
72bytecode representation is used to store the intermediate representation on
73disk in compacted form.
74</p>
75</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]76
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]77<!-- *********************************************************************** -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]78<div class="doc_section"> <a name="concepts">Concepts</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]79<!-- *********************************************************************** -->
80<div class="doc_text">
81<p>This section describes the general concepts of the bytecode file format
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]82without getting into bit and byte level specifics. Note that the LLVM bytecode
83format may change in the future, but will always be backwards compatible with
84older formats. This document only describes the most current version of the
85bytecode format.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]86</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]87
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]88<!-- _______________________________________________________________________ -->
89<div class="doc_subsection"><a name="blocks">Blocks</a> </div>
90<div class="doc_text">
91<p>LLVM bytecode files consist simply of a sequence of blocks of bytes.
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]92Each block begins with an header of two unsigned integers. The first value
93identifies the type of block and the second value provides the size of the
94block in bytes. The block identifier is used because it is possible for entire
95blocks to be omitted from the file if they are empty. The block identifier helps
96the reader determine which kind of block is next in the file. Note that blocks
97can be nested within other blocks.</p>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]98<p> All blocks are variable length, and the block header specifies the size of
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]99the block. All blocks begin on a byte index that is aligned to an even 32-bit
100boundary. That is, the first block is 32-bit aligned because it starts at offset
1010. Each block is padded with zero fill bytes to ensure that the next block also
102starts on a 32-bit boundary.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]103</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]104
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]105<!-- _______________________________________________________________________ -->
106<div class="doc_subsection"><a name="lists">Lists</a> </div>
107<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]108 <p>LLVM Bytecode blocks often contain lists of things of a similar type. For
109 example, a function contains a list of instructions and a function type
110 contains a list of argument types. There are two basic types of lists:
111 length lists, and null terminated lists, as described here:</p>
112 <ul>
113 <li><b>Length Lists</b>. Length lists are simply preceded by the number
114 of items in the list. The bytecode reader will read the count first and
115 then iterate that many times to read in the list contents.</li>
116 <li><b>Null Terminated Lists</b>. For some lists, the number of elements
117 in the list is not readily available at the time of writing the bytecode.
118 In these cases, the list is terminated by some null value. What constitutes
119 a null value differs, but it almost always boils down to a zero value.</li>
120 </ul>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]121</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]122
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]123<!-- _______________________________________________________________________ -->
124<div class="doc_subsection"><a name="fields">Fields</a> </div>
125<div class="doc_text">
126<p>Fields are units of information that LLVM knows how to write atomically.
127Most fields have a uniform length or some kind of length indication built into
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]128their encoding. For example, a constant string (array of bytes) is
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]129written simply as the length followed by the characters. Although this is
130similar to a list, constant strings are treated atomically and are thus
131fields.</p>
132<p>Fields use a condensed bit format specific to the type of information
133they must contain. As few bits as possible are written for each field. The
134sections that follow will provide the details on how these fields are
135written and how the bits are to be interpreted.</p>
136</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]137
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]138<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]139<div class="doc_subsection"><a name="align">Alignment</a> </div>
Reid Spencer7aa940d2004-05-25 15:47:57 +0000[diff] [blame]140<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]141 <p>To support cross-platform differences, the bytecode file is aligned on
142 certain boundaries. This means that a small amount of padding (at most 3
143 bytes) will be added to ensure that the next entry is aligned to a 32-bit
144 boundary.</p>
Chris Lattner8dabb502004-05-25 17:44:58 +0000[diff] [blame]145</div>
146
Reid Spencer7aa940d2004-05-25 15:47:57 +0000[diff] [blame]147<!-- _______________________________________________________________________ -->
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]148<div class="doc_subsection"><a name="encoding">Encoding Primitives</a> </div>
149<div class="doc_text">
150<p>Each field that can be put out is encoded into the file using a small set
151of primitives. The rules for these primitives are described below.</p>
152<h3>Variable Bit Rate Encoding</h3>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]153<p>Most of the values written to LLVM bytecode files are small integers. To
154minimize the number of bytes written for these quantities, an encoding
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]155scheme similar to UTF-8 is used to write integer data. The scheme is known as
156variable bit rate (vbr) encoding. In this encoding, the high bit of each
157byte is used to indicate if more bytes follow. If (byte &amp; 0x80) is non-zero
158in any given byte, it means there is another byte immediately following that
159also contributes to the value. For the final byte (byte &amp; 0x80) is false
160(the high bit is not set). In each byte only the low seven bits contribute to
161the value. Consequently 32-bit quantities can take from one to <em>five</em>
162bytes to encode. In general, smaller quantities will encode in fewer bytes,
163as follows:</p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]164<table>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]165 <tr>
166 <th>Byte #</th>
167 <th>Significant Bits</th>
168 <th>Maximum Value</th>
169 </tr>
170 <tr><td>1</td><td>0-6</td><td>127</td></tr>
171 <tr><td>2</td><td>7-13</td><td>16,383</td></tr>
172 <tr><td>3</td><td>14-20</td><td>2,097,151</td></tr>
173 <tr><td>4</td><td>21-27</td><td>268,435,455</td></tr>
174 <tr><td>5</td><td>28-34</td><td>34,359,738,367</td></tr>
175 <tr><td>6</td><td>35-41</td><td>4,398,046,511,103</td></tr>
176 <tr><td>7</td><td>42-48</td><td>562,949,953,421,311</td></tr>
177 <tr><td>8</td><td>49-55</td><td>72,057,594,037,927,935</td></tr>
178 <tr><td>9</td><td>56-62</td><td>9,223,372,036,854,775,807</td></tr>
179 <tr><td>10</td><td>63-69</td><td>1,180,591,620,717,411,303,423</td></tr>
180</table>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]181<p>Note that in practice, the tenth byte could only encode bit 63
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]182since the maximum quantity to use this encoding is a 64-bit integer.</p>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]183
184<p><em>Signed</em> VBR values are encoded with the standard vbr encoding, but
185with the sign bit as the low order bit instead of the high order bit. This
186allows small negative quantities to be encoded efficiently. For example, -3
187is encoded as "((3 &lt;&lt; 1) | 1)" and 3 is encoded as "(3 &lt;&lt; 1) |
1880)", emitted with the standard vbr encoding above.</p>
189
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]190<p>The table below defines the encoding rules for type names used in the
191descriptions of blocks and fields in the next section. Any type name with
192the suffix <em>_vbr</em> indicate a quantity that is encoded using
193variable bit rate encoding as described above.</p>
194<table class="doc_table" >
195 <tr>
196 <th><b>Type</b></th>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]197 <th class="td_left"><b>Rule</b></th>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]198 </tr>
199 <tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]200 <td><a name="unsigned">unsigned</a></td>
201 <td class="td_left">A 32-bit unsigned integer that always occupies four
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]202 consecutive bytes. The unsigned integer is encoded using LSB first
203 ordering. That is bits 2<sup>0</sup> through 2<sup>7</sup> are in the
204 byte with the lowest file offset (little endian).</td>
205 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]206 <td><a name="uint_vbr">uint_vbr</a></td>
207 <td class="td_left">A 32-bit unsigned integer that occupies from one to five
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]208 bytes using variable bit rate encoding.</td>
209 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]210 <td><a name="uint64_vbr">uint64_vbr</a></td>
211 <td class="td_left">A 64-bit unsigned integer that occupies from one to ten
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]212 bytes using variable bit rate encoding.</td>
213 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]214 <td><a name="int64_vbr">int64_vbr</a></td>
215 <td class="td_left">A 64-bit signed integer that occupies from one to ten
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]216 bytes using the signed variable bit rate encoding.</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]217 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]218 <td><a name="char">char</a></td>
219 <td class="td_left">A single unsigned character encoded into one byte</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]220 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]221 <td><a name="bit">bit</a></td>
222 <td class="td_left">A single bit within some larger integer field.</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]223 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]224 <td><a name="string">string</a></td>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]225 <td class="td_left">A uint_vbr indicating the type of the character string
226 which also includes its length, immediately followed by the characters of
227 the string. There is no terminating null byte in the string.</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]228 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]229 <td><a name="data">data</a></td>
230 <td class="td_left">An arbitrarily long segment of data to which no
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]231 interpretation is implied. This is used for float, double, and constant
232 initializers.</td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]233 </tr><tr>
234 <td><a name="block">block</a></td>
235 <td class="td_left">A block of data that is logically related. A block
236 begins with an <a href="#unsigned">unsigned</a> that provides the block
237 identifier (constant value) and an <a href="#unsigned">unsigned</a> that
238 provides the length of the block. Blocks may compose other blocks.
239 </td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]240 </tr>
241</table>
242</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]243
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]244<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]245<div class="doc_subsection"><a name="slots">Slots</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]246<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]247<p>The bytecode format uses the notion of a "slot" to reference Types and
248Values. Since the bytecode file is a <em>direct</em> representation of LLVM's
249intermediate representation, there is a need to represent pointers in the file.
250Slots are used for this purpose. For example, if one has the following assembly:
251</p>
252<div class="doc_code">
253 %MyType = type { int, sbyte }<br>
254 %MyVar = external global %MyType
255</div>
256<p>there are two definitions. The definition of <tt>%MyVar</tt> uses
257<tt>%MyType</tt>. In the C++ IR this linkage between <tt>%MyVar</tt> and
258<tt>%MyType</tt> is
259explicit through the use of C++ pointers. In bytecode, however, there's no
260ability to store memory addresses. Instead, we compute and write out slot
261numbers for every Type and Value written to the file.</p>
262<p>A slot number is simply an unsigned 32-bit integer encoded in the variable
263bit rate scheme (see <a href="#encoding">encoding</a>). This ensures that
264low slot numbers are encoded in one byte. Through various bits of magic LLVM
265attempts to always keep the slot numbers low. The first attempt is to associate
266slot numbers with their "type plane". That is, Values of the same type are
267written to the bytecode file in a list (sequentially). Their order in that list
268determines their slot number. This means that slot #1 doesn't mean anything
269unless you also specify for which type you want slot #1. Types are handled
270specially and are always written to the file first (in the
271<a href="#globaltypes">Global Type Pool</a>) and
272in such a way that both forward and backward references of the types can often be
273resolved with a single pass through the type pool. </p>
274<p>Slot numbers are also kept small by rearranging their order. Because of the
275structure of LLVM, certain values are much more likely to be used frequently
276in the body of a function. For this reason, a compaction table is provided in
277the body of a function if its use would make the function body smaller.
278Suppose you have a function body that uses just the types "int*" and "{double}"
279but uses them thousands of time. Its worthwhile to ensure that the slot number
280for these types are low so they can be encoded in a single byte (via vbr).
281This is exactly what the compaction table does.</p>
282</div>
283
284<!-- *********************************************************************** -->
285<div class="doc_section"> <a name="general">General Layout</a> </div>
286<!-- *********************************************************************** -->
287<div class="doc_text">
288 <p>This section provides the general layout of the LLVM bytecode file format.
289 The detailed layout can be found in the <a href="#details">next section</a>.
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]290</p>
291</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]292
293<!-- _______________________________________________________________________ -->
294<div class="doc_subsection"><a name="structure">Structure</a> </div>
295<div class="doc_text">
296<p>The bytecode file format requires blocks to be in a certain order and
297nested in a particular way so that an LLVM module can be constructed
298efficiently from the contents of the file. This ordering defines a general
299structure for bytecode files as shown below. The table below shows the order
300in which all block types may appear. Please note that some of the blocks are
301optional and some may be repeated. The structure is fairly loose because
302optional blocks, if empty, are completely omitted from the file.
303</p>
304<table>
305 <tr>
306 <th>ID</th>
307 <th>Parent</th>
308 <th>Optional?</th>
309 <th>Repeated?</th>
310 <th>Level</th>
311 <th>Block Type</th>
312 </tr>
313 <tr><td>N/A</td><td>File</td><td>No</td><td>No</td><td>0</td>
314 <td class="td_left"><a href="#signature">Signature</a></td>
315 </tr>
316 <tr><td>0x01</td><td>File</td><td>No</td><td>No</td><td>0</td>
317 <td class="td_left"><a href="#module">Module</a></td>
318 </tr>
319 <tr><td>0x15</td><td>Module</td><td>No</td><td>No</td><td>1</td>
320 <td class="td_left">&nbsp;&nbsp;&nbsp;
321 <a href="#globaltypes">Global Type Pool</a></td>
322 </tr>
323 <tr><td>0x14</td><td>Module</td><td>No</td><td>No</td><td>1</td>
324 <td class="td_left">&nbsp;&nbsp;&nbsp;
325 <a href="#globalinfo">Module Globals Info</a></td>
326 </tr>
327 <tr><td>0x12</td><td>Module</td><td>Yes</td><td>No</td><td>1</td>
328 <td class="td_left">&nbsp;&nbsp;&nbsp;
329 <a href="#constantpool">Module Constant Pool</a></td>
330 </tr>
331 <tr><td>0x11</td><td>Module</td><td>Yes</td><td>Yes</td><td>1</td>
332 <td class="td_left">&nbsp;&nbsp;&nbsp;
333 <a href="#functiondefs">Function Definitions</a></td>
334 </tr>
335 <tr><td>0x12</td><td>Function</td><td>Yes</td><td>No</td><td>2</td>
336 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
337 <a href="#constantpool">Function Constant Pool</a></td>
338 </tr>
339 <tr><td>0x33</td><td>Function</td><td>Yes</td><td>No</td><td>2</td>
340 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
341 <a href="#compactiontable">Compaction Table</a></td>
342 </tr>
343 <tr><td>0x32</td><td>Function</td><td>No</td><td>No</td><td>2</td>
344 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
345 <a href="#instructionlist">Instruction List</a></td>
346 </tr>
347 <tr><td>0x13</td><td>Function</td><td>Yes</td><td>No</td><td>2</td>
348 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;
349 <a href="#symboltable">Function Symbol Table</a></td>
350 </tr>
351 <tr><td>0x13</td><td>Module</td><td>Yes</td><td>No</td><td>1</td>
352 <td class="td_left">&nbsp;&nbsp;&nbsp;
353 <a href="#symboltable">Module Symbol Table</a></td>
354 </tr>
355</table>
356<p>Use the links in the table or see <a href="#blocktypes">Block Types</a> for
357details about the contents of each of the block types.</p>
358</div>
359
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]360<!-- *********************************************************************** -->
361<div class="doc_section"> <a name="details">Detailed Layout</a> </div>
362<!-- *********************************************************************** -->
363<div class="doc_text">
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]364<p>This section provides the detailed layout of the LLVM bytecode file format.
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]365</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]366</div>
367<!-- _______________________________________________________________________ -->
368<div class="doc_subsection"><a name="notation">Notation</a></div>
369<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]370<p>The descriptions of the bytecode format that follow describe the order, type
371and bit fields in detail. These descriptions are provided in tabular form.
372Each table has four columns that specify:</p>
373<ol>
374 <li><b>Byte(s)</b>: The offset in bytes of the field from the start of
375 its container (block, list, other field).</li>
376 <li><b>Bit(s)</b>: The offset in bits of the field from the start of
377 the byte field. Bits are always little endian. That is, bit addresses with
378 smaller values have smaller address (i.e. 2<sup>0</sup> is at bit 0,
379 2<sup>1</sup> at 1, etc.)
380 </li>
381 <li><b>Align?</b>: Indicates if this field is aligned to 32 bits or not.
382 This indicates where the <em>next</em> field starts, always on a 32 bit
383 boundary.</li>
384 <li><b>Type</b>: The basic type of information contained in the field.</li>
385 <li><b>Description</b>: Describes the contents of the field.</li>
386</ol>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]387</div>
388<!-- _______________________________________________________________________ -->
389<div class="doc_subsection"><a name="blocktypes">Block Types</a></div>
390<div class="doc_text">
391 <p>The bytecode format encodes the intermediate representation into groups
392 of bytes known as blocks. The blocks are written sequentially to the file in
393 the following order:</p>
394<ol>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]395 <li><a href="#signature">Signature</a>: This contains the file signature
396 (magic number) that identifies the file as LLVM bytecode and the bytecode
397 version number.</li>
398 <li><a href="#module">Module Block</a>: This is the top level block in a
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]399 bytecode file. It contains all the other blocks.</li>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]400 <li><a href="#gtypepool">Global Type Pool</a>: This block contains all the
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]401 global (module) level types.</li>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]402 <li><a href="#modinfo">Module Info</a>: This block contains the types of the
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]403 global variables and functions in the module as well as the constant
404 initializers for the global variables</li>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]405 <li><a href="#constants">Constants</a>: This block contains all the global
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]406 constants except function arguments, global values and constant strings.</li>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]407 <li><a href="#functions">Functions</a>: One function block is written for
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]408 each function in the module. </li>
Tanya Lattner61f1d2b2004-06-21 23:29:40 +0000[diff] [blame]409 <li><a href="#symtab">Symbol Table</a>: The module level symbol table that
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]410 provides names for the various other entries in the file is the final block
411 written.</li>
412</ol>
413</div>
414<!-- _______________________________________________________________________ -->
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]415<div class="doc_subsection"><a name="signature">Signature Block</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]416<div class="doc_text">
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]417<p>The signature occurs in every LLVM bytecode file and is always first.
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]418It simply provides a few bytes of data to identify the file as being an LLVM
419bytecode file. This block is always four bytes in length and differs from the
420other blocks because there is no identifier and no block length at the start
421of the block. Essentially, this block is just the "magic number" for the file.
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]422<table>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]423 <tr>
Reid Spencer939290f2004-05-22 05:56:41 +0000[diff] [blame]424 <th><b>Type</b></th>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]425 <th class="td_left"><b>Field Description</b></th>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]426 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]427 <td><a href="#char">char</a></td>
428 <td class="td_left">Constant "l" (0x6C)</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]429 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]430 <td><a href="#char">char</a></td>
431 <td class="td_left">Constant "l" (0x6C)</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]432 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]433 <td><a href="#char">char</a></td>
434 <td class="td_left">Constant "v" (0x76)</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]435 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]436 <td><a href="#char">char</a></td>
437 <td class="td_left">Constant "m" (0x6D)</td>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]438 </tr>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]439</table>
440</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]441
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]442<!-- _______________________________________________________________________ -->
443<div class="doc_subsection"><a name="module">Module Block</a> </div>
444<div class="doc_text">
445<p>The module block contains a small pre-amble and all the other blocks in
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]446the file. The table below shows the structure of the module block. Note that it
447only provides the module identifier, size of the module block, and the format
448information. Everything else is contained in other blocks, described in other
449sections.</p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]450<table>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]451 <tr>
452 <th><b>Type</b></th>
453 <th class="td_left"><b>Field Description</b></th>
454 </tr><tr>
455 <td><a href="#unsigned">unsigned</a></td>
456 <td class="td_left">Module Identifier (0x01)</td>
457 </tr><tr>
458 <td><a href="#unsigned">unsigned</a></td>
459 <td class="td_left">Size of the module block in bytes</td>
460 </tr><tr>
461 <td><a href="#uint32_vbr">uint32_vbr</a></td>
462 <td class="td_left"><a href="#format">Format Information</a></td>
463 </tr><tr>
464 <td><a href="#block">block</a></td>
465 <td class="td_left"><a href="#globaltypes">Global Type Pool</a></td>
466 </tr><tr>
467 <td><a href="#block">block</a></td>
468 <td class="td_left"><a href="#globalinfo">Module Globals Info</a></td>
469 </tr><tr>
470 <td><a href="#block">block</a></td>
471 <td class="td_left"><a href="#constantpool">Module Constant Pool</a></td>
472 </tr><tr>
473 <td><a href="#block">block</a></td>
474 <td class="td_left"><a href="#functiondefs">Function Definitions</a></td>
475 </tr><tr>
476 <td><a href="#block">block</a></td>
477 <td class="td_left"><a href="#symboltable">Module Symbol Table</a></td>
478 </tr>
479</table>
480</div>
481
482<!-- _______________________________________________________________________ -->
483<div class="doc_subsubsection"><a name="format">Format Information</a></div>
484<div class="doc_text">
485<p>The format information field is encoded into a 32-bit vbr-encoded unsigned
486integer as shown in the following table.</p>
487<table>
488 <tr>
489 <th><b>Bit(s)</b></th>
490 <th><b>Type</b></th>
491 <th class="td_left"><b>Description</b></th>
492 </tr><tr>
493 <td>0</td><td>bit</td>
494 <td class="td_left">Big Endian?</td>
495 </tr><tr>
496 <td>1</td><td>bit</td>
497 <td class="td_left">Pointers Are 64-bit?</td>
498 </tr><tr>
499 <td>2</td><td>bit</td>
500 <td class="td_left">Has No Endianess?</td>
501 </tr><tr>
502 <td>3</td><td>bit</td>
503 <td class="td_left">Has No Pointer Size?</td>
504 </tr><tr>
505 <td>4-31</td><td>bit</td>
506 <td class="td_left">Bytecode Format Version</td>
507 </tr>
508</table>
509<p>
510Of particular note, the bytecode format number is simply a 28-bit
511monotonically increase integer that identifies the version of the bytecode
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]512format (which is not directly related to the LLVM release number). The
513bytecode versions defined so far are (note that this document only describes
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]514the latest version, 1.3):</p>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]515<ul>
516<li>#0: LLVM 1.0 &amp; 1.1</li>
517<li>#1: LLVM 1.2</li>
518<li>#2: LLVM 1.3</li>
519</ul>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]520<p>Note that we plan to eventually expand the target description capabilities
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]521of bytecode files to <a href="http://llvm.cs.uiuc.edu/PR263">target triples</a>.
522</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]523</div>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]524
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]525<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]526<div class="doc_subsection"><a name="globaltypes">Global Type Pool</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]527<div class="doc_text">
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]528<p>The global type pool consists of type definitions. Their order of appearance
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]529in the file determines their slot number (0 based). Slot numbers are used to
530replace pointers in the intermediate representation. Each slot number uniquely
531identifies one entry in a type plane (a collection of values of the same type).
532Since all values have types and are associated with the order in which the type
533pool is written, the global type pool <em>must</em> be written as the first
534block of a module. If it is not, attempts to read the file will fail because
535both forward and backward type resolution will not be possible.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]536<p>The type pool is simply a list of type definitions, as shown in the table
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]537below.</p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]538<table>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]539 <tr>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]540 <th><b>Type</b></th>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]541 <th class="td_left"><b>Field Description</b></th>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]542 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]543 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]544 <td class="td_left">Type Pool Identifier (0x15)</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]545 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]546 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]547 <td class="td_left">Size in bytes of the type pool block.</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]548 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]549 <td><a href="#uint32_vbr">uint32_vbr</a></td>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]550 <td class="td_left">Number of type definitions that follow in the next
551 field.</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]552 </tr><tr>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]553 <td><a href="#type">type</a></td>
554 <td class="td_left">Each of the type definitions (see below)<sup>1</sup></td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]555 </tr>
556</table>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]557Notes:
558<ol>
559 <li>Repeated field.</li>
560</ol>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]561</div>
562<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]563<div class="doc_subsubsection"><a name="type">Type Definitions</a></div>
564<div class="doc_text">
565<p>Types in the type pool are defined using a different format for each
566basic type of type as given in the following sections.</p>
567<h3>Primitive Types</h3>
568<p>The primitive types encompass the basic integer and floating point types</p>
569<table>
570 <tr>
571 <th><b>Type</b></th>
572 <th class="td_left"><b>Description</b></th>
573 </tr><tr>
574 <td><a href="#uint32_vbr">uint32_vbr</td>
575 <td class="td_left">Type ID For The Primitive (1-11)<sup>1</sup></td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]576 </tr>
577</table>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]578Notes:
579<ol>
580 <li>See the definition of Type::TypeID in Type.h for the numeric equivalents
581 of the primitive type ids.</li>
582</ol>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]583<h3>Function Types</h3>
584<table>
585 <tr>
586 <th><b>Type</b></th>
587 <th class="td_left"><b>Description</b></th>
588 </tr><tr>
589 <td><a href="#uint32_vbr">uint32_vbr</td>
590 <td class="td_left">Type ID for function types (13)</td>
591 </tr><tr>
592 <td><a href="#uint32_vbr">uint32_vbr</td>
593 <td class="td_left">Slot number of function's return type.</td>
594 </tr><tr>
595 <td><a href="#uint32_vbr">uint32_vbr</td>
596 <td class="td_left">The number of arguments in the function.</td>
597 </tr><tr>
598 <td><a href="#uint32_vbr">uint32_vbr</td>
599 <td class="td_left">Slot number of each argument's type.<sup>1</sup></td>
600 </tr><tr>
601 <td><a href="#uint32_vbr">uint32_vbr</td>
602 <td class="td_left">Value 0 if this is a varargs function.<sup>2</sup></td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]603 </tr>
604</table>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]605Notes:
606<ol>
607 <li>Repeated field.</li>
608 <li>Optional field.</li>
609</ol>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]610<h3>Structure Types</h3>
611<table>
612 <tr>
613 <th><b>Type</b></th>
614 <th class="td_left"><b>Description</b></th>
615 </tr><tr>
616 <td><a href="#uint32_vbr">uint32_vbr</td>
617 <td class="td_left">Type ID for structure types (14)</td>
618 </tr><tr>
619 <td><a href="#uint32_vbr">uint32_vbr</td>
620 <td class="td_left">Slot number of each of the element's fields.<sup>1</sup></td>
621 </tr><tr>
622 <td><a href="#uint32_vbr">uint32_vbr</td>
623 <td class="td_left">Null Terminator (VoidTy type id)</td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]624 </tr>
625</table>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]626Notes:
627<ol>
628 <li>Repeatable field.</li>
629</ol>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]630<h3>Array Types</h3>
631<table>
632 <tr>
633 <th><b>Type</b></th>
634 <th class="td_left"><b>Description</b></th>
635 </tr><tr>
636 <td><a href="#uint32_vbr">uint32_vbr</td>
637 <td class="td_left">Type ID for Array Types (15)</td>
638 </tr><tr>
639 <td><a href="#uint32_vbr">uint32_vbr</td>
640 <td class="td_left">Slot number of array's element type.</td>
641 </tr><tr>
642 <td><a href="#uint32_vbr">uint32_vbr</td>
643 <td class="td_left">The number of elements in the array.</td>
644 </tr>
645</table>
646<h3>Pointer Types</h3>
647<table>
648 <tr>
649 <th><b>Type</b></th>
650 <th class="td_left"><b>Description</b></th>
651 </tr><tr>
652 <td><a href="#uint32_vbr">uint32_vbr</td>
653 <td class="td_left">Type ID For Pointer Types (16)</td>
654 </tr><tr>
655 <td><a href="#uint32_vbr">uint32_vbr</td>
656 <td class="td_left">Slot number of pointer's element type.</td>
657 </tr>
658</table>
659<h3>Opaque Types</h3>
660<table>
661 <tr>
662 <th><b>Type</b></th>
663 <th class="td_left"><b>Description</b></th>
664 </tr><tr>
665 <td><a href="#uint32_vbr">uint32_vbr</td>
666 <td class="td_left">Type ID For Opaque Types (17)</td>
667 </tr>
668</table>
669</div>
670<!-- _______________________________________________________________________ -->
671<div class="doc_subsection"><a name="globalinfo">Module Global Info</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]672<div class="doc_text">
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]673 <p>The module global info block contains the definitions of all global
674 variables including their initializers and the <em>declaration</em> of all
675 functions. The format is shown in the table below</p>
676 <table>
677 <tr>
678 <th><b>Type</b></th>
679 <th class="td_left"><b>Field Description</b></th>
680 </tr><tr>
681 <td><a href="#unsigned">unsigned</a></td>
682 <td class="td_left">Module global info identifier (0x14)</td>
683 </tr><tr>
684 <td><a href="#unsigned">unsigned</a></td>
685 <td class="td_left">Size in bytes of the module global info block.</td>
686 </tr><tr>
687 <td><a href="#globalvar">globalvar</a></td>
688 <td class="td_left">Definition of the global variable (see below).
689 <sup>1</sup>
690 </td>
691 </tr><tr>
692 <td><a href="#uint32_vbr">uint32_vbr</a></td>
693 <td class="td_left">Slot number of the global variable's constant
694 initializer.<sup>1,2</sup>
695 </td>
696 </tr><tr>
697 <td><a href="#uint32_vbr">uint32_vbr</a></td>
698 <td class="td_left">Zero. This terminates the list of global variables.
699 </td>
700 </tr><tr>
701 <td><a href="#uint32_vbr">uint32_vbr</a></td>
702 <td class="td_left">Type slot number of a function defined in this
703 bytecode file.<sup>3</sup>
704 </td>
705 </tr><tr>
706 <td><a href="#uint32_vbr">uint32_vbr</a></td>
707 <td class="td_left">Zero. This terminates the list of function
708 declarations.
709 </tr>
710 </table>
711 Notes:<ol>
712 <li>Both these fields are repeatable but in pairs.</li>
713 <li>Optional field.</li>
714 <li>Repeatable field.</li>
715 </ol>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]716</div>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]717
718<!-- _______________________________________________________________________ -->
719<div class="doc_subsubsection"><a name="globalvar">Global Variable Field</a>
720</div>
721<div class="doc_text">
722 <p>Global variables are written using a single
723 <a href="#uint32_vbr">uint32_vbr</a> that encodes information about the global
724 variable. The table below provides the bit layout of the value written for
725 each global variable.</p>
726 <table>
727 <tr>
728 <th><b>Bit(s)</b></th>
729 <th><b>Type</b></th>
730 <th class="td_left"><b>Description</b></th>
731 </tr><tr>
732 <td>0</td><td>bit</td>
733 <td class="td_left">Is constant?</td>
734 </tr><tr>
735 <td>1</td><td>bit</td>
736 <td class="td_left">Has initializer?<sup>1</sup></td>
737 </tr><tr>
738 <td>2-4</td><td>enumeration</td>
739 <td class="td_left">Linkage type: 0=External, 1=Weak, 2=Appending,
740 3=Internal, 4=LinkOnce</td>
741 </tr><tr>
742 <td>5-31</td><td>type slot</td>
743 <td class="td_left">Slot number of type for the global variable.</td>
744 </tr>
745 </table>
746 Notes:
747 <ol>
748 <li>This bit determines whether the constant initializer field follows
749 immediately after this field</li>
750 </ol>
751</div>
752
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]753<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]754<div class="doc_subsection"><a name="constantpool">Constant Pool</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]755<div class="doc_text">
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]756 <p>A constant pool defines as set of constant values. There are actually two
757 types of constant pool blocks: one for modules and one for functions. For
758 modules, the block begins with the constant strings encountered anywhere in
759 the module. For functions, the block begins with types only encountered in
760 the function. In both cases the header is identical. The tables the follow,
761 show the header, module constant pool preamble, function constant pool
762 preamble, and the part common to both function and module constant pools.</p>
763 <p><b>Common Block Header</b></p>
764 <table>
765 <tr>
766 <th><b>Type</b></th>
767 <th class="td_left"><b>Field Description</b></th>
768 </tr><tr>
769 <td><a href="#unsigned">unsigned</a></td>
770 <td class="td_left">Constant pool identifier (0x12)</td>
771 </tr>
772 </table>
773 <p><b>Module Constant Pool Preamble (constant strings)</b></p>
774 <table>
775 <tr>
776 <th><b>Type</b></th>
777 <th class="td_left"><b>Field Description</b></th>
778 </tr><tr>
779 <td><a href="#uint32_vbr">uint32_vbr</a></td>
780 <td class="td_left">The number of constant strings that follow.</td>
781 </tr><tr>
782 <td><a href="#uint32_vbr">uint32_vbr</a></td>
783 <td class="td_left">Zero. This identifies the following "plane" as
784 containing the constant strings.
785 </td>
786 </tr><tr>
787 <td><a href="#string">string</a></td>
788 <td class="td_left">Slot number of the constant string's type which
789 includes the length of the string.<sup>1</sup>
790 </td>
791 </tr>
792 </table>
793 Notes:
794 <ol>
795 <li>Repeated field.</li>
796 </ol>
797 <p><b>Function Constant Pool Preamble (function types)</b></p>
798 <p>The structure of the types for functions is identical to the
799 <a href="#globaltypes">Global Type Pool</a>. Please refer to that section
800 for the details.
801 <p><b>Common Part (other constants)</b></p>
802 <table>
803 <tr>
804 <th><b>Type</b></th>
805 <th class="td_left"><b>Field Description</b></th>
806 </tr><tr>
807 <td><a href="#uint32_vbr">uint32_vbr</a></td>
808 <td class="td_left">Number of entries in this type plane.</td>
809 </tr><tr>
810 <td><a href="#uint32_vbr">uint32_vbr</a></td>
811 <td class="td_left">Type slot number of this plane.</td>
812 </tr><tr>
813 <td><a href="#constant">constant</a></td>
814 <td class="td_left">The definition of a constant (see below).</td>
815 </tr>
816 </table>
817</div>
818<!-- _______________________________________________________________________ -->
819<div class="doc_subsubsection"><a name="constant">Constant Field</a></div>
820<div class="doc_text">
821 <p>Constants come in many shapes and flavors. The sections that followe define
822 the format for each of them. All constants start with a
823 <a href="#uint32_vbr">uint32_vbr</a> encoded integer that provides the number
824 of operands for the constant. For primitive, structure, and array constants,
825 this will always be zero since those types of constants have no operands.
826 In this case, we have the following field definitions:</p>
827 <ul>
828 <li><b>Bool</b>. This is written as an <a href="#uint32_vbr">uint32_vbr</a>
829 of value 1U or 0U.</li>
830 <li><b>Signed Integers (sbyte,short,int,long)</b>. These are written as
831 an <a href="#int64_vbr">int64_vbr</a> with the corresponding value.</li>
832 <li><b>Unsigned Integers (ubyte,ushort,uint,ulong)</b>. These are written
833 as an <a href="#uint64_vbr">uint64_vbr</a> with the corresponding value.
834 </li>
835 <li><b>Floating Point</b>. Both the float and double types are written
836 literally in binary format.</li>
837 <li><b>Arrays</b>. Arrays are written simply as a list of
838 <a href="#uint32_vbr">uint32_vbr</a> encoded slot numbers to the constant
839 element values.</li>
840 <li><b>Structures</b>. Structures are written simply as a list of
841 <a href="#uint32_vbr">uint32_vbr</a> encoded slot numbers to the constant
842 field values of the structure.</li>
843 </ul>
844 <p>When the number of operands to the constant is non-zero, we have a
845 constant expression and its field format is provided in the table below.</p>
846 <table>
847 <tr>
848 <th><b>Type</b></th>
849 <th class="td_left"><b>Field Description</b></th>
850 </tr><tr>
851 <td><a href="#uint32_vbr">uint32_vbr</a></td>
852 <td class="td_left">Op code of the instruction for the constant
853 expression.</td>
854 </tr><tr>
855 <td><a href="#uint32_vbr">uint32_vbr</a></td>
856 <td class="td_left">The slot number of the constant value for an
857 operand.<sup>1</sup></td>
858 </tr><tr>
859 <td><a href="#uint32_vbr">uint32_vbr</a></td>
860 <td class="td_left">The slot number for the type of the constant value
861 for an operand.<sup>1</sup></td>
862 </tr>
863 </table>
864 Notes:<ol>
865 <li>Both these fields are repeatable but only in pairs.</li>
866 </ol>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]867</div>
868<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]869<div class="doc_subsection"><a name="functiondefs">Function Definition</a> </div>
870<div class="doc_text">
871 <p>To be determined.</p>
872</div>
873<!-- _______________________________________________________________________ -->
874<div class="doc_subsection"><a name="compactiontable">Compaction Table</a> </div>
875<div class="doc_text">
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]876 <p>Compaction tables are part of a function definition. They are merely a
877 device for reducing the size of bytecode files. The size of a bytecode
878 file is dependent on the <em>value</em> of the slot numbers used because
879 larger values use more bytes in the variable bit rate encoding scheme.
880 Furthermore, the compresses instruction format reserves only six bits for
881 the type of the instruction. In large modules, declaring hundreds or thousands
882 of types, the values of the slot numbers can be quite large. However,
883 functions may use only a small fraction of the global types. In such cases
884 a compaction table is created that maps the global type and value slot
885 numbers to smaller values used by a function. Compaction tables have the
886 format shown in the table below.</p>
887 <table>
888 <tr>
889 <th><b>Type</b></th>
890 <th class="td_left"><b>Field Description</b></th>
891 </tr><tr>
892 <td><a href="#uint32_vbr">uint32_vbr</a></td>
893 <td class="td_left">The number of types that follow</td>
894 </tr><tr>
895 <td><a href="#uint32_vbr">uint32_vbr</a></td>
896 <td class="td_left">The slot number in the global type plane of the
897 type that will be referenced in the function with the index of
898 this entry in the compaction table.<sup>1</sup></td>
899 </tr><tr>
900 <td><a href="#type_len">type_len</a></td>
901 <td class="td_left">An encoding of the type and number of values that
902 follow.<sup>2</sup></td>
903 </tr><tr>
904 <td><a href="#uint32_vbr">uint32_vbr</a></td>
905 <td class="td_left">The slot number in the globals of the value that
906 will be referenced in the function with the index of this entry in
907 the compaction table<sup>1</sup></td>
908 </tr>
909 </table>
910 Notes:<ol>
911 <li>Repeated field.</li>
912 <li>This field's encoding varies depending on the size of the type plane.
913 See <a href="#type_len">Type and Length</a> for further details.
914 </ol>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]915</div>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]916
917<!-- _______________________________________________________________________ -->
918<div class="doc_subsubsection"><a name="type_len">Type and Length</a></div>
919<div class="doc_text">
920 <p>The type and length of a compaction table type plane is encoded differently
921 depending on the length of the plane. For planes of length 1 or 2, the length
922 is encoded into bits 0 and 1 of a <a href="#uint32_vbr">uint32_vbr</a> and the
923 type is encoded into bits 2-31. Because type numbers are often small, this
924 often saves an extra byte per plane. If the length of the plane is greater
925 than 2 then the encoding uses a <a href="#uint32_vbr">uint32_vbr</a> for each
926 of the length and type, in that order.</p>
927</div>
928
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]929<!-- _______________________________________________________________________ -->
930<div class="doc_subsection"><a name="instructionlist">Instruction List</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]931<div class="doc_text">
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]932 <p>To be determined.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]933</div>
934<!-- _______________________________________________________________________ -->
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]935<div class="doc_subsection"><a name="symtab">Symbol Table</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]936<div class="doc_text">
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]937<p>A symbol table can be put out in conjunction with a module or a function.
938A symbol table is a list of type planes. Each type plane starts with the number
939of entries in the plane and the type plane's slot number (so the type can be
940looked up in the global type pool). For each entry in a type plane, the slot
941number of the value and the name associated with that value are written. The
942format is given in the table below. </p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]943<table>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]944 <tr>
945 <th><b>Byte(s)</b></th>
946 <th><b>Bit(s)</b></th>
947 <th><b>Align?</b></th>
948 <th><b>Type</b></th>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]949 <th class="td_left"><b>Field Description</b></th>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]950 </tr><tr>
951 <td>00-03</td><td>-</td><td>No</td><td>unsigned</td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]952 <td class="td_left">Symbol Table Identifier (0x13)</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]953 </tr><tr>
954 <td>04-07</td><td>-</td><td>No</td><td>unsigned</td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]955 <td class="td_left">Size in bytes of the symbol table block.</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]956 </tr><tr>
957 <td>08-11<sup>1</sup></td><td>-</td><td>No</td><td>uint32_vbr</td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]958 <td class="td_left">Number of entries in type plane</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]959 </tr><tr>
960 <td>12-15<sup>1</sup></td><td>-</td><td>No</td><td>uint32_vbr</td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]961 <td class="td_left">Type plane index for following entries</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]962 </tr><tr>
963 <td>16-19<sup>1,2</sup></td><td>-</td><td>No</td><td>uint32_vbr</td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]964 <td class="td_left">Slot number of a value.</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]965 </tr><tr>
966 <td>variable<sup>1,2</sup></td><td>-</td><td>No</td><td>string</td>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]967 <td class="td_left">Name of the value in the symbol table.</td>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]968 </tr>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]969 </tr>
970</table>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame^]971Notes:
972<ol>
973 <li>Maximum length shown, may be smaller</li>
974 <li>Repeated field.</li>
975</ol>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]976</div>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]977<!-- *********************************************************************** -->
978<div class="doc_section"> <a name="versiondiffs">Version Differences</a> </div>
979<!-- *********************************************************************** -->
980<div class="doc_text">
981<p>This section describes the differences in the Bytecode Format across LLVM
982versions. The versions are listed in reverse order because it assumes the
983current version is as documented in the previous sections. Each section here
Chris Lattner1cc070c2004-07-05 18:05:48 +0000[diff] [blame]984describes the differences between that version and the one that <i>follows</i>.
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]985</p>
986</div>
987<!-- _______________________________________________________________________ -->
988<div class="doc_subsection">
989<a name="vers12">Version 1.2 Differences From 1.3</a></div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]990<!-- _______________________________________________________________________ -->
991<div class="doc_subsubsection">Type Derives From Value</div>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]992<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]993 <p>In version 1.2, the Type class in the LLVM IR derives from the Value class.
994 This is not the case in version 1.3. Consequently, in version 1.2 the notion
995 of a "Type Type" was used to write out values that were Types. The types
996 always occuped plane 12 (corresponding to the TypeTyID) of any type planed
997 set of values. In 1.3 this representation is not convenient because the
998 TypeTyID (12) is not present and its value is now used for LabelTyID.
999 Consequently, the data structures written that involve types do so by writing
1000 all the types first and then each of the value planes according to those
1001 types. In version 1.2, the types would have been written intermingled with
1002 the values.</p>
1003</div>
1004
1005<!-- _______________________________________________________________________ -->
1006<div class="doc_subsubsection">Restricted getelementptr Types</a></div>
1007<div class="doc_text">
1008 <p>In version 1.2, the getelementptr instruction required a ubyte type index
1009 for accessing a structure field and a long type index for accessing an array
1010 element. Consequently, it was only possible to access structures of 255 or
1011 fewer elements. Starting in version 1.3, this restriction was lifted.
Chris Lattner7c66ab32004-07-05 17:55:28 +0000[diff] [blame]1012 Structures must now be indexed with uint constants. Arrays may now be
1013 indexed with int, uint, long, or ulong typed values.
1014 The consequence of this was that the bytecode format had to
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1015 change in order to accommodate the larger range of structure indices.</p>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1016</div>
1017
1018<!-- _______________________________________________________________________ -->
1019<div class="doc_subsection">
1020<a name="vers11">Version 1.1 Differences From 1.2 </a></div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1021<!-- _______________________________________________________________________ -->
1022<div class="doc_subsubsection">Explicit Primitive Zeros</div>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1023<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1024 <p>In version 1.1, the zero value for primitives was explicitly encoded into
1025 the bytecode format. Since these zero values are constant values in the
1026 LLVM IR and never change, there is no reason to explicitly encode them. This
1027 explicit encoding was removed in version 1.2.</p>
1028</div>
1029
1030<!-- _______________________________________________________________________ -->
1031<div class="doc_subsubsection">Inconsistent Module Global Info</div>
1032<div class="doc_text">
1033 <p>In version 1.1, the Module Global Info block was not aligned causing the
1034 next block to be read in on an unaligned boundary. This problem was corrected
1035 in version 1.2.</p>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1036</div>
1037
1038<!-- _______________________________________________________________________ -->
1039<div class="doc_subsection">
1040<a name="vers11">Version 1.0 Differences From 1.1</a></div>
1041<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1042<p>None. Version 1.0 and 1.1 bytecode formats are identical.</p>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1043</div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]1044
1045<!-- *********************************************************************** -->
1046<hr>
1047<address>
1048 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1049 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
1050 <a href="http://validator.w3.org/check/referer"><img
1051 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
1052
1053 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a> and
1054 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1055 <a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
1056 Last modified: $Date$
1057</address>
1058</body>
1059</html>
1060<!-- vim: sw=2
1061-->