Gitiles
Code Review Sign In
gerrit-public.fairphone.software / platform / external / llvm / 09daa63597e24c704a253ccb9091f7968ae27047 / . / docs / BytecodeFormat.html
blob: ad353155a6ae3a8ba775725691f87770517d6287 [file] [log] [blame]
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]2<html>
3<head>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]4 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]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 Spencer2de29992004-08-03 20:21:05 +0000[diff] [blame]8 TR, TD { border: 2px solid gray; padding-left: 4pt; padding-right: 4pt;
9 padding-top: 2pt; padding-bottom: 2pt; }
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]10 TH { border: 2px solid gray; font-weight: bold; font-size: 105%; }
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]11 TABLE { text-align: center; border: 2px solid black;
Reid Spencer2de29992004-08-03 20:21:05 +0000[diff] [blame]12 border-collapse: collapse; margin-top: 1em; margin-left: 1em;
13 margin-right: 1em; margin-bottom: 1em; }
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]14 .td_left { border: 2px solid gray; text-align: left; }
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]15 </style>
16</head>
17<body>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]18<div class="doc_title"> LLVM Bytecode File Format </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]19<ol>
20 <li><a href="#abstract">Abstract</a></li>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]21 <li><a href="#concepts">Concepts</a>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]22 <ol>
23 <li><a href="#blocks">Blocks</a></li>
24 <li><a href="#lists">Lists</a></li>
25 <li><a href="#fields">Fields</a></li>
26 <li><a href="#align">Alignment</a></li>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]27 <li><a href="#vbr">Variable Bit-Rate Encoding</a></li>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]28 <li><a href="#encoding">Encoding Primitives</a></li>
29 <li><a href="#slots">Slots</a></li>
30 </ol>
31 </li>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]32 <li><a href="#general">General Structure</a> </li>
33 <li><a href="#blockdefs">Block Definitions</a>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]34 <ol>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]35 <li><a href="#signature">Signature Block</a></li>
36 <li><a href="#module">Module Block</a></li>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]37 <li><a href="#globaltypes">Global Type Pool</a></li>
38 <li><a href="#globalinfo">Module Info Block</a></li>
39 <li><a href="#constantpool">Global Constant Pool</a></li>
40 <li><a href="#functiondefs">Function Definition</a></li>
41 <li><a href="#compactiontable">Compaction Table</a></li>
42 <li><a href="#instructionlist">Instruction List</a></li>
43 <li><a href="#symtab">Symbol Table</a></li>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]44 </ol>
45 </li>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]46 <li><a href="#versiondiffs">Version Differences</a>
47 <ol>
48 <li><a href="#vers12">Version 1.2 Differences From 1.3</a></li>
49 <li><a href="#vers11">Version 1.1 Differences From 1.2</a></li>
50 <li><a href="#vers10">Version 1.0 Differences From 1.1</a></li>
51 </ol>
52 </li>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]53</ol>
Chris Lattner8dabb502004-05-25 17:44:58 +0000[diff] [blame]54<div class="doc_author">
55<p>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a>
56</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]57</div>
58<!-- *********************************************************************** -->
59<div class="doc_section"> <a name="abstract">Abstract </a></div>
60<!-- *********************************************************************** -->
61<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]62<p>This document describes the LLVM bytecode file format. It specifies
63the binary encoding rules of the bytecode file format so that
64equivalent systems can encode bytecode files correctly. The LLVM
65bytecode representation is used to store the intermediate
66representation on disk in compacted form.</p>
67<p>The LLVM bytecode format may change in the future, but LLVM will
68always be backwards compatible with older formats. This document will
69only describe the most current version of the bytecode format. See <a
70 href="#versiondiffs">Version Differences</a> for the details on how
71the current version is different from previous versions.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]72</div>
73<!-- *********************************************************************** -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]74<div class="doc_section"> <a name="concepts">Concepts</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]75<!-- *********************************************************************** -->
76<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]77<p>This section describes the general concepts of the bytecode file
78format without getting into specific layout details. It is recommended
79that you read this section thoroughly before interpreting the detailed
80descriptions.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]81</div>
82<!-- _______________________________________________________________________ -->
83<div class="doc_subsection"><a name="blocks">Blocks</a> </div>
84<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]85<p>LLVM bytecode files consist simply of a sequence of blocks of bytes
86using a binary encoding Each block begins with an header of two
87unsigned integers. The first value identifies the type of block and the
88second value provides the size of the block in bytes. The block
89identifier is used because it is possible for entire blocks to be
90omitted from the file if they are empty. The block identifier helps the
91reader determine which kind of block is next in the file. Note that
92blocks can be nested within other blocks.</p>
93<p> All blocks are variable length, and the block header specifies the
94size of the block. All blocks begin on a byte index that is aligned to
95an even 32-bit boundary. That is, the first block is 32-bit aligned
96because it starts at offset 0. Each block is padded with zero fill
97bytes to ensure that the next block also starts on a 32-bit boundary.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]98</div>
99<!-- _______________________________________________________________________ -->
100<div class="doc_subsection"><a name="lists">Lists</a> </div>
101<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]102<p>LLVM Bytecode blocks often contain lists of things of a similar
103type. For example, a function contains a list of instructions and a
104function type contains a list of argument types. There are two basic
105types of lists: length lists (<a href="#llist">llist</a>), and null
106terminated lists (<a href="#zlist">zlist</a>), as described below in
107the <a href="#encoding">Encoding Primitives</a>.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]108</div>
109<!-- _______________________________________________________________________ -->
110<div class="doc_subsection"><a name="fields">Fields</a> </div>
111<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]112<p>Fields are units of information that LLVM knows how to write atomically. Most
113fields have a uniform length or some kind of length indication built into their
114encoding. For example, a constant string (array of bytes) is written simply as
115the length followed by the characters. Although this is similar to a list,
116constant strings are treated atomically and are thus fields.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]117<p>Fields use a condensed bit format specific to the type of information
118they must contain. As few bits as possible are written for each field. The
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]119sections that follow will provide the details on how these fields are
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]120written and how the bits are to be interpreted.</p>
121</div>
122<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]123<div class="doc_subsection"><a name="align">Alignment</a> </div>
Reid Spencer7aa940d2004-05-25 15:47:57 +0000[diff] [blame]124<div class="doc_text">
Reid Spencer267660f2004-08-03 20:33:56 +0000[diff] [blame]125 <p>To support cross-platform differences, the bytecode file is aligned on
126 certain boundaries. This means that a small amount of padding (at most 3
127 bytes) will be added to ensure that the next entry is aligned to a 32-bit
128 boundary.</p>
Chris Lattner8dabb502004-05-25 17:44:58 +0000[diff] [blame]129</div>
Reid Spencer7aa940d2004-05-25 15:47:57 +0000[diff] [blame]130<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]131<div class="doc_subsection"><a name="vbr">Variable Bit-Rate Encoding</a>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]132</div>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]133<div class="doc_text">
134<p>Most of the values written to LLVM bytecode files are small integers. To
135minimize the number of bytes written for these quantities, an encoding scheme
136similar to UTF-8 is used to write integer data. The scheme is known as
137variable bit rate (vbr) encoding. In this encoding, the high bit of
138each byte is used to indicate if more bytes follow. If (byte &amp;
1390x80) is non-zero in any given byte, it means there is another byte
140immediately following that also contributes to the value. For the final
141byte (byte &amp; 0x80) is false (the high bit is not set). In each byte
142only the low seven bits contribute to the value. Consequently 32-bit
143quantities can take from one to <em>five</em> bytes to encode. In
144general, smaller quantities will encode in fewer bytes, as follows:</p>
145<table>
146 <tbody>
147 <tr>
148 <th>Byte #</th>
149 <th>Significant Bits</th>
150 <th>Maximum Value</th>
151 </tr>
152 <tr>
153 <td>1</td>
154 <td>0-6</td>
155 <td>127</td>
156 </tr>
157 <tr>
158 <td>2</td>
159 <td>7-13</td>
160 <td>16,383</td>
161 </tr>
162 <tr>
163 <td>3</td>
164 <td>14-20</td>
165 <td>2,097,151</td>
166 </tr>
167 <tr>
168 <td>4</td>
169 <td>21-27</td>
170 <td>268,435,455</td>
171 </tr>
172 <tr>
173 <td>5</td>
174 <td>28-34</td>
175 <td>34,359,738,367</td>
176 </tr>
177 <tr>
178 <td>6</td>
179 <td>35-41</td>
180 <td>4,398,046,511,103</td>
181 </tr>
182 <tr>
183 <td>7</td>
184 <td>42-48</td>
185 <td>562,949,953,421,311</td>
186 </tr>
187 <tr>
188 <td>8</td>
189 <td>49-55</td>
190 <td>72,057,594,037,927,935</td>
191 </tr>
192 <tr>
193 <td>9</td>
194 <td>56-62</td>
195 <td>9,223,372,036,854,775,807</td>
196 </tr>
197 <tr>
198 <td>10</td>
199 <td>63-69</td>
200 <td>1,180,591,620,717,411,303,423</td>
201 </tr>
202 </tbody>
203</table>
204<p>Note that in practice, the tenth byte could only encode bit 63 since
205the maximum quantity to use this encoding is a 64-bit integer.</p>
206<p><em>Signed</em> VBR values are encoded with the standard vbr
207encoding, but with the sign bit as the low order bit instead of the
208high order bit. This allows small negative quantities to be encoded
209efficiently. For example, -3
210is encoded as "((3 &lt;&lt; 1) | 1)" and 3 is encoded as "(3 &lt;&lt;
2111) | 0)", emitted with the standard vbr encoding above.</p>
212</div>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]213<!-- _______________________________________________________________________ -->
214<div class="doc_subsection"><a name="encoding">Encoding Primitives</a> </div>
215<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]216<p>Each field in the bytecode format is encoded into the file using a
217small set of primitive formats. The table below defines the encoding
218rules for the various primitives used and gives them each a type name.
219The type names used in the descriptions of blocks and fields in the <a
220 href="#details">Detailed Layout</a>next section. Any type name with
221the suffix <em>_vbr</em> indicates a quantity that is encoded using
222variable bit rate encoding as described above.</p>
223<table class="doc_table">
224 <tbody>
225 <tr>
226 <th><b>Type</b></th>
227 <th class="td_left"><b>Rule</b></th>
228 </tr>
229 <tr>
230 <td><a name="unsigned"><b>unsigned</b></a></td>
231 <td class="td_left">A 32-bit unsigned integer that always occupies four
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]232 consecutive bytes. The unsigned integer is encoded using LSB first
233 ordering. That is bits 2<sup>0</sup> through 2<sup>7</sup> are in the
234 byte with the lowest file offset (little endian).</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]235 </tr>
236 <tr>
Reid Spencer301fe482004-08-03 20:57:56 +0000[diff] [blame]237 <td style="vertical-align: top;"><a name="uint24_vbr">
238 <b>uint24_vbr</b></a></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]239 <td style="vertical-align: top; text-align: left;">A 24-bit unsigned
240 integer that occupies from one to four bytes using variable bit rate
241 encoding.</td>
242 </tr>
243 <tr>
244 <td><a name="uint32_vbr"><b>uint32_vbr</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000[diff] [blame]245 <td class="td_left">A 32-bit unsigned integer that occupies from one to
246 five bytes using variable bit rate encoding.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]247 </tr>
248 <tr>
249 <td><a name="uint64_vbr"><b>uint64_vbr</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000[diff] [blame]250 <td class="td_left">A 64-bit unsigned integer that occupies from one to ten
251 bytes using variable bit rate encoding.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]252 </tr>
253 <tr>
254 <td><a name="int64_vbr"><b>int64_vbr</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000[diff] [blame]255 <td class="td_left">A 64-bit signed integer that occupies from one to ten
256 bytes using the signed variable bit rate encoding.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]257 </tr>
258 <tr>
259 <td><a name="char"><b>char</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000[diff] [blame]260 <td class="td_left">A single unsigned character encoded into one byte</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]261 </tr>
262 <tr>
263 <td><a name="bit"><b>bit(n-m)</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000[diff] [blame]264 <td class="td_left">A set of bit within some larger integer field. The values
265 of <code>n</code> and <code>m</code> specify the inclusive range of bits
266 that define the subfield. The value for <code>m</code> may be omitted if
267 its the same as <code>n</code>.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]268 </tr>
269 <tr>
270 <td style="vertical-align: top;"><b><a name="float"><b>float</b></a></b></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000[diff] [blame]271 <td style="vertical-align: top; text-align: left;">A floating point value encoded
272 as a 32-bit IEEE value written in little-endian form.<br>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]273 </td>
274 </tr>
275 <tr>
276 <td style="vertical-align: top;"><b><b><a name="double"><b>double</b></a></b></b></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000[diff] [blame]277 <td style="vertical-align: top; text-align: left;">A floating point value encoded
278 as a64-bit IEEE value written in little-endian form</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]279 </tr>
280 <tr>
281 <td><a name="string"><b>string</b></a></td>
282 <td class="td_left">A uint32_vbr indicating the type of the
283constant string which also includes its length, immediately followed by
284the characters of the string. There is no terminating null byte in the
285string.</td>
286 </tr>
287 <tr>
288 <td><a name="data"><b>data</b></a></td>
289 <td class="td_left">An arbitrarily long segment of data to which
290no interpretation is implied. This is used for constant initializers.<br>
291 </td>
292 </tr>
293 <tr>
294 <td><a name="llist"><b>llist(x)</b></a></td>
295 <td class="td_left">A length list of x. This means the list is
296encoded as an <a href="#uint32_vbr">uint32_vbr</a> providing the
297length of the list, followed by a sequence of that many "x" items. This
298implies that the reader should iterate the number of times provided by
299the length.</td>
300 </tr>
301 <tr>
302 <td><a name="zlist"><b>zlist(x)</b></a></td>
303 <td class="td_left">A zero-terminated list of x. This means the
304list is encoded as a sequence of an indeterminate number of "x" items,
305followed by an <a href="#uint32_vbr">uint32_vbr</a> terminating value.
306This implies that none of the "x" items can have a zero value (or else
307the list terminates).</td>
308 </tr>
309 <tr>
310 <td><a name="block"><b>block</b></a></td>
311 <td class="td_left">A block of data that is logically related. A
312block is an unsigned 32-bit integer that encodes the type of the block
313in the low 5 bits and the size of the block in the high 27 bits. The
314length does not include the block header or any alignment bytes at the
315end of the block. Blocks may compose other blocks. </td>
316 </tr>
317 </tbody>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]318</table>
319</div>
320<!-- _______________________________________________________________________ -->
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]321<div class="doc_subsection"><a name="notation">Field Notation</a> </div>
322<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]323<p>In the detailed block and field descriptions that follow, a regex
324like notation is used to describe optional and repeated fields. A very
325limited subset of regex is used to describe these, as given in the
326following table: </p>
327<table class="doc_table">
328 <tbody>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]329 <tr>
330 <th><b>Character</b></th>
331 <th class="td_left"><b>Meaning</b></th>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]332 </tr>
333 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]334 <td><b><code>?</code></b></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]335 <td class="td_left">The question mark indicates 0 or 1
336occurrences of the thing preceding it.</td>
337 </tr>
338 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]339 <td><b><code>*</code></b></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]340 <td class="td_left">The asterisk indicates 0 or more occurrences
341of the thing preceding it.</td>
342 </tr>
343 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]344 <td><b><code>+</code></b></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]345 <td class="td_left">The plus sign indicates 1 or more occurrences
346of the thing preceding it.</td>
347 </tr>
348 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]349 <td><b><code>()</code></b></td>
350 <td class="td_left">Parentheses are used for grouping.</td>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]351 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]352 <tr>
353 <td><b><code>,</code></b></td>
354 <td class="td_left">The comma separates sequential fields.</td>
355 </tr>
356 </tbody>
357</table>
358<p>So, for example, consider the following specifications:</p>
359<div class="doc_code">
360<ol>
361 <li><code>string?</code></li>
362 <li><code>(uint32_vbr,uin32_vbr)+</code></li>
363 <li><code>(unsigned?,uint32_vbr)*</code></li>
364 <li><code>(llist(unsigned))?</code></li>
365</ol>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]366</div>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]367<p>with the following interpretations:</p>
368<ol>
369 <li>An optional string. Matches either nothing or a single string</li>
370 <li>One or more pairs of uint32_vbr.</li>
371 <li>Zero or more occurrences of either an unsigned followed by a
372uint32_vbr or just a uint32_vbr.</li>
373 <li>An optional length list of unsigned values.</li>
374</ol>
375</div>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]376<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]377<div class="doc_subsection"><a name="slots">Slots</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]378<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]379<p>The bytecode format uses the notion of a "slot" to reference Types
380and Values. Since the bytecode file is a <em>direct</em> representation of
381LLVM's intermediate representation, there is a need to represent pointers in
382the file. Slots are used for this purpose. For example, if one has the following
383assembly:
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]384</p>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]385<div class="doc_code"><code> %MyType = type { int, sbyte }<br>
386%MyVar = external global %MyType
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]387</code></div>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]388<p>there are two definitions. The definition of <tt>%MyVar</tt> uses <tt>%MyType</tt>.
389In the C++ IR this linkage between <tt>%MyVar</tt> and <tt>%MyType</tt>
390is explicit through the use of C++ pointers. In bytecode, however, there's no
391ability to store memory addresses. Instead, we compute and write out
392slot numbers for every Type and Value written to the file.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]393<p>A slot number is simply an unsigned 32-bit integer encoded in the variable
394bit rate scheme (see <a href="#encoding">encoding</a>). This ensures that
395low slot numbers are encoded in one byte. Through various bits of magic LLVM
396attempts to always keep the slot numbers low. The first attempt is to associate
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]397slot numbers with their "type plane". That is, Values of the same type
398are written to the bytecode file in a list (sequentially). Their order in
399that list determines their slot number. This means that slot #1 doesn't mean
400anything unless you also specify for which type you want slot #1. Types are
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]401always written to the file first (in the <a href="#globaltypes">Global Type
402Pool</a>) and in such a way that both forward and backward references of the
403types can often be resolved with a single pass through the type pool. </p>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]404<p>Slot numbers are also kept small by rearranging their order. Because
405of the structure of LLVM, certain values are much more likely to be used
406frequently in the body of a function. For this reason, a compaction table is
407provided in the body of a function if its use would make the function body
408smaller. Suppose you have a function body that uses just the types "int*" and
409"{double}" but uses them thousands of time. Its worthwhile to ensure that the
410slot number for these types are low so they can be encoded in a single byte
411(via vbr). This is exactly what the compaction table does.</p>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]412<p>In summary then, a slot number can be though of as just a vbr encoded index
413into a list of Type* or Value*. To keep slot numbers low, Value* are indexed by
414two slot numbers: the "type plane index" (type slot) and the "value index"
415(value slot).</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]416</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]417<!-- *********************************************************************** -->
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]418<div class="doc_section"> <a name="general">General Structure</a> </div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]419<!-- *********************************************************************** -->
420<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]421<p>This section provides the general structure of the LLVM bytecode
422file format. The bytecode file format requires blocks to be in a
423certain order and nested in a particular way so that an LLVM module can
424be constructed efficiently from the contents of the file. This ordering
425defines a general structure for bytecode files as shown below. The
426table below shows the order in which all block types may appear. Please
427note that some of the blocks are optional and some may be repeated. The
428structure is fairly loose because optional blocks, if empty, are
429completely omitted from the file.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]430<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]431 <tbody>
432 <tr>
433 <th>ID</th>
434 <th>Parent</th>
435 <th>Optional?</th>
436 <th>Repeated?</th>
437 <th>Level</th>
438 <th>Block Type</th>
439 <th>Description</th>
440 </tr>
441 <tr>
442 <td>N/A</td>
443 <td>File</td>
444 <td>No</td>
445 <td>No</td>
446 <td>0</td>
447 <td class="td_left"><a href="#signature">Signature</a></td>
448 <td class="td_left">This contains the file signature (magic
449number) that identifies the file as LLVM bytecode.</td>
450 </tr>
451 <tr>
452 <td>0x01</td>
453 <td>File</td>
454 <td>No</td>
455 <td>No</td>
456 <td>0</td>
457 <td class="td_left"><a href="#module">Module</a></td>
458 <td class="td_left">This is the top level block in a bytecode
459file. It contains all the other blocks. </td>
460 </tr>
461 <tr>
462 <td>0x06</td>
463 <td>Module</td>
464 <td>No</td>
465 <td>No</td>
466 <td>1</td>
467 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#globaltypes">Global&nbsp;Type&nbsp;Pool</a></td>
468 <td class="td_left">This block contains all the global (module)
469level types.</td>
470 </tr>
471 <tr>
472 <td>0x05</td>
473 <td>Module</td>
474 <td>No</td>
475 <td>No</td>
476 <td>1</td>
477 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#globalinfo">Module&nbsp;Globals&nbsp;Info</a></td>
478 <td class="td_left">This block contains the type, constness, and
479linkage for each of the global variables in the module. It also
480contains the type of the functions and the constant initializers.</td>
481 </tr>
482 <tr>
483 <td>0x03</td>
484 <td>Module</td>
485 <td>Yes</td>
486 <td>No</td>
487 <td>1</td>
488 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#constantpool">Module&nbsp;Constant&nbsp;Pool</a></td>
489 <td class="td_left">This block contains all the global constants
490except function arguments, global values and constant strings.</td>
491 </tr>
492 <tr>
493 <td>0x02</td>
494 <td>Module</td>
495 <td>Yes</td>
496 <td>Yes</td>
497 <td>1</td>
498 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#functiondefs">Function&nbsp;Definitions</a>*</td>
499 <td class="td_left">One function block is written for each
500function in the module. The function block contains the instructions,
501compaction table, type constant pool, and symbol table for the function.</td>
502 </tr>
503 <tr>
504 <td>0x03</td>
505 <td>Function</td>
506 <td>Yes</td>
507 <td>No</td>
508 <td>2</td>
509 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a
510 href="#constantpool">Function&nbsp;Constant&nbsp;Pool</a></td>
511 <td class="td_left">Any constants (including types) used solely
512within the function are emitted here in the function constant pool. </td>
513 </tr>
514 <tr>
515 <td>0x08</td>
516 <td>Function</td>
517 <td>Yes</td>
518 <td>No</td>
519 <td>2</td>
520 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a
521 href="#compactiontable">Compaction&nbsp;Table</a></td>
522 <td class="td_left">This table reduces bytecode size by providing
523a funtion-local mapping of type and value slot numbers to their global
524slot numbers</td>
525 </tr>
526 <tr>
527 <td>0x07</td>
528 <td>Function</td>
529 <td>No</td>
530 <td>No</td>
531 <td>2</td>
532 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a
533 href="#instructionlist">Instruction&nbsp;List</a></td>
534 <td class="td_left">This block contains all the instructions of
535the function. The basic blocks are inferred by terminating
536instructions. </td>
537 </tr>
538 <tr>
539 <td>0x04</td>
540 <td>Function</td>
541 <td>Yes</td>
542 <td>No</td>
543 <td>2</td>
544 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a
545 href="#symtab">Function&nbsp;Symbol&nbsp;Table</a></td>
546 <td class="td_left">This symbol table provides the names for the
547function specific values used (basic block labels mostly).</td>
548 </tr>
549 <tr>
550 <td>0x04</td>
551 <td>Module</td>
552 <td>Yes</td>
553 <td>No</td>
554 <td>1</td>
555 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#symtab">Module&nbsp;Symbol&nbsp;Table</a></td>
556 <td class="td_left">This symbol table provides the names for the
557various entries in the file that are not function specific (global
558vars, and functions mostly).</td>
559 </tr>
560 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]561</table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]562<p>Use the links in the table for details about the contents of each of
563the block types.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]564</div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]565<!-- *********************************************************************** -->
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]566<div class="doc_section"> <a name="blockdefs">Block Definitions</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]567<!-- *********************************************************************** -->
568<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]569<p>This section provides the detailed layout of the individual block
570types in the LLVM bytecode file format. </p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]571</div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]572<!-- _______________________________________________________________________ -->
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]573<div class="doc_subsection"><a name="signature">Signature Block</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]574<div class="doc_text">
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]575<p>The signature occurs in every LLVM bytecode file and is always first.
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]576It simply provides a few bytes of data to identify the file as being an LLVM
577bytecode file. This block is always four bytes in length and differs from the
578other blocks because there is no identifier and no block length at the start
579of the block. Essentially, this block is just the "magic number" for the file.
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]580</p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]581<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]582 <tbody>
583 <tr>
584 <th><b>Type</b></th>
585 <th class="td_left"><b>Field Description</b></th>
586 </tr>
587 <tr>
588 <td><a href="#char">char</a></td>
589 <td class="td_left">Constant "l" (0x6C)</td>
590 </tr>
591 <tr>
592 <td><a href="#char">char</a></td>
593 <td class="td_left">Constant "l" (0x6C)</td>
594 </tr>
595 <tr>
596 <td><a href="#char">char</a></td>
597 <td class="td_left">Constant "v" (0x76)</td>
598 </tr>
599 <tr>
600 <td><a href="#char">char</a></td>
601 <td class="td_left">Constant "m" (0x6D)</td>
602 </tr>
603 </tbody>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]604</table>
605</div>
606<!-- _______________________________________________________________________ -->
607<div class="doc_subsection"><a name="module">Module Block</a> </div>
608<div class="doc_text">
609<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]610the file. The table below shows the structure of the module block. Note that it
611only provides the module identifier, size of the module block, and the format
612information. Everything else is contained in other blocks, described in other
613sections.</p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]614<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]615 <tbody>
616 <tr>
617 <th><b>Type</b></th>
618 <th class="td_left"><b>Field Description</b></th>
619 </tr>
620 <tr>
Reid Spencer5bc74d52004-08-16 19:24:36 +0000[diff] [blame]621 <td><a href="#unsigned">unsigned</a><br></td>
622 <td class="td_left"><a href="#mod_header">Module Block Identifier
623 (0x01)</a></td>
624 </tr>
625 <tr>
626 <td><a href="#unsigned">unsigned</a></td>
627 <td class="td_left"><a href="#mod_header">Module Block Size</a></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]628 </tr>
629 <tr>
630 <td><a href="#uint32_vbr">uint32_vbr</a></td>
631 <td class="td_left"><a href="#format">Format Information</a></td>
632 </tr>
633 <tr>
634 <td><a href="#block">block</a></td>
635 <td class="td_left"><a href="#globaltypes">Global Type Pool</a></td>
636 </tr>
637 <tr>
638 <td><a href="#block">block</a></td>
639 <td class="td_left"><a href="#globalinfo">Module Globals Info</a></td>
640 </tr>
641 <tr>
642 <td><a href="#block">block</a></td>
643 <td class="td_left"><a href="#constantpool">Module Constant Pool</a></td>
644 </tr>
645 <tr>
646 <td><a href="#block">block</a>*</td>
647 <td class="td_left"><a href="#functiondefs">Function Definitions</a></td>
648 </tr>
649 <tr>
650 <td><a href="#block">block</a></td>
Reid Spencer8996e552004-08-17 00:49:03 +0000[diff] [blame]651 <td class="td_left"><a href="#symtab">Module Symbol Table</a></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]652 </tr>
653 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]654</table>
655</div>
Reid Spencer5bc74d52004-08-16 19:24:36 +0000[diff] [blame]656
657<!-- _______________________________________________________________________ -->
658<div class="doc_subsubsection"><a name="mod_header">Module Block Header</a></div>
659<div class="doc_text">
660 <p>The block header for the module block uses a longer format than the other
661 blocks in a bytecode file. Specifically, instead of encoding the type and size
662 of the block into a 32-bit integer with 5-bits for type and 27-bits for size,
663 the module block header uses two 32-bit unsigned values, one for type, and one
664 for size. While the 2<sup>27</sup> byte limit on block size is sufficient for the blocks
665 contained in the module, it isn't sufficient for the module block itself
666 because we want to ensure that bytecode files as large as 2<sup>32</sup> bytes
667 are possible. For this reason, the module block (and only the module block)
668 uses a long format header.</p>
669</div>
670
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]671<!-- _______________________________________________________________________ -->
672<div class="doc_subsubsection"><a name="format">Format Information</a></div>
673<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]674<p>The format information field is encoded into a <a href="#uint32_vbr">uint32_vbr</a>
675as shown in the following table.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]676<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]677 <tbody>
678 <tr>
679 <th><b>Type</b></th>
680 <th class="td_left"><b>Description</b></th>
681 </tr>
682 <tr>
683 <td><a href="#bit">bit(0)</a></td>
684 <td class="td_left">Target is big endian?</td>
685 </tr>
686 <tr>
687 <td><a href="#bit">bit(1)</a></td>
688 <td class="td_left">On target pointers are 64-bit?</td>
689 </tr>
690 <tr>
691 <td><a href="#bit">bit(2)</a></td>
692 <td class="td_left">Target has no endianess?</td>
693 </tr>
694 <tr>
695 <td><a href="#bit">bit(3)</a></td>
696 <td class="td_left">Target has no pointer size?</td>
697 </tr>
698 <tr>
699 <td><a href="#bit">bit(4-31)</a></td>
700 <td class="td_left">Bytecode format version</td>
701 </tr>
702 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]703</table>
704<p>
705Of particular note, the bytecode format number is simply a 28-bit
706monotonically increase integer that identifies the version of the bytecode
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]707format (which is not directly related to the LLVM release number). The
708bytecode versions defined so far are (note that this document only
709describes the latest version, 1.3):</p>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]710<ul>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]711 <li>#0: LLVM 1.0 &amp; 1.1</li>
712 <li>#1: LLVM 1.2</li>
713 <li>#2: LLVM 1.2.5 (not released)</li>
714 <li>#3: LLVM 1.3<br>
715 </li>
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]716</ul>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]717<p>Note that we plan to eventually expand the target description
718capabilities
719of bytecode files to <a href="http://llvm.cs.uiuc.edu/PR263">target
720triples</a>.
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]721</p>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]722</div>
723<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]724<div class="doc_subsection"><a name="globaltypes">Global Type Pool</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]725<div class="doc_text">
Chris Lattner2b905652004-05-24 05:35:17 +0000[diff] [blame]726<p>The global type pool consists of type definitions. Their order of appearance
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]727in the file determines their type slot number (0 based). Slot numbers are
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]728used to replace pointers in the intermediate representation. Each slot number
729uniquely identifies one entry in a type plane (a collection of values of the
730same type). Since all values have types and are associated with the order in
731which the type pool is written, the global type pool <em>must</em> be written
732as the first block of a module. If it is not, attempts to read the file will
733fail because both forward and backward type resolution will not be possible.</p>
734<p>The type pool is simply a list of type definitions, as shown in the
735table below.</p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]736<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]737 <tbody>
738 <tr>
739 <th><b>Type</b></th>
740 <th class="td_left"><b>Field Description</b></th>
741 </tr>
742 <tr>
743 <td><a href="#unsigned">block</a></td>
744 <td class="td_left">Type Pool Identifier (0x06) + Size<br>
745 </td>
746 </tr>
747 <tr>
748 <td><a href="#llist">llist</a>(<a href="#type">type</a>)</td>
749 <td class="td_left">A length list of type definitions.</td>
750 </tr>
751 </tbody>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]752</table>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]753</div>
754<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]755<div class="doc_subsubsection"><a name="type">Type Definitions</a></div>
756<div class="doc_text">
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]757<p>Types in the type pool are defined using a different format for each kind
758of type, as given in the following sections.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]759<h3>Primitive Types</h3>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]760<p>The primitive types encompass the basic integer and floating point
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]761types. They are encoded simply as their TypeID.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]762<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]763 <tbody>
764 <tr>
765 <th><b>Type</b></th>
766 <th class="td_left"><b>Description</b></th>
767 </tr>
768 <tr>
769 <td><a href="#uint24_vbr">uint24_vbr</a></td>
770 <td class="td_left">Type ID for the primitive types (values 1 to
77111) <sup>1</sup></td>
772 </tr>
773 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]774</table>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]775Notes:
776<ol>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]777 <li>The values for the Type IDs for the primitive types are provided
778by the definition of the <code>llvm::Type::TypeID</code> enumeration
779in <code>include/llvm/Type.h</code>. The enumeration gives the
780following mapping:
781 <ol>
782 <li>bool</li>
783 <li>ubyte</li>
784 <li>sbyte</li>
785 <li>ushort</li>
786 <li>short</li>
787 <li>uint</li>
788 <li>int</li>
789 <li>ulong</li>
790 <li>long</li>
791 <li>float</li>
792 <li>double</li>
793 </ol>
794 </li>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]795</ol>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]796<h3>Function Types</h3>
797<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]798 <tbody>
799 <tr>
800 <th><b>Type</b></th>
801 <th class="td_left"><b>Description</b></th>
802 </tr>
803 <tr>
804 <td><a href="#uint24_vbr">uint24_vbr</a></td>
805 <td class="td_left">Type ID for function types (13)</td>
806 </tr>
807 <tr>
808 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]809 <td class="td_left">Type slot number of function's return type.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]810 </tr>
811 <tr>
812 <td><a href="#llist">llist</a>(<a href="#uint24_vbr">uint24_vbr</a>)</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]813 <td class="td_left">Type slot number of each argument's type.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]814 </tr>
815 <tr>
816 <td><a href="#uint32_vbr">uint32_vbr</a>?</td>
817 <td class="td_left">Value 0 if this is a varargs function,
818missing otherwise.</td>
819 </tr>
820 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]821</table>
822<h3>Structure Types</h3>
823<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]824 <tbody>
825 <tr>
826 <th><b>Type</b></th>
827 <th class="td_left"><b>Description</b></th>
828 </tr>
829 <tr>
830 <td><a href="#uint24_vbr">uint24_vbr</a></td>
831 <td class="td_left">Type ID for structure types (14)</td>
832 </tr>
833 <tr>
834 <td><a href="#zlist">zlist</a>(<a href="#uint24_vbr">uint24_vbr</a>)</td>
835 <td class="td_left">Slot number of each of the element's fields.</td>
836 </tr>
837 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]838</table>
839<h3>Array Types</h3>
840<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]841 <tbody>
842 <tr>
843 <th><b>Type</b></th>
844 <th class="td_left"><b>Description</b></th>
845 </tr>
846 <tr>
847 <td><a href="#uint24_vbr">uint24_vbr</a></td>
848 <td class="td_left">Type ID for Array Types (15)</td>
849 </tr>
850 <tr>
851 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]852 <td class="td_left">Type slot number of array's element type.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]853 </tr>
854 <tr>
855 <td><a href="#uint32_vbr">uint32_vbr</a></td>
856 <td class="td_left">The number of elements in the array.</td>
857 </tr>
858 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]859</table>
860<h3>Pointer Types</h3>
861<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]862 <tbody>
863 <tr>
864 <th><b>Type</b></th>
865 <th class="td_left"><b>Description</b></th>
866 </tr>
867 <tr>
868 <td><a href="#uint24_vbr">uint24_vbr</a></td>
869 <td class="td_left">Type ID For Pointer Types (16)</td>
870 </tr>
871 <tr>
872 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]873 <td class="td_left">Type slot number of pointer's element type.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]874 </tr>
875 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]876</table>
877<h3>Opaque Types</h3>
878<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]879 <tbody>
880 <tr>
881 <th><b>Type</b></th>
882 <th class="td_left"><b>Description</b></th>
883 </tr>
884 <tr>
885 <td><a href="#uint24_vbr">uint24_vbr</a></td>
886 <td class="td_left">Type ID For Opaque Types (17)</td>
887 </tr>
888 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]889</table>
890</div>
891<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]892<div class="doc_subsection"><a name="globalinfo">Module Global Info</a>
893</div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]894<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]895<p>The module global info block contains the definitions of all global
896variables including their initializers and the <em>declaration</em> of
897all functions. The format is shown in the table below:</p>
898<table>
899 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]900 <tr>
901 <th><b>Type</b></th>
902 <th class="td_left"><b>Field Description</b></th>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]903 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]904 <tr>
905 <td><a href="#block">block</a></td>
906 <td class="td_left">Module global info identifier (0x05) + size<br>
907 </td>
908 </tr>
909 <tr>
910 <td><a href="#zlist">zlist</a>(<a href="#globalvar">globalvar</a>)</td>
911 <td class="td_left">A zero terminated list of global var
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]912definitions occurring in the module.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]913 </tr>
914 <tr>
915 <td><a href="#zlist">zlist</a>(<a href="#uint24_vbr">uint24_vbr</a>)</td>
916 <td class="td_left">A zero terminated list of function types
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]917occurring in the module.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]918 </tr>
919 <tr>
920 <td style="vertical-align: top;"><a href="#llist">llist</a>(<a
921 href="#string">string</a>)<br>
922 </td>
923 <td style="vertical-align: top; text-align: left;">A length list
924of strings that specify the names of the libraries that this module
925depends upon.<br>
926 </td>
927 </tr>
928 <tr>
929 <td style="vertical-align: top;"><a href="#string">string</a><br>
930 </td>
931 <td style="vertical-align: top; text-align: left;">The target
932triple for the module (blank means no target triple specified, i.e. a
933platform independent module).<br>
934 </td>
935 </tr>
936 </tbody>
937</table>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]938</div>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]939<!-- _______________________________________________________________________ -->
940<div class="doc_subsubsection"><a name="globalvar">Global Variable Field</a>
941</div>
942<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]943<p>Global variables are written using an <a href="#uint32_vbr">uint32_vbr</a>
944that encodes information about the global variable and a list of the
945constant initializers for the global var, if any.</p>
946<p>The table below provides the bit layout of the first <a
947 href="#uint32_vbr">uint32_vbr</a> that describes the global variable.</p>
948<table>
949 <tbody>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]950 <tr>
951 <th><b>Type</b></th>
952 <th class="td_left"><b>Description</b></th>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]953 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]954 <tr>
955 <td><a href="#bit">bit(0)</a></td>
956 <td class="td_left">Is constant?</td>
957 </tr>
958 <tr>
959 <td><a href="#bit">bit(1)</a></td>
960 <td class="td_left">Has initializer? Note that this bit
961determines whether the constant initializer field (described below)
962follows. </td>
963 </tr>
964 <tr>
965 <td><a href="#bit">bit(2-4)</a></td>
966 <td class="td_left">Linkage type: 0=External, 1=Weak,
9672=Appending, 3=Internal, 4=LinkOnce</td>
968 </tr>
969 <tr>
970 <td><a href="#bit">bit(5-31)</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]971 <td class="td_left">Type slot number of type for the global variable.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]972 </tr>
973 </tbody>
974</table>
975<p>The table below provides the format of the constant initializers for
976the global variable field, if it has one.</p>
977<table>
978 <tbody>
979 <tr>
980 <th><b>Type</b></th>
981 <th class="td_left"><b>Description</b></th>
982 </tr>
983 <tr>
984 <td>(<a href="#zlist">zlist</a>(<a href="#uint32_vbr">uint32_vbr</a>))?
985 </td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]986 <td class="td_left">An optional zero-terminated list of value slot
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]987numbers of the global variable's constant initializer.</td>
988 </tr>
989 </tbody>
990</table>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]991</div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]992<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]993<div class="doc_subsection"><a name="constantpool">Constant Pool</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]994<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]995<p>A constant pool defines as set of constant values. There are
996actually two types of constant pool blocks: one for modules and one for
997functions. For modules, the block begins with the constant strings
998encountered anywhere in the module. For functions, the block begins
999with types only encountered in the function. In both cases the header
1000is identical. The tables that follow, show the header, module constant
1001pool preamble, function constant pool preamble, and the part common to
1002both function and module constant pools.</p>
1003<p><b>Common Block Header</b></p>
1004<table>
1005 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1006 <tr>
1007 <th><b>Type</b></th>
1008 <th class="td_left"><b>Field Description</b></th>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1009 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1010 <tr>
1011 <td><a href="#block">block</a></td>
1012 <td class="td_left">Constant pool identifier (0x03) + size<br>
1013 </td>
1014 </tr>
1015 </tbody>
1016</table>
1017<p><b>Module Constant Pool Preamble (constant strings)</b></p>
1018<table>
1019 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1020 <tr>
1021 <th><b>Type</b></th>
1022 <th class="td_left"><b>Field Description</b></th>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1023 </tr>
1024 <tr>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1025 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1026 <td class="td_left">The number of constant strings that follow.</td>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1027 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1028 <tr>
1029 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1030 <td class="td_left">Zero. This identifies the following "plane"
1031as containing the constant strings. This is needed to identify it
1032uniquely from other constant planes that follow. </td>
1033 </tr>
1034 <tr>
1035 <td><a href="#uint24_vbr">uint24_vbr</a>+</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1036 <td class="td_left">Type slot number of the constant string's type.
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1037Note that the constant string's type implicitly defines the length of
1038the string. </td>
1039 </tr>
1040 </tbody>
1041</table>
1042<p><b>Function Constant Pool Preamble (function types)</b></p>
1043<p>The structure of the types for functions is identical to the <a
1044 href="#globaltypes">Global Type Pool</a>. Please refer to that section
1045for the details. </p>
1046<p><b>Common Part (other constants)</b></p>
1047<table>
1048 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1049 <tr>
1050 <th><b>Type</b></th>
1051 <th class="td_left"><b>Field Description</b></th>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1052 </tr>
1053 <tr>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1054 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1055 <td class="td_left">Number of entries in this type plane.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1056 </tr>
1057 <tr>
1058 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1059 <td class="td_left">Type slot number of this plane.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1060 </tr>
1061 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]1062 <td><a href="#constant">constant</a>+</td>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1063 <td class="td_left">The definition of a constant (see below).</td>
1064 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1065 </tbody>
1066</table>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1067</div>
1068<!-- _______________________________________________________________________ -->
1069<div class="doc_subsubsection"><a name="constant">Constant Field</a></div>
1070<div class="doc_text">
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1071<p>Constants come in many shapes and flavors. The sections that follow
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1072define the format for each of them. All constants start with a <a
1073 href="#uint32_vbr">uint32_vbr</a> encoded integer that provides the
1074number of operands for the constant. For primitive, structure, and
1075array constants, this will always be zero since those types of
1076constants have no operands. In this case, we have the following field
1077definitions:</p>
1078<ul>
1079 <li><b>Bool</b>. This is written as an <a href="#uint32_vbr">uint32_vbr</a>
1080of value 1U or 0U.</li>
1081 <li><b>Signed Integers (sbyte,short,int,long)</b>. These are written
1082as an <a href="#int64_vbr">int64_vbr</a> with the corresponding value.</li>
1083 <li><b>Unsigned Integers (ubyte,ushort,uint,ulong)</b>. These are
1084written as an <a href="#uint64_vbr">uint64_vbr</a> with the
1085corresponding value. </li>
1086 <li><b>Floating Point</b>. Both the float and double types are
1087written literally in binary format.</li>
1088 <li><b>Arrays</b>. Arrays are written simply as a list of <a
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1089 href="#uint32_vbr">uint32_vbr</a> encoded value slot numbers to the constant
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1090element values.</li>
1091 <li><b>Structures</b>. Structures are written simply as a list of <a
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1092 href="#uint32_vbr">uint32_vbr</a> encoded value slot numbers to the constant
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1093field values of the structure.</li>
1094</ul>
1095<p>When the number of operands to the constant is non-zero, we have a
1096constant expression and its field format is provided in the table below.</p>
1097<table>
1098 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1099 <tr>
1100 <th><b>Type</b></th>
1101 <th class="td_left"><b>Field Description</b></th>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1102 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1103 <tr>
1104 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1105 <td class="td_left">Op code of the instruction for the constant
1106expression.</td>
1107 </tr>
1108 <tr>
1109 <td><a href="#uint32_vbr">uint32_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1110 <td class="td_left">The value slot number of the constant value for an
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1111operand.<sup>1</sup></td>
1112 </tr>
1113 <tr>
1114 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1115 <td class="td_left">The type slot number for the type of the constant
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1116value for an operand.<sup>1</sup></td>
1117 </tr>
1118 </tbody>
1119</table>
1120Notes:
1121<ol>
1122 <li>Both these fields are repeatable but only in pairs.</li>
1123</ol>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]1124</div>
1125<!-- _______________________________________________________________________ -->
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1126<div class="doc_subsection"><a name="functiondefs">Function Definition</a></div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1127<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1128<p>Function definitions contain the linkage, constant pool or
1129compaction table, instruction list, and symbol table for a function.
1130The following table shows the structure of a function definition.</p>
1131<table>
1132 <tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1133 <tr>
1134 <th><b>Type</b></th>
1135 <th class="td_left"><b>Field Description</b></th>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1136 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1137 <tr>
1138 <td><a href="#block">block</a><br>
1139 </td>
1140 <td class="td_left">Function definition block identifier (0x02) +
1141size<br>
1142 </td>
1143 </tr>
1144 <tr>
1145 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1146 <td class="td_left">The linkage type of the function: 0=External,
11471=Weak, 2=Appending, 3=Internal, 4=LinkOnce<sup>1</sup></td>
1148 </tr>
1149 <tr>
1150 <td><a href="#block">block</a></td>
1151 <td class="td_left">The <a href="#constantpool">constant pool</a>
1152block for this function.<sup>2</sup></td>
1153 </tr>
1154 <tr>
1155 <td><a href="#block">block</a></td>
1156 <td class="td_left">The <a href="#compactiontable">compaction
1157table</a> block for the function.<sup>2</sup></td>
1158 </tr>
1159 <tr>
1160 <td><a href="#block">block</a></td>
1161 <td class="td_left">The <a href="#instructionlist">instruction
1162list</a> for the function.</td>
1163 </tr>
1164 <tr>
1165 <td><a href="#block">block</a></td>
Reid Spencer8996e552004-08-17 00:49:03 +0000[diff] [blame]1166 <td class="td_left">The function's <a href="#symtab">symbol
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1167table</a> containing only those symbols pertinent to the function
1168(mostly block labels).</td>
1169 </tr>
1170 </tbody>
1171</table>
1172Notes:
1173<ol>
1174 <li>Note that if the linkage type is "External" then none of the
1175other fields will be present as the function is defined elsewhere.</li>
1176 <li>Note that only one of the constant pool or compaction table will
1177be written. Compaction tables are only written if they will actually
1178save bytecode space. If not, then a regular constant pool is written.</li>
1179</ol>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1180</div>
1181<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1182<div class="doc_subsection"><a name="compactiontable">Compaction Table</a>
1183</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1184<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1185<p>Compaction tables are part of a function definition. They are merely
1186a device for reducing the size of bytecode files. The size of a
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1187bytecode file is dependent on the <em>values</em> of the slot numbers
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1188used because larger values use more bytes in the variable bit rate
1189encoding scheme. Furthermore, the compressed instruction format
1190reserves only six bits for the type of the instruction. In large
1191modules, declaring hundreds or thousands of types, the values of the
1192slot numbers can be quite large. However, functions may use only a
1193small fraction of the global types. In such cases a compaction table is
1194created that maps the global type and value slot numbers to smaller
1195values used by a function. Functions will contain either a
1196function-specific constant pool <em>or</em> a compaction table but not
1197both. Compaction tables have the format shown in the table below.</p>
1198<table>
1199 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1200 <tr>
1201 <th><b>Type</b></th>
1202 <th class="td_left"><b>Field Description</b></th>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1203 </tr>
1204 <tr>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1205 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1206 <td class="td_left">The number of types that follow</td>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1207 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1208 <tr>
1209 <td><a href="#uint24_vbr">uint24_vbr</a>+</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1210 <td class="td_left">The type slot number in the global types of
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1211the type that will be referenced in the function with the index of this
1212entry in the compaction table.</td>
1213 </tr>
1214 <tr>
1215 <td><a href="#type_len">type_len</a></td>
1216 <td class="td_left">An encoding of the type and number of values
1217that follow. This field's encoding varies depending on the size of the
1218type plane. See <a href="#type_len">Type and Length</a> for further
1219details.</td>
1220 </tr>
1221 <tr>
1222 <td><a href="#uint32_vbr">uint32_vbr</a>+</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1223 <td class="td_left">The value slot number in the global values
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1224that will be referenced in the function with the index of this entry in
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1225the compaction table.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1226 </tr>
1227 </tbody>
1228</table>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1229</div>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1230<!-- _______________________________________________________________________ -->
1231<div class="doc_subsubsection"><a name="type_len">Type and Length</a></div>
1232<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1233<p>The type and length of a compaction table type plane is encoded
1234differently depending on the length of the plane. For planes of length
12351 or 2, the length is encoded into bits 0 and 1 of a <a
1236 href="#uint32_vbr">uint32_vbr</a> and the type is encoded into bits
12372-31. Because type numbers are often small, this often saves an extra
1238byte per plane. If the length of the plane is greater than 2 then the
1239encoding uses a <a href="#uint32_vbr">uint32_vbr</a> for each of the
1240length and type, in that order.</p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1241</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1242<!-- _______________________________________________________________________ -->
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1243<div class="doc_subsection"><a name="instructionlist">Instruction List</a></div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]1244<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1245<p>The instructions in a function are written as a simple list. Basic
1246blocks are inferred by the terminating instruction types. The format of
1247the block is given in the following table.</p>
1248<table>
1249 <tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1250 <tr>
1251 <th><b>Type</b></th>
1252 <th class="td_left"><b>Field Description</b></th>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1253 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1254 <tr>
1255 <td><a href="#block">block</a><br>
1256 </td>
1257 <td class="td_left">Instruction list identifier (0x07) + size<br>
1258 </td>
1259 </tr>
1260 <tr>
1261 <td><a href="#instruction">instruction</a>+</td>
1262 <td class="td_left">An instruction. Instructions have a variety
1263of formats. See <a href="#instruction">Instructions</a> for details.</td>
1264 </tr>
1265 </tbody>
1266</table>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]1267</div>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1268<!-- _______________________________________________________________________ -->
1269<div class="doc_subsubsection"><a name="instruction">Instructions</a></div>
1270<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1271<p>For brevity, instructions are written in one of four formats,
1272depending on the number of operands to the instruction. Each
1273instruction begins with a <a href="#uint32_vbr">uint32_vbr</a> that
1274encodes the type of the instruction as well as other things. The tables
Reid Spencer8996e552004-08-17 00:49:03 +0000[diff] [blame]1275that follow describe the format of this first part of each instruction.</p>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1276<p><b>Instruction Format 0</b></p>
1277<p>This format is used for a few instructions that can't easily be
Reid Spencer8996e552004-08-17 00:49:03 +0000[diff] [blame]1278shortened because they have large numbers of operands (e.g. PHI Node or
1279getelementptr). Each of the opcode, type, and operand fields is found in
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1280successive fields.</p>
1281<table>
1282 <tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1283 <tr>
1284 <th><b>Type</b></th>
1285 <th class="td_left"><b>Field Description</b></th>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1286 </tr>
1287 <tr>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1288 <td><a href="#uint32_vbr">uint32_vbr</a></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1289 <td class="td_left">Specifies the opcode of the instruction. Note
1290that for compatibility with the other instruction formats, the opcode
1291is shifted left by 2 bits. Bits 0 and 1 must have value zero for this
1292format.</td>
1293 </tr>
1294 <tr>
1295 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1296 <td class="td_left">Provides the type slot number of the result type of
1297 the instruction.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1298 </tr>
1299 <tr>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1300 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1301 <td class="td_left">The number of operands that follow.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1302 </tr>
1303 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000[diff] [blame]1304 <td><a href="#uint32_vbr">uint32_vbr</a>+</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1305 <td class="td_left">The slot number of the value(s) for the operand(s).
1306 <sup>1</sup></td>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1307 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1308 </tbody>
1309</table>
1310Notes:
1311<ol>
1312 <li>Note that if the instruction is a getelementptr and the type of
1313the operand is a sequential type (array or pointer) then the slot
1314number is shifted up two bits and the low order bits will encode the
1315type of index used, as follows: 0=uint, 1=int, 2=ulong, 3=long.</li>
1316</ol>
1317<p><b>Instruction Format 1</b></p>
1318<p>This format encodes the opcode, type and a single operand into a
1319single <a href="#uint32_vbr">uint32_vbr</a> as follows:</p>
1320<table>
1321 <tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1322 <tr>
1323 <th><b>Bits</b></th>
1324 <th><b>Type</b></th>
1325 <th class="td_left"><b>Field Description</b></th>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1326 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1327 <tr>
1328 <td>0-1</td>
1329 <td>constant "1"</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1330 <td class="td_left">These two bits must be the value 1 which identifies
1331 this as an instruction of format 1.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1332 </tr>
1333 <tr>
1334 <td>2-7</td>
Reid Spencer8996e552004-08-17 00:49:03 +0000[diff] [blame]1335 <td><a href="#opcode">opcode</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1336 <td class="td_left">Specifies the opcode of the instruction. Note that
1337 the maximum opcode value is 63.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1338 </tr>
1339 <tr>
1340 <td>8-19</td>
1341 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1342 <td class="td_left">Specifies the slot number of the type for this
1343 instruction. Maximum slot number is 2<sup>12</sup>-1=4095.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1344 </tr>
1345 <tr>
1346 <td>20-31</td>
1347 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1348 <td class="td_left">Specifies the slot number of the value for the
1349 first operand. Maximum slot number is 2<sup>12</sup>-1=4095. Note that
1350 the value 2<sup>12</sup>-1 denotes zero operands.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1351 </tr>
1352 </tbody>
1353</table>
1354<p><b>Instruction Format 2</b></p>
1355<p>This format encodes the opcode, type and two operands into a single <a
1356 href="#uint32_vbr">uint32_vbr</a> as follows:</p>
1357<table>
1358 <tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1359 <tr>
1360 <th><b>Bits</b></th>
1361 <th><b>Type</b></th>
1362 <th class="td_left"><b>Field Description</b></th>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1363 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1364 <tr>
1365 <td>0-1</td>
1366 <td>constant "2"</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1367 <td class="td_left">These two bits must be the value 2 which identifies
1368 this as an instruction of format 2.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1369 </tr>
1370 <tr>
1371 <td>2-7</td>
1372 <td><a href="#opcodes">opcode</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1373 <td class="td_left">Specifies the opcode of the instruction. Note that
1374 the maximum opcode value is 63.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1375 </tr>
1376 <tr>
1377 <td>8-15</td>
1378 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1379 <td class="td_left">Specifies the slot number of the type for this
1380 instruction. Maximum slot number is 2<sup>8</sup>-1=255.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1381 </tr>
1382 <tr>
1383 <td>16-23</td>
1384 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1385 <td class="td_left">Specifies the slot number of the value for the first
1386 operand. Maximum slot number is 2<sup>8</sup>-1=255.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1387 </tr>
1388 <tr>
1389 <td>24-31</td>
1390 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1391 <td class="td_left">Specifies the slot number of the value for the second
1392 operand. Maximum slot number is 2<sup>8</sup>-1=255.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1393 </tr>
1394 </tbody>
1395</table>
1396<p><b>Instruction Format 3</b></p>
1397<p>This format encodes the opcode, type and three operands into a
1398single <a href="#uint32_vbr">uint32_vbr</a> as follows:</p>
1399<table>
1400 <tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1401 <tr>
1402 <th><b>Bits</b></th>
1403 <th><b>Type</b></th>
1404 <th class="td_left"><b>Field Description</b></th>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1405 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1406 <tr>
1407 <td>0-1</td>
1408 <td>constant "3"</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1409 <td class="td_left">These two bits must be the value 3 which identifies
1410 this as an instruction of format 3.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1411 </tr>
1412 <tr>
1413 <td>2-7</td>
1414 <td><a href="#opcodes">opcode</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1415 <td class="td_left">Specifies the opcode of the instruction. Note that
1416 the maximum opcode value is 63.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1417 </tr>
1418 <tr>
1419 <td>8-13</td>
1420 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1421 <td class="td_left">Specifies the slot number of the type for this
1422 instruction. Maximum slot number is 2<sup>6</sup>-1=63.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1423 </tr>
1424 <tr>
1425 <td>14-19</td>
1426 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1427 <td class="td_left">Specifies the slot number of the value for the first
1428 operand. Maximum slot number is 2<sup>6</sup>-1=63.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1429 </tr>
1430 <tr>
1431 <td>20-25</td>
1432 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1433 <td class="td_left">Specifies the slot number of the value for the second
1434 operand. Maximum slot number is 2<sup>6</sup>-1=63.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1435 </tr>
1436 <tr>
1437 <td>26-31</td>
1438 <td><a href="#unsigned">unsigned</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1439 <td class="td_left">Specifies the slot number of the value for the third
1440 operand. Maximum slot number is 2<sup>6</sup>-1=63.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1441 </tr>
1442 </tbody>
1443</table>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1444</div>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1445
1446<!-- _______________________________________________________________________ -->
1447<div class="doc_subsection"><a name="opcodes">Opcodes</a></div>
1448<div class="doc_text">
1449 <p>Instructions encode an opcode that identifies the kind of instruction.
1450 Opcodes are an enumerated integer value. The specific values used depend on
1451 the version of LLVM you're using. The opcode values are defined in the
1452 <a href="http://llvm.org/cvsweb/cvsweb.cgi/llvm/include/llvm/Instruction.def">
1453 <tt>include/llvm/Instruction.def</tt></a> file. You should check there for the
1454 most recent definitions. The table below provides the opcodes defined as of
1455 the writing of this document. The table associates each opcode mneumonic with
1456 its enumeration value and the bytecode and LLVM version numbers in which the
1457 opcode was introduced.</p>
1458 <table>
1459 <tbody>
1460 <tr>
1461 <th>Opcode</th>
1462 <th>Number</th>
1463 <th>Bytecode Version</th>
1464 <th>LLVM Version</th>
1465 </tr>
1466 <tr><td colspan="4"><b>Terminator Instructions</b></td></tr>
1467 <tr><td>Ret</td><td>1</td><td>1</td><td>1.0</td></tr>
1468 <tr><td>Br</td><td>2</td><td>1</td><td>1.0</td></tr>
1469 <tr><td>Switch</td><td>3</td><td>1</td><td>1.0</td></tr>
1470 <tr><td>Invoke</td><td>4</td><td>1</td><td>1.0</td></tr>
1471 <tr><td>Unwind</td><td>5</td><td>1</td><td>1.0</td></tr>
1472 <tr><td colspan="4"><b>Binary Operators</b></td></tr>
1473 <tr><td>Add</td><td>6</td><td>1</td><td>1.0</td></tr>
1474 <tr><td>Sub</td><td>7</td><td>1</td><td>1.0</td></tr>
1475 <tr><td>Mul</td><td>8</td><td>1</td><td>1.0</td></tr>
1476 <tr><td>Div</td><td>9</td><td>1</td><td>1.0</td></tr>
1477 <tr><td>Rem</td><td>10</td><td>1</td><td>1.0</td></tr>
1478 <tr><td colspan="4"><b>Logical Operators</b></td></tr>
1479 <tr><td>And</td><td>11</td><td>1</td><td>1.0</td></tr>
1480 <tr><td>Or</td><td>12</td><td>1</td><td>1.0</td></tr>
1481 <tr><td>Xor</td><td>13</td><td>1</td><td>1.0</td></tr>
1482 <tr><td colspan="4"><b>Binary Comparison Operators</b></td></tr>
1483 <tr><td>SetEQ</td><td>14</td><td>1</td><td>1.0</td></tr>
1484 <tr><td>SetNE</td><td>15</td><td>1</td><td>1.0</td></tr>
1485 <tr><td>SetLE</td><td>16</td><td>1</td><td>1.0</td></tr>
1486 <tr><td>SetGE</td><td>17</td><td>1</td><td>1.0</td></tr>
1487 <tr><td>SetLT</td><td>18</td><td>1</td><td>1.0</td></tr>
1488 <tr><td>SetGT</td><td>19</td><td>1</td><td>1.0</td></tr>
1489 <tr><td colspan="4"><b>Memory Operators</b></td></tr>
1490 <tr><td>Malloc</td><td>20</td><td>1</td><td>1.0</td></tr>
1491 <tr><td>Free</td><td>21</td><td>1</td><td>1.0</td></tr>
1492 <tr><td>Alloca</td><td>22</td><td>1</td><td>1.0</td></tr>
1493 <tr><td>Load</td><td>23</td><td>1</td><td>1.0</td></tr>
1494 <tr><td>Store</td><td>24</td><td>1</td><td>1.0</td></tr>
1495 <tr><td>GetElementPtr</td><td>25</td><td>1</td><td>1.0</td></tr>
1496 <tr><td colspan="4"><b>Other Operators</b></td></tr>
1497 <tr><td>PHI</td><td>26</td><td>1</td><td>1.0</td></tr>
1498 <tr><td>Cast</td><td>27</td><td>1</td><td>1.0</td></tr>
1499 <tr><td>Call</td><td>28</td><td>1</td><td>1.0</td></tr>
1500 <tr><td>Shl</td><td>29</td><td>1</td><td>1.0</td></tr>
1501 <tr><td>Shr</td><td>30</td><td>1</td><td>1.0</td></tr>
1502 <tr><td>VANext</td><td>31</td><td>1</td><td>1.0</td></tr>
1503 <tr><td>VAArg</td><td>32</td><td>1</td><td>1.0</td></tr>
1504 <tr><td>Select</td><td>33</td><td>2</td><td>1.2</td></tr>
1505 <tr><td>UserOp1</td><td>34</td><td>1</td><td>1.0</td></tr>
1506 <tr><td>UserOp2</td><td>35</td><td>1</td><td>1.0</td></tr>
1507 </tbody>
1508 </table>
1509</div>
1510
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]1511<!-- _______________________________________________________________________ -->
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]1512<div class="doc_subsection"><a name="symtab">Symbol Table</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]1513<div class="doc_text">
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1514<p>A symbol table can be put out in conjunction with a module or a function. A
1515symbol table has a list of name/type associations followed by a list of
1516name/value associations. The name/value associations are organized into "type
1517planes" so that all values of a common type are listed together. Each type
1518plane starts with the number of entries in the plane and the type slot number
1519for all the values in that plane (so the type can be looked up in the global
1520type pool). For each entry in a type plane, the slot number of the value and
1521the name associated with that value are written. The format is given in the
1522table below. </p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000[diff] [blame]1523<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1524 <tbody>
1525 <tr>
1526 <th><b>Type</b></th>
1527 <th class="td_left"><b>Field Description</b></th>
1528 </tr>
1529 <tr>
1530 <td><a href="#block">block</a><br>
1531 </td>
1532 <td class="td_left">Symbol Table Identifier (0x04)</td>
1533 </tr>
1534 <tr>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1535 <td><a href="#llist">llist</a>(<a href="#symtab_entry">type_entry</a>)</td>
Reid Spencer8996e552004-08-17 00:49:03 +0000[diff] [blame]1536 <td class="td_left">A length list of symbol table entries for
1537 <tt>Type</tt>s
1538 </td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1539 </tr>
1540 <tr>
Reid Spencer8996e552004-08-17 00:49:03 +0000[diff] [blame]1541 <td><a href="#zlist">llist</a>(<a href="#symtab_plane">symtab_plane</a>)</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1542 <td class="td_left">A length list of "type planes" of symbol table
Reid Spencer8996e552004-08-17 00:49:03 +0000[diff] [blame]1543 entries for <tt>Value</tt>s</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1544 </tr>
1545 </tbody>
Reid Spencerb39021b2004-05-23 17:05:09 +0000[diff] [blame]1546</table>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]1547</div>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1548
1549<!-- _______________________________________________________________________ -->
1550<div class="doc_subsubsection"> <a name="type_entry">Symbol Table Type
1551Entry</a>
1552</div>
1553<div class="doc_text">
1554<p>A symbol table type entry associates a name with a type. The name is provided
1555simply as an array of chars. The type is provided as a type slot number (index)
1556into the global type pool. The format is given in the following table:</p>
1557<table>
1558 <tbody>
1559 <tr>
1560 <th><b>Type</b></th>
1561 <th class="td_left"><b>Field Description</b></th>
1562 </tr>
1563 <tr>
1564 <td><a href="#uint32_vbr">uint24_vbr</a></td>
1565 <td class="td_left">Type slot number of the type being given a
1566 name relative to the global type pool.
1567 </td>
1568 </tr>
1569 <tr>
1570 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1571 <td class="td_left">Length of the character array that follows.</td>
1572 </tr>
1573 <tr>
1574 <td><a href="#char">char</a>+</td>
1575 <td class="td_left">The characters of the name.</td>
1576 </tr>
1577 </tbody>
1578</table>
1579</div>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1580<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1581<div class="doc_subsubsection"> <a name="symtab_plane">Symbol Table
1582Plane</a>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1583</div>
1584<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1585<p>A symbol table plane provides the symbol table entries for all
1586values of a common type. The encoding is given in the following table:</p>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1587<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1588 <tbody>
1589 <tr>
1590 <th><b>Type</b></th>
1591 <th class="td_left"><b>Field Description</b></th>
1592 </tr>
1593 <tr>
1594 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1595 <td class="td_left">Number of entries in this plane.</td>
1596 </tr>
1597 <tr>
1598 <td><a href="#uint32_vbr">uint32_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1599 <td class="td_left">Type slot number of type for all values in this plane..</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1600 </tr>
1601 <tr>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1602 <td><a href="#value_entry">value_entry</a>+</td>
1603 <td class="td_left">The symbol table entries for to associate values with
1604 names.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1605 </tr>
1606 </tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1607</table>
1608</div>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1609<!-- _______________________________________________________________________ -->
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1610<div class="doc_subsubsection"><a name="value_entry">Symbol Table Value
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1611Entry</a>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1612</div>
1613<div class="doc_text">
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1614<p>A symbol table value entry provides the assocation between a value and the
1615name given to the value. The value is referenced by its slot number. The
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1616format is given in the following table:</p>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1617<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1618 <tbody>
1619 <tr>
1620 <th><b>Type</b></th>
1621 <th class="td_left"><b>Field Description</b></th>
1622 </tr>
1623 <tr>
1624 <td><a href="#uint32_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1625 <td class="td_left">Value slot number of the value being given a name.
Reid Spencer8996e552004-08-17 00:49:03 +0000[diff] [blame]1626 </td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1627 </tr>
1628 <tr>
1629 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1630 <td class="td_left">Length of the character array that follows.</td>
1631 </tr>
1632 <tr>
1633 <td><a href="#char">char</a>+</td>
1634 <td class="td_left">The characters of the name.</td>
1635 </tr>
1636 </tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +0000[diff] [blame]1637</table>
1638</div>
Reid Spencer09daa632004-08-18 20:06:19 +0000[diff] [blame^]1639
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1640<!-- *********************************************************************** -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1641<div class="doc_section"> <a name="versiondiffs">Version Differences</a>
1642</div>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1643<!-- *********************************************************************** -->
1644<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1645<p>This section describes the differences in the Bytecode Format across
1646LLVM
1647versions. The versions are listed in reverse order because it assumes
1648the current version is as documented in the previous sections. Each
1649section here
Chris Lattner1cc070c2004-07-05 18:05:48 +0000[diff] [blame]1650describes the differences between that version and the one that <i>follows</i>.
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1651</p>
1652</div>
1653<!-- _______________________________________________________________________ -->
Reid Spencercf549e12004-08-17 07:43:43 +0000[diff] [blame]1654<div class="doc_subsection"><a name="vers12">Version 1.3 Differences From
1655 1.4</a></div>
1656<!-- _______________________________________________________________________ -->
1657<div class="doc_subsubsection">Aligned Data</div>
1658<div class="doc_text">
1659 <p>In version 1.3, certain data items were aligned to 32-bit boundaries. In
1660 version 1.4, alignment of data was done away with completely. The need for
1661 alignment has gone away and the only thing it adds is bytecode file size
1662 overhead. In most cases this overhead was small. However, in functions with
1663 large numbers of format 0 instructions (GEPs and PHIs with lots of parameters)
1664 or regular instructions with large valued operands (e.g. because there's just
1665 a lot of instructions in the function) the overhead can be extreme. In one
1666 test case, the overhead was 44,000 bytes (34% of the total file size).
1667 Consequently in release 1.4, the decision was made to eliminate alignment
1668 altogether.</p>
1669 <p>In version 1.3 format, the following bytecode constructs were aligned (i.e.
1670 they were followed by one to three bytes of padding):</p>
1671 <ul>
1672 <li>All blocks.</li>
1673 <li>Instructions using the long format (format 0).</li>
1674 <li>All call instructions that called a var args function.</li>
1675 <li>The target triple (a string field at the end of the module block).</li>
1676 <li>The version field (immediately following the signature).</li>
1677 </ul>
1678 <p>None of these constructs are aligned in version 1.4</p>
1679</div>
1680
1681<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1682<div class="doc_subsection"><a name="vers12">Version 1.2 Differences
1683From 1.3</a></div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1684<!-- _______________________________________________________________________ -->
1685<div class="doc_subsubsection">Type Derives From Value</div>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1686<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1687<p>In version 1.2, the Type class in the LLVM IR derives from the Value
1688class. This is not the case in version 1.3. Consequently, in version
16891.2 the notion of a "Type Type" was used to write out values that were
1690Types. The types always occuped plane 12 (corresponding to the
1691TypeTyID) of any type planed set of values. In 1.3 this representation
1692is not convenient because the TypeTyID (12) is not present and its
1693value is now used for LabelTyID. Consequently, the data structures
1694written that involve types do so by writing all the types first and
1695then each of the value planes according to those types. In version 1.2,
1696the types would have been written intermingled with the values.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1697</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1698<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1699<div class="doc_subsubsection">Restricted getelementptr Types</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1700<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1701<p>In version 1.2, the getelementptr instruction required a ubyte type
1702index for accessing a structure field and a long type index for
1703accessing an array element. Consequently, it was only possible to
1704access structures of 255 or fewer elements. Starting in version 1.3,
1705this restriction was lifted. Structures must now be indexed with uint
1706constants. Arrays may now be indexed with int, uint, long, or ulong
1707typed values. The consequence of this was that the bytecode format had
1708to change in order to accommodate the larger range of structure indices.</p>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1709</div>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1710<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1711<div class="doc_subsubsection">Short Block Headers</div>
1712<div class="doc_text">
1713<p>In version 1.2, block headers were always 8 bytes being comprised of
1714both an unsigned integer type and an unsigned integer size. For very
1715small modules, these block headers turn out to be a large fraction of
1716the total bytecode file size. In an attempt to make these small files
1717smaller, the type and size information was encoded into a single
1718unsigned integer (4 bytes) comprised of 5 bits for the block type
1719(maximum 31 block types) and 27 bits for the block size (max
1720~134MBytes). These limits seemed sufficient for any blocks or sizes
1721forseen in the future. Note that the module block, which encloses all
1722the other blocks is still written as 8 bytes since bytecode files
1723larger than 134MBytes might be possible.</p>
1724</div>
1725<!-- _______________________________________________________________________ -->
1726<div class="doc_subsubsection">Dependent Libraries and Target Triples</div>
1727<div class="doc_text">
1728<p>In version 1.2, the bytecode format does not store module's target
1729triple or dependent. These fields have been added to the end of the <a
1730 href="#globalinfo">module global info block</a>. The purpose of these
1731fields is to allow a front end compiler to specifiy that the generated
1732module is specific to a particular target triple (operating
1733system/manufacturer/processor) which makes it non-portable; and to
1734allow front end compilers to specify the list of libraries that the
1735module depends on for successful linking.</p>
1736</div>
1737<!-- _______________________________________________________________________ -->
1738<div class="doc_subsubsection">Types Restricted to 24-bits</div>
1739<div class="doc_text">
1740<p>In version 1.2, type slot identifiers were written as 32-bit VBR
1741quantities. In 1.3 this has been reduced to 24-bits in order to ensure
1742that it is not possible to overflow the type field of a global variable
1743definition. 24-bits for type slot numbers is deemed sufficient for any
1744practical use of LLVM.</p>
1745</div>
1746<!-- _______________________________________________________________________ -->
1747<!-- _______________________________________________________________________ -->
1748<div class="doc_subsection"><a name="vers11">Version 1.1 Differences
1749From 1.2 </a></div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1750<!-- _______________________________________________________________________ -->
1751<div class="doc_subsubsection">Explicit Primitive Zeros</div>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1752<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1753<p>In version 1.1, the zero value for primitives was explicitly encoded
1754into the bytecode format. Since these zero values are constant values
1755in the LLVM IR and never change, there is no reason to explicitly
1756encode them. This explicit encoding was removed in version 1.2.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1757</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1758<!-- _______________________________________________________________________ -->
1759<div class="doc_subsubsection">Inconsistent Module Global Info</div>
1760<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1761<p>In version 1.1, the Module Global Info block was not aligned causing
1762the next block to be read in on an unaligned boundary. This problem was
1763corrected in version 1.2.<br>
1764<br>
1765</p>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1766</div>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1767<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1768<div class="doc_subsection"><a name="vers10">Version 1.0 Differences
1769From 1.1</a></div>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1770<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +0000[diff] [blame]1771<p>None. Version 1.0 and 1.1 bytecode formats are identical.</p>
Reid Spencer7c76d332004-06-08 07:41:41 +0000[diff] [blame]1772</div>
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]1773<!-- *********************************************************************** -->
1774<hr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1775<address> <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
1776 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
1777<a href="http://validator.w3.org/check/referer"><img
1778 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
1779<a href="mailto:rspencer@x10sys.com">Reid Spencer</a> and <a
1780 href="mailto:sabre@nondot.org">Chris Lattner</a><br>
1781<a href="http://llvm.cs.uiuc.edu">The LLVM Compiler Infrastructure</a><br>
1782Last modified: $Date$
Reid Spencer50026612004-05-22 02:28:36 +0000[diff] [blame]1783</address>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000[diff] [blame]1784</body>
1785</html>