blob: d6b555242d75343313068acbc8bf9f7b136f417a [file] [log] [blame]
Reid Spencer9bd2be22004-07-29 00:13:04 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
Reid Spencer50026612004-05-22 02:28:36 +00002<html>
3<head>
Reid Spencer09daa632004-08-18 20:06:19 +00004 <meta http-equiv="Content-Type" content="text/html; charset=utf-8">
Reid Spencer50026612004-05-22 02:28:36 +00005 <title>LLVM Bytecode File Format</title>
6 <link rel="stylesheet" href="llvm.css" type="text/css">
Reid Spencer1ab929c2004-07-05 08:18:07 +00007 <style type="text/css">
Reid Spencer2de29992004-08-03 20:21:05 +00008 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 +000010 TH { border: 2px solid gray; font-weight: bold; font-size: 105%; }
Reid Spencer2cc36152004-07-05 19:04:27 +000011 TABLE { text-align: center; border: 2px solid black;
Reid Spencer2de29992004-08-03 20:21:05 +000012 border-collapse: collapse; margin-top: 1em; margin-left: 1em;
13 margin-right: 1em; margin-bottom: 1em; }
Reid Spencer2cc36152004-07-05 19:04:27 +000014 .td_left { border: 2px solid gray; text-align: left; }
Reid Spencer50026612004-05-22 02:28:36 +000015 </style>
16</head>
17<body>
Reid Spencer9bd2be22004-07-29 00:13:04 +000018<div class="doc_title"> LLVM Bytecode File Format </div>
Reid Spencer50026612004-05-22 02:28:36 +000019<ol>
20 <li><a href="#abstract">Abstract</a></li>
Reid Spencer1ab929c2004-07-05 08:18:07 +000021 <li><a href="#concepts">Concepts</a>
Reid Spencer50026612004-05-22 02:28:36 +000022 <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 +000027 <li><a href="#vbr">Variable Bit-Rate Encoding</a></li>
Reid Spencer1ab929c2004-07-05 08:18:07 +000028 <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 +000032 <li><a href="#general">General Structure</a> </li>
33 <li><a href="#blockdefs">Block Definitions</a>
Reid Spencer1ab929c2004-07-05 08:18:07 +000034 <ol>
Reid Spencerb39021b2004-05-23 17:05:09 +000035 <li><a href="#signature">Signature Block</a></li>
36 <li><a href="#module">Module Block</a></li>
Reid Spencer1ab929c2004-07-05 08:18:07 +000037 <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>
Chris Lattnerf97fa592005-11-05 22:20:06 +000041 <li><a href="#instructionlist">Instructions List</a></li>
42 <li><a href="#instructions">Instructions</a></li>
Reid Spencer1ab929c2004-07-05 08:18:07 +000043 <li><a href="#symtab">Symbol Table</a></li>
Reid Spencer50026612004-05-22 02:28:36 +000044 </ol>
45 </li>
Reid Spencer7c76d332004-06-08 07:41:41 +000046 <li><a href="#versiondiffs">Version Differences</a>
47 <ol>
Chris Lattnera1dad812004-10-16 18:03:55 +000048 <li><a href="#vers13">Version 1.3 Differences From 1.4</a></li>
Reid Spencer7c76d332004-06-08 07:41:41 +000049 <li><a href="#vers12">Version 1.2 Differences From 1.3</a></li>
50 <li><a href="#vers11">Version 1.1 Differences From 1.2</a></li>
51 <li><a href="#vers10">Version 1.0 Differences From 1.1</a></li>
52 </ol>
53 </li>
Reid Spencer50026612004-05-22 02:28:36 +000054</ol>
Chris Lattner8dabb502004-05-25 17:44:58 +000055<div class="doc_author">
56<p>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a>
57</p>
Reid Spencer50026612004-05-22 02:28:36 +000058</div>
59<!-- *********************************************************************** -->
60<div class="doc_section"> <a name="abstract">Abstract </a></div>
61<!-- *********************************************************************** -->
62<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +000063<p>This document describes the LLVM bytecode file format. It specifies
64the binary encoding rules of the bytecode file format so that
65equivalent systems can encode bytecode files correctly. The LLVM
66bytecode representation is used to store the intermediate
67representation on disk in compacted form.</p>
68<p>The LLVM bytecode format may change in the future, but LLVM will
69always be backwards compatible with older formats. This document will
70only describe the most current version of the bytecode format. See <a
71 href="#versiondiffs">Version Differences</a> for the details on how
72the current version is different from previous versions.</p>
Reid Spencer50026612004-05-22 02:28:36 +000073</div>
74<!-- *********************************************************************** -->
Reid Spencer1ab929c2004-07-05 08:18:07 +000075<div class="doc_section"> <a name="concepts">Concepts</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +000076<!-- *********************************************************************** -->
77<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +000078<p>This section describes the general concepts of the bytecode file
79format without getting into specific layout details. It is recommended
80that you read this section thoroughly before interpreting the detailed
81descriptions.</p>
Reid Spencer50026612004-05-22 02:28:36 +000082</div>
83<!-- _______________________________________________________________________ -->
84<div class="doc_subsection"><a name="blocks">Blocks</a> </div>
85<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +000086<p>LLVM bytecode files consist simply of a sequence of blocks of bytes
87using a binary encoding Each block begins with an header of two
88unsigned integers. The first value identifies the type of block and the
89second value provides the size of the block in bytes. The block
90identifier is used because it is possible for entire blocks to be
91omitted from the file if they are empty. The block identifier helps the
92reader determine which kind of block is next in the file. Note that
93blocks can be nested within other blocks.</p>
94<p> All blocks are variable length, and the block header specifies the
95size of the block. All blocks begin on a byte index that is aligned to
96an even 32-bit boundary. That is, the first block is 32-bit aligned
97because it starts at offset 0. Each block is padded with zero fill
98bytes to ensure that the next block also starts on a 32-bit boundary.</p>
Reid Spencer50026612004-05-22 02:28:36 +000099</div>
100<!-- _______________________________________________________________________ -->
101<div class="doc_subsection"><a name="lists">Lists</a> </div>
102<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000103<p>LLVM Bytecode blocks often contain lists of things of a similar
104type. For example, a function contains a list of instructions and a
105function type contains a list of argument types. There are two basic
106types of lists: length lists (<a href="#llist">llist</a>), and null
107terminated lists (<a href="#zlist">zlist</a>), as described below in
108the <a href="#encoding">Encoding Primitives</a>.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000109</div>
110<!-- _______________________________________________________________________ -->
111<div class="doc_subsection"><a name="fields">Fields</a> </div>
112<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000113<p>Fields are units of information that LLVM knows how to write atomically. Most
114fields have a uniform length or some kind of length indication built into their
115encoding. For example, a constant string (array of bytes) is written simply as
116the length followed by the characters. Although this is similar to a list,
117constant strings are treated atomically and are thus fields.</p>
Reid Spencer50026612004-05-22 02:28:36 +0000118<p>Fields use a condensed bit format specific to the type of information
119they must contain. As few bits as possible are written for each field. The
Reid Spencer9bd2be22004-07-29 00:13:04 +0000120sections that follow will provide the details on how these fields are
Reid Spencer50026612004-05-22 02:28:36 +0000121written and how the bits are to be interpreted.</p>
122</div>
123<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000124<div class="doc_subsection"><a name="align">Alignment</a> </div>
Reid Spencer7aa940d2004-05-25 15:47:57 +0000125<div class="doc_text">
Reid Spencer267660f2004-08-03 20:33:56 +0000126 <p>To support cross-platform differences, the bytecode file is aligned on
127 certain boundaries. This means that a small amount of padding (at most 3
128 bytes) will be added to ensure that the next entry is aligned to a 32-bit
129 boundary.</p>
Chris Lattner8dabb502004-05-25 17:44:58 +0000130</div>
Reid Spencer7aa940d2004-05-25 15:47:57 +0000131<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000132<div class="doc_subsection"><a name="vbr">Variable Bit-Rate Encoding</a>
Reid Spencer82c46712004-07-07 13:34:26 +0000133</div>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000134<div class="doc_text">
135<p>Most of the values written to LLVM bytecode files are small integers. To
136minimize the number of bytes written for these quantities, an encoding scheme
137similar to UTF-8 is used to write integer data. The scheme is known as
138variable bit rate (vbr) encoding. In this encoding, the high bit of
139each byte is used to indicate if more bytes follow. If (byte &amp;
1400x80) is non-zero in any given byte, it means there is another byte
141immediately following that also contributes to the value. For the final
142byte (byte &amp; 0x80) is false (the high bit is not set). In each byte
143only the low seven bits contribute to the value. Consequently 32-bit
144quantities can take from one to <em>five</em> bytes to encode. In
145general, smaller quantities will encode in fewer bytes, as follows:</p>
146<table>
147 <tbody>
148 <tr>
149 <th>Byte #</th>
150 <th>Significant Bits</th>
151 <th>Maximum Value</th>
152 </tr>
153 <tr>
154 <td>1</td>
155 <td>0-6</td>
156 <td>127</td>
157 </tr>
158 <tr>
159 <td>2</td>
160 <td>7-13</td>
161 <td>16,383</td>
162 </tr>
163 <tr>
164 <td>3</td>
165 <td>14-20</td>
166 <td>2,097,151</td>
167 </tr>
168 <tr>
169 <td>4</td>
170 <td>21-27</td>
171 <td>268,435,455</td>
172 </tr>
173 <tr>
174 <td>5</td>
175 <td>28-34</td>
176 <td>34,359,738,367</td>
177 </tr>
178 <tr>
179 <td>6</td>
180 <td>35-41</td>
181 <td>4,398,046,511,103</td>
182 </tr>
183 <tr>
184 <td>7</td>
185 <td>42-48</td>
186 <td>562,949,953,421,311</td>
187 </tr>
188 <tr>
189 <td>8</td>
190 <td>49-55</td>
191 <td>72,057,594,037,927,935</td>
192 </tr>
193 <tr>
194 <td>9</td>
195 <td>56-62</td>
196 <td>9,223,372,036,854,775,807</td>
197 </tr>
198 <tr>
199 <td>10</td>
200 <td>63-69</td>
201 <td>1,180,591,620,717,411,303,423</td>
202 </tr>
203 </tbody>
204</table>
205<p>Note that in practice, the tenth byte could only encode bit 63 since
206the maximum quantity to use this encoding is a 64-bit integer.</p>
207<p><em>Signed</em> VBR values are encoded with the standard vbr
208encoding, but with the sign bit as the low order bit instead of the
209high order bit. This allows small negative quantities to be encoded
210efficiently. For example, -3
211is encoded as "((3 &lt;&lt; 1) | 1)" and 3 is encoded as "(3 &lt;&lt;
2121) | 0)", emitted with the standard vbr encoding above.</p>
213</div>
Reid Spencer82c46712004-07-07 13:34:26 +0000214<!-- _______________________________________________________________________ -->
215<div class="doc_subsection"><a name="encoding">Encoding Primitives</a> </div>
216<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000217<p>Each field in the bytecode format is encoded into the file using a
218small set of primitive formats. The table below defines the encoding
219rules for the various primitives used and gives them each a type name.
220The type names used in the descriptions of blocks and fields in the <a
221 href="#details">Detailed Layout</a>next section. Any type name with
222the suffix <em>_vbr</em> indicates a quantity that is encoded using
223variable bit rate encoding as described above.</p>
224<table class="doc_table">
225 <tbody>
226 <tr>
227 <th><b>Type</b></th>
228 <th class="td_left"><b>Rule</b></th>
229 </tr>
230 <tr>
231 <td><a name="unsigned"><b>unsigned</b></a></td>
232 <td class="td_left">A 32-bit unsigned integer that always occupies four
Reid Spencerb39021b2004-05-23 17:05:09 +0000233 consecutive bytes. The unsigned integer is encoded using LSB first
234 ordering. That is bits 2<sup>0</sup> through 2<sup>7</sup> are in the
235 byte with the lowest file offset (little endian).</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000236 </tr>
237 <tr>
Reid Spencer301fe482004-08-03 20:57:56 +0000238 <td style="vertical-align: top;"><a name="uint24_vbr">
239 <b>uint24_vbr</b></a></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000240 <td style="vertical-align: top; text-align: left;">A 24-bit unsigned
241 integer that occupies from one to four bytes using variable bit rate
242 encoding.</td>
243 </tr>
244 <tr>
245 <td><a name="uint32_vbr"><b>uint32_vbr</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000246 <td class="td_left">A 32-bit unsigned integer that occupies from one to
247 five bytes using variable bit rate encoding.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000248 </tr>
249 <tr>
250 <td><a name="uint64_vbr"><b>uint64_vbr</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000251 <td class="td_left">A 64-bit unsigned integer that occupies from one to ten
252 bytes using variable bit rate encoding.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000253 </tr>
254 <tr>
255 <td><a name="int64_vbr"><b>int64_vbr</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000256 <td class="td_left">A 64-bit signed integer that occupies from one to ten
257 bytes using the signed variable bit rate encoding.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000258 </tr>
259 <tr>
260 <td><a name="char"><b>char</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000261 <td class="td_left">A single unsigned character encoded into one byte</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000262 </tr>
263 <tr>
264 <td><a name="bit"><b>bit(n-m)</b></a></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000265 <td class="td_left">A set of bit within some larger integer field. The values
266 of <code>n</code> and <code>m</code> specify the inclusive range of bits
267 that define the subfield. The value for <code>m</code> may be omitted if
268 its the same as <code>n</code>.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000269 </tr>
270 <tr>
271 <td style="vertical-align: top;"><b><a name="float"><b>float</b></a></b></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000272 <td style="vertical-align: top; text-align: left;">A floating point value encoded
273 as a 32-bit IEEE value written in little-endian form.<br>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000274 </td>
275 </tr>
276 <tr>
277 <td style="vertical-align: top;"><b><b><a name="double"><b>double</b></a></b></b></td>
Reid Spencerf08561f2004-08-03 19:20:18 +0000278 <td style="vertical-align: top; text-align: left;">A floating point value encoded
279 as a64-bit IEEE value written in little-endian form</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000280 </tr>
281 <tr>
282 <td><a name="string"><b>string</b></a></td>
283 <td class="td_left">A uint32_vbr indicating the type of the
284constant string which also includes its length, immediately followed by
285the characters of the string. There is no terminating null byte in the
286string.</td>
287 </tr>
288 <tr>
289 <td><a name="data"><b>data</b></a></td>
290 <td class="td_left">An arbitrarily long segment of data to which
291no interpretation is implied. This is used for constant initializers.<br>
292 </td>
293 </tr>
294 <tr>
295 <td><a name="llist"><b>llist(x)</b></a></td>
296 <td class="td_left">A length list of x. This means the list is
297encoded as an <a href="#uint32_vbr">uint32_vbr</a> providing the
298length of the list, followed by a sequence of that many "x" items. This
299implies that the reader should iterate the number of times provided by
300the length.</td>
301 </tr>
302 <tr>
303 <td><a name="zlist"><b>zlist(x)</b></a></td>
304 <td class="td_left">A zero-terminated list of x. This means the
305list is encoded as a sequence of an indeterminate number of "x" items,
306followed by an <a href="#uint32_vbr">uint32_vbr</a> terminating value.
307This implies that none of the "x" items can have a zero value (or else
308the list terminates).</td>
309 </tr>
310 <tr>
311 <td><a name="block"><b>block</b></a></td>
312 <td class="td_left">A block of data that is logically related. A
313block is an unsigned 32-bit integer that encodes the type of the block
314in the low 5 bits and the size of the block in the high 27 bits. The
315length does not include the block header or any alignment bytes at the
316end of the block. Blocks may compose other blocks. </td>
317 </tr>
318 </tbody>
Reid Spencerb39021b2004-05-23 17:05:09 +0000319</table>
320</div>
321<!-- _______________________________________________________________________ -->
Reid Spencer82c46712004-07-07 13:34:26 +0000322<div class="doc_subsection"><a name="notation">Field Notation</a> </div>
323<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000324<p>In the detailed block and field descriptions that follow, a regex
325like notation is used to describe optional and repeated fields. A very
326limited subset of regex is used to describe these, as given in the
327following table: </p>
328<table class="doc_table">
329 <tbody>
Reid Spencer82c46712004-07-07 13:34:26 +0000330 <tr>
331 <th><b>Character</b></th>
332 <th class="td_left"><b>Meaning</b></th>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000333 </tr>
334 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000335 <td><b><code>?</code></b></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000336 <td class="td_left">The question mark indicates 0 or 1
337occurrences of the thing preceding it.</td>
338 </tr>
339 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000340 <td><b><code>*</code></b></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000341 <td class="td_left">The asterisk indicates 0 or more occurrences
342of the thing preceding it.</td>
343 </tr>
344 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000345 <td><b><code>+</code></b></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000346 <td class="td_left">The plus sign indicates 1 or more occurrences
347of the thing preceding it.</td>
348 </tr>
349 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +0000350 <td><b><code>()</code></b></td>
351 <td class="td_left">Parentheses are used for grouping.</td>
Reid Spencer82c46712004-07-07 13:34:26 +0000352 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000353 <tr>
354 <td><b><code>,</code></b></td>
355 <td class="td_left">The comma separates sequential fields.</td>
356 </tr>
357 </tbody>
358</table>
359<p>So, for example, consider the following specifications:</p>
360<div class="doc_code">
361<ol>
362 <li><code>string?</code></li>
363 <li><code>(uint32_vbr,uin32_vbr)+</code></li>
364 <li><code>(unsigned?,uint32_vbr)*</code></li>
365 <li><code>(llist(unsigned))?</code></li>
366</ol>
Reid Spencer82c46712004-07-07 13:34:26 +0000367</div>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000368<p>with the following interpretations:</p>
369<ol>
370 <li>An optional string. Matches either nothing or a single string</li>
371 <li>One or more pairs of uint32_vbr.</li>
372 <li>Zero or more occurrences of either an unsigned followed by a
373uint32_vbr or just a uint32_vbr.</li>
374 <li>An optional length list of unsigned values.</li>
375</ol>
376</div>
Reid Spencer82c46712004-07-07 13:34:26 +0000377<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000378<div class="doc_subsection"><a name="slots">Slots</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000379<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000380<p>The bytecode format uses the notion of a "slot" to reference Types
381and Values. Since the bytecode file is a <em>direct</em> representation of
382LLVM's intermediate representation, there is a need to represent pointers in
383the file. Slots are used for this purpose. For example, if one has the following
384assembly:
Reid Spencer1ab929c2004-07-05 08:18:07 +0000385</p>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000386<div class="doc_code"><code> %MyType = type { int, sbyte }<br>
387%MyVar = external global %MyType
Reid Spencer82c46712004-07-07 13:34:26 +0000388</code></div>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000389<p>there are two definitions. The definition of <tt>%MyVar</tt> uses <tt>%MyType</tt>.
390In the C++ IR this linkage between <tt>%MyVar</tt> and <tt>%MyType</tt>
391is explicit through the use of C++ pointers. In bytecode, however, there's no
392ability to store memory addresses. Instead, we compute and write out
393slot numbers for every Type and Value written to the file.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000394<p>A slot number is simply an unsigned 32-bit integer encoded in the variable
395bit rate scheme (see <a href="#encoding">encoding</a>). This ensures that
396low slot numbers are encoded in one byte. Through various bits of magic LLVM
397attempts to always keep the slot numbers low. The first attempt is to associate
Reid Spencer9bd2be22004-07-29 00:13:04 +0000398slot numbers with their "type plane". That is, Values of the same type
399are written to the bytecode file in a list (sequentially). Their order in
400that list determines their slot number. This means that slot #1 doesn't mean
401anything unless you also specify for which type you want slot #1. Types are
Reid Spencer09daa632004-08-18 20:06:19 +0000402always written to the file first (in the <a href="#globaltypes">Global Type
403Pool</a>) and in such a way that both forward and backward references of the
404types can often be resolved with a single pass through the type pool. </p>
Reid Spencerc4a8d392007-01-31 00:26:08 +0000405<p>In summary then, a slot number can be thought of as just a vbr encoded index
Reid Spencer09daa632004-08-18 20:06:19 +0000406into a list of Type* or Value*. To keep slot numbers low, Value* are indexed by
407two slot numbers: the "type plane index" (type slot) and the "value index"
408(value slot).</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000409</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000410<!-- *********************************************************************** -->
Reid Spencer51f31e02004-07-05 22:28:02 +0000411<div class="doc_section"> <a name="general">General Structure</a> </div>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000412<!-- *********************************************************************** -->
413<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000414<p>This section provides the general structure of the LLVM bytecode
415file format. The bytecode file format requires blocks to be in a
416certain order and nested in a particular way so that an LLVM module can
417be constructed efficiently from the contents of the file. This ordering
418defines a general structure for bytecode files as shown below. The
419table below shows the order in which all block types may appear. Please
420note that some of the blocks are optional and some may be repeated. The
421structure is fairly loose because optional blocks, if empty, are
422completely omitted from the file.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000423<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000424 <tbody>
425 <tr>
426 <th>ID</th>
427 <th>Parent</th>
428 <th>Optional?</th>
429 <th>Repeated?</th>
430 <th>Level</th>
431 <th>Block Type</th>
432 <th>Description</th>
433 </tr>
434 <tr>
435 <td>N/A</td>
436 <td>File</td>
437 <td>No</td>
438 <td>No</td>
439 <td>0</td>
440 <td class="td_left"><a href="#signature">Signature</a></td>
441 <td class="td_left">This contains the file signature (magic
442number) that identifies the file as LLVM bytecode.</td>
443 </tr>
444 <tr>
445 <td>0x01</td>
446 <td>File</td>
447 <td>No</td>
448 <td>No</td>
449 <td>0</td>
450 <td class="td_left"><a href="#module">Module</a></td>
451 <td class="td_left">This is the top level block in a bytecode
452file. It contains all the other blocks. </td>
453 </tr>
454 <tr>
455 <td>0x06</td>
456 <td>Module</td>
457 <td>No</td>
458 <td>No</td>
459 <td>1</td>
460 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#globaltypes">Global&nbsp;Type&nbsp;Pool</a></td>
461 <td class="td_left">This block contains all the global (module)
462level types.</td>
463 </tr>
464 <tr>
465 <td>0x05</td>
466 <td>Module</td>
467 <td>No</td>
468 <td>No</td>
469 <td>1</td>
470 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#globalinfo">Module&nbsp;Globals&nbsp;Info</a></td>
471 <td class="td_left">This block contains the type, constness, and
472linkage for each of the global variables in the module. It also
473contains the type of the functions and the constant initializers.</td>
474 </tr>
475 <tr>
476 <td>0x03</td>
477 <td>Module</td>
478 <td>Yes</td>
479 <td>No</td>
480 <td>1</td>
481 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#constantpool">Module&nbsp;Constant&nbsp;Pool</a></td>
482 <td class="td_left">This block contains all the global constants
483except function arguments, global values and constant strings.</td>
484 </tr>
485 <tr>
486 <td>0x02</td>
487 <td>Module</td>
488 <td>Yes</td>
489 <td>Yes</td>
490 <td>1</td>
491 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#functiondefs">Function&nbsp;Definitions</a>*</td>
492 <td class="td_left">One function block is written for each
493function in the module. The function block contains the instructions,
Reid Spencerc4a8d392007-01-31 00:26:08 +0000494type constant pool, and symbol table for the function.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000495 </tr>
496 <tr>
497 <td>0x03</td>
498 <td>Function</td>
499 <td>Yes</td>
500 <td>No</td>
501 <td>2</td>
502 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a
503 href="#constantpool">Function&nbsp;Constant&nbsp;Pool</a></td>
504 <td class="td_left">Any constants (including types) used solely
505within the function are emitted here in the function constant pool. </td>
506 </tr>
507 <tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000508 <td>0x07</td>
509 <td>Function</td>
510 <td>No</td>
511 <td>No</td>
512 <td>2</td>
513 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a
514 href="#instructionlist">Instruction&nbsp;List</a></td>
515 <td class="td_left">This block contains all the instructions of
516the function. The basic blocks are inferred by terminating
517instructions. </td>
518 </tr>
519 <tr>
520 <td>0x04</td>
521 <td>Function</td>
522 <td>Yes</td>
523 <td>No</td>
524 <td>2</td>
525 <td class="td_left">&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;<a
526 href="#symtab">Function&nbsp;Symbol&nbsp;Table</a></td>
527 <td class="td_left">This symbol table provides the names for the
528function specific values used (basic block labels mostly).</td>
529 </tr>
530 <tr>
531 <td>0x04</td>
532 <td>Module</td>
533 <td>Yes</td>
534 <td>No</td>
535 <td>1</td>
536 <td class="td_left">&nbsp;&nbsp;&nbsp;<a href="#symtab">Module&nbsp;Symbol&nbsp;Table</a></td>
537 <td class="td_left">This symbol table provides the names for the
538various entries in the file that are not function specific (global
539vars, and functions mostly).</td>
540 </tr>
541 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000542</table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000543<p>Use the links in the table for details about the contents of each of
544the block types.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000545</div>
Reid Spencer50026612004-05-22 02:28:36 +0000546<!-- *********************************************************************** -->
Reid Spencer51f31e02004-07-05 22:28:02 +0000547<div class="doc_section"> <a name="blockdefs">Block Definitions</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000548<!-- *********************************************************************** -->
549<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000550<p>This section provides the detailed layout of the individual block
551types in the LLVM bytecode file format. </p>
Reid Spencer50026612004-05-22 02:28:36 +0000552</div>
Reid Spencer50026612004-05-22 02:28:36 +0000553<!-- _______________________________________________________________________ -->
Reid Spencerb39021b2004-05-23 17:05:09 +0000554<div class="doc_subsection"><a name="signature">Signature Block</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000555<div class="doc_text">
Chris Lattner2b905652004-05-24 05:35:17 +0000556<p>The signature occurs in every LLVM bytecode file and is always first.
Reid Spencerb39021b2004-05-23 17:05:09 +0000557It simply provides a few bytes of data to identify the file as being an LLVM
558bytecode file. This block is always four bytes in length and differs from the
559other blocks because there is no identifier and no block length at the start
560of the block. Essentially, this block is just the "magic number" for the file.
Reid Spencer9bd2be22004-07-29 00:13:04 +0000561</p>
Reid Spencerda794e72004-11-08 08:55:21 +0000562<p>There are two types of signatures for LLVM bytecode: uncompressed and
563compressed as shown in the table below. </p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000564<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000565 <tbody>
566 <tr>
567 <th><b>Type</b></th>
Reid Spencerda794e72004-11-08 08:55:21 +0000568 <th class="td_left"><b>Uncompressed</b></th>
569 <th class="td_left"><b>Compressed</b></th>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000570 </tr>
571 <tr>
572 <td><a href="#char">char</a></td>
573 <td class="td_left">Constant "l" (0x6C)</td>
Reid Spencerda794e72004-11-08 08:55:21 +0000574 <td class="td_left">Constant "l" (0x6C)</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000575 </tr>
576 <tr>
577 <td><a href="#char">char</a></td>
578 <td class="td_left">Constant "l" (0x6C)</td>
Reid Spencerda794e72004-11-08 08:55:21 +0000579 <td class="td_left">Constant "l" (0x6C)</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000580 </tr>
581 <tr>
582 <td><a href="#char">char</a></td>
583 <td class="td_left">Constant "v" (0x76)</td>
Reid Spencerda794e72004-11-08 08:55:21 +0000584 <td class="td_left">Constant "v" (0x76)</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000585 </tr>
586 <tr>
587 <td><a href="#char">char</a></td>
588 <td class="td_left">Constant "m" (0x6D)</td>
Reid Spencerda794e72004-11-08 08:55:21 +0000589 <td class="td_left">Constant "c" (0x63)</td>
590 </tr>
591 <tr>
592 <td><a href="#char">char</a></td>
593 <td class="td_left">N/A</td>
594 <td class="td_left">'0'=null,'1'=gzip,'2'=bzip2</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000595 </tr>
596 </tbody>
Reid Spencerb39021b2004-05-23 17:05:09 +0000597</table>
Reid Spencerda794e72004-11-08 08:55:21 +0000598<p>In other words, the uncompressed signature is just the characters 'llvm'
599while the compressed signature is the characters 'llvc' followed by an ascii
600digit ('0', '1', or '2') that indicates the kind of compression used. A value of
601'0' indicates that null compression was used. This can happen when compression
602was requested on a platform that wasn't configured for gzip or bzip2. A value of
603'1' means that the rest of the file is compressed using the gzip algorithm and
604should be uncompressed before interpretation. A value of '2' means that the rest
605of the file is compressed using the bzip2 algorithm and should be uncompressed
606before interpretation. In all cases, the data resulting from uncompression
607should be interpreted as if it occurred immediately after the 'llvm'
608signature (i.e. the uncompressed data begins with the
609<a href="#module">Module Block</a></p>
610<p><b>NOTE:</b> As of LLVM 1.4, all bytecode files produced by the LLVM tools
Reid Spencera4213b22004-11-08 09:10:50 +0000611are compressed by default. To disable compression, pass the
Reid Spencerda794e72004-11-08 08:55:21 +0000612<tt>--disable-compression</tt> option to the tool, if it supports it.
Reid Spencerb39021b2004-05-23 17:05:09 +0000613</div>
614<!-- _______________________________________________________________________ -->
615<div class="doc_subsection"><a name="module">Module Block</a> </div>
616<div class="doc_text">
617<p>The module block contains a small pre-amble and all the other blocks in
Reid Spencer1ab929c2004-07-05 08:18:07 +0000618the file. The table below shows the structure of the module block. Note that it
619only provides the module identifier, size of the module block, and the format
620information. Everything else is contained in other blocks, described in other
621sections.</p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000622<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000623 <tbody>
624 <tr>
625 <th><b>Type</b></th>
626 <th class="td_left"><b>Field Description</b></th>
627 </tr>
628 <tr>
Reid Spencer5bc74d52004-08-16 19:24:36 +0000629 <td><a href="#unsigned">unsigned</a><br></td>
630 <td class="td_left"><a href="#mod_header">Module Block Identifier
631 (0x01)</a></td>
632 </tr>
633 <tr>
634 <td><a href="#unsigned">unsigned</a></td>
635 <td class="td_left"><a href="#mod_header">Module Block Size</a></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000636 </tr>
637 <tr>
638 <td><a href="#uint32_vbr">uint32_vbr</a></td>
639 <td class="td_left"><a href="#format">Format Information</a></td>
640 </tr>
641 <tr>
642 <td><a href="#block">block</a></td>
643 <td class="td_left"><a href="#globaltypes">Global Type Pool</a></td>
644 </tr>
645 <tr>
646 <td><a href="#block">block</a></td>
647 <td class="td_left"><a href="#globalinfo">Module Globals Info</a></td>
648 </tr>
649 <tr>
650 <td><a href="#block">block</a></td>
651 <td class="td_left"><a href="#constantpool">Module Constant Pool</a></td>
652 </tr>
653 <tr>
654 <td><a href="#block">block</a>*</td>
655 <td class="td_left"><a href="#functiondefs">Function Definitions</a></td>
656 </tr>
657 <tr>
658 <td><a href="#block">block</a></td>
Reid Spencer8996e552004-08-17 00:49:03 +0000659 <td class="td_left"><a href="#symtab">Module Symbol Table</a></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000660 </tr>
661 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000662</table>
663</div>
Reid Spencer5bc74d52004-08-16 19:24:36 +0000664
665<!-- _______________________________________________________________________ -->
666<div class="doc_subsubsection"><a name="mod_header">Module Block Header</a></div>
667<div class="doc_text">
668 <p>The block header for the module block uses a longer format than the other
669 blocks in a bytecode file. Specifically, instead of encoding the type and size
670 of the block into a 32-bit integer with 5-bits for type and 27-bits for size,
671 the module block header uses two 32-bit unsigned values, one for type, and one
672 for size. While the 2<sup>27</sup> byte limit on block size is sufficient for the blocks
673 contained in the module, it isn't sufficient for the module block itself
674 because we want to ensure that bytecode files as large as 2<sup>32</sup> bytes
675 are possible. For this reason, the module block (and only the module block)
676 uses a long format header.</p>
677</div>
678
Reid Spencer1ab929c2004-07-05 08:18:07 +0000679<!-- _______________________________________________________________________ -->
680<div class="doc_subsubsection"><a name="format">Format Information</a></div>
681<div class="doc_text">
Owen Andersond753fbc2007-01-27 19:23:25 +0000682<p>The format information field is encoded into a <a href="#uint32_vbr">uint32_vbr</a>.</p>
683
684<p>Of particular note, the bytecode format number is simply a 32-bit
John Criswell611a8d12005-10-24 16:20:10 +0000685monotonically increasing integer that identifies the version of the bytecode
Reid Spencer9bd2be22004-07-29 00:13:04 +0000686format (which is not directly related to the LLVM release number). The
687bytecode versions defined so far are (note that this document only
Owen Andersond753fbc2007-01-27 19:23:25 +0000688describes the latest version, 2.0):</p>
Chris Lattner2b905652004-05-24 05:35:17 +0000689<ul>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000690 <li>#0: LLVM 1.0 &amp; 1.1</li>
691 <li>#1: LLVM 1.2</li>
692 <li>#2: LLVM 1.2.5 (not released)</li>
John Criswell1d9be9e2005-10-24 17:10:57 +0000693 <li>#3: LLVM 1.3</li>
694 <li>#4: LLVM 1.3.x (not released)</li>
Owen Andersond753fbc2007-01-27 19:23:25 +0000695 <li>#5: LLVM 1.4 through 1.8</li>
696 <li>#6: LLVM 1.9</li>
697 <li>#7: LLVM 2.0 and newer</li>
Chris Lattner2b905652004-05-24 05:35:17 +0000698</ul>
Reid Spencer50026612004-05-22 02:28:36 +0000699</div>
700<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000701<div class="doc_subsection"><a name="globaltypes">Global Type Pool</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +0000702<div class="doc_text">
Chris Lattner2b905652004-05-24 05:35:17 +0000703<p>The global type pool consists of type definitions. Their order of appearance
Reid Spencer09daa632004-08-18 20:06:19 +0000704in the file determines their type slot number (0 based). Slot numbers are
Reid Spencer9bd2be22004-07-29 00:13:04 +0000705used to replace pointers in the intermediate representation. Each slot number
706uniquely identifies one entry in a type plane (a collection of values of the
707same type). Since all values have types and are associated with the order in
708which the type pool is written, the global type pool <em>must</em> be written
709as the first block of a module. If it is not, attempts to read the file will
710fail because both forward and backward type resolution will not be possible.</p>
711<p>The type pool is simply a list of type definitions, as shown in the
712table below.</p>
Reid Spencer2cc36152004-07-05 19:04:27 +0000713<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000714 <tbody>
715 <tr>
716 <th><b>Type</b></th>
717 <th class="td_left"><b>Field Description</b></th>
718 </tr>
719 <tr>
720 <td><a href="#unsigned">block</a></td>
721 <td class="td_left">Type Pool Identifier (0x06) + Size<br>
722 </td>
723 </tr>
724 <tr>
725 <td><a href="#llist">llist</a>(<a href="#type">type</a>)</td>
726 <td class="td_left">A length list of type definitions.</td>
727 </tr>
728 </tbody>
Reid Spencerb39021b2004-05-23 17:05:09 +0000729</table>
Reid Spencer50026612004-05-22 02:28:36 +0000730</div>
731<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +0000732<div class="doc_subsubsection"><a name="type">Type Definitions</a></div>
733<div class="doc_text">
Reid Spencer82c46712004-07-07 13:34:26 +0000734<p>Types in the type pool are defined using a different format for each kind
735of type, as given in the following sections.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000736<h3>Primitive Types</h3>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000737<p>The primitive types encompass the basic integer and floating point
Reid Spencer09daa632004-08-18 20:06:19 +0000738types. They are encoded simply as their TypeID.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000739<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000740 <tbody>
741 <tr>
742 <th><b>Type</b></th>
743 <th class="td_left"><b>Description</b></th>
744 </tr>
745 <tr>
746 <td><a href="#uint24_vbr">uint24_vbr</a></td>
747 <td class="td_left">Type ID for the primitive types (values 1 to
74811) <sup>1</sup></td>
749 </tr>
750 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000751</table>
Reid Spencer2cc36152004-07-05 19:04:27 +0000752Notes:
753<ol>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000754 <li>The values for the Type IDs for the primitive types are provided
755by the definition of the <code>llvm::Type::TypeID</code> enumeration
756in <code>include/llvm/Type.h</code>. The enumeration gives the
757following mapping:
758 <ol>
759 <li>bool</li>
760 <li>ubyte</li>
761 <li>sbyte</li>
762 <li>ushort</li>
763 <li>short</li>
764 <li>uint</li>
765 <li>int</li>
766 <li>ulong</li>
767 <li>long</li>
768 <li>float</li>
769 <li>double</li>
770 </ol>
771 </li>
Reid Spencer2cc36152004-07-05 19:04:27 +0000772</ol>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000773<h3>Function Types</h3>
774<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000775 <tbody>
776 <tr>
777 <th><b>Type</b></th>
778 <th class="td_left"><b>Description</b></th>
779 </tr>
780 <tr>
781 <td><a href="#uint24_vbr">uint24_vbr</a></td>
782 <td class="td_left">Type ID for function types (13)</td>
783 </tr>
784 <tr>
785 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000786 <td class="td_left">Type slot number of function's return type.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000787 </tr>
788 <tr>
789 <td><a href="#llist">llist</a>(<a href="#uint24_vbr">uint24_vbr</a>)</td>
Reid Spencer09daa632004-08-18 20:06:19 +0000790 <td class="td_left">Type slot number of each argument's type.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000791 </tr>
792 <tr>
793 <td><a href="#uint32_vbr">uint32_vbr</a>?</td>
794 <td class="td_left">Value 0 if this is a varargs function,
795missing otherwise.</td>
796 </tr>
797 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000798</table>
799<h3>Structure Types</h3>
800<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000801 <tbody>
802 <tr>
803 <th><b>Type</b></th>
804 <th class="td_left"><b>Description</b></th>
805 </tr>
806 <tr>
807 <td><a href="#uint24_vbr">uint24_vbr</a></td>
808 <td class="td_left">Type ID for structure types (14)</td>
809 </tr>
810 <tr>
811 <td><a href="#zlist">zlist</a>(<a href="#uint24_vbr">uint24_vbr</a>)</td>
812 <td class="td_left">Slot number of each of the element's fields.</td>
813 </tr>
814 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000815</table>
816<h3>Array Types</h3>
817<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000818 <tbody>
819 <tr>
820 <th><b>Type</b></th>
821 <th class="td_left"><b>Description</b></th>
822 </tr>
823 <tr>
824 <td><a href="#uint24_vbr">uint24_vbr</a></td>
825 <td class="td_left">Type ID for Array Types (15)</td>
826 </tr>
827 <tr>
828 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000829 <td class="td_left">Type slot number of array's element type.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000830 </tr>
831 <tr>
832 <td><a href="#uint32_vbr">uint32_vbr</a></td>
833 <td class="td_left">The number of elements in the array.</td>
834 </tr>
835 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000836</table>
837<h3>Pointer Types</h3>
838<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000839 <tbody>
840 <tr>
841 <th><b>Type</b></th>
842 <th class="td_left"><b>Description</b></th>
843 </tr>
844 <tr>
845 <td><a href="#uint24_vbr">uint24_vbr</a></td>
846 <td class="td_left">Type ID For Pointer Types (16)</td>
847 </tr>
848 <tr>
849 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +0000850 <td class="td_left">Type slot number of pointer's element type.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000851 </tr>
852 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000853</table>
854<h3>Opaque Types</h3>
855<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000856 <tbody>
857 <tr>
858 <th><b>Type</b></th>
859 <th class="td_left"><b>Description</b></th>
860 </tr>
861 <tr>
862 <td><a href="#uint24_vbr">uint24_vbr</a></td>
863 <td class="td_left">Type ID For Opaque Types (17)</td>
864 </tr>
865 </tbody>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000866</table>
Brian Gaeke715c90b2004-08-20 06:00:58 +0000867<h3>Packed Types</h3>
868<table>
869 <tbody>
870 <tr>
871 <th><b>Type</b></th>
872 <th class="td_left"><b>Description</b></th>
873 </tr>
874 <tr>
875 <td><a href="#uint24_vbr">uint24_vbr</a></td>
876 <td class="td_left">Type ID for Packed Types (18)</td>
877 </tr>
878 <tr>
879 <td><a href="#uint24_vbr">uint24_vbr</a></td>
880 <td class="td_left">Slot number of packed vector's element type.</td>
881 </tr>
882 <tr>
883 <td><a href="#uint32_vbr">uint32_vbr</a></td>
884 <td class="td_left">The number of elements in the packed vector.</td>
885 </tr>
886 </tbody>
887</table>
Andrew Lenharth88942552006-12-08 17:50:05 +0000888<h3>Packed Structure Types</h3>
889<table>
890 <tbody>
891 <tr>
892 <th><b>Type</b></th>
893 <th class="td_left"><b>Description</b></th>
894 </tr>
895 <tr>
896 <td><a href="#uint24_vbr">uint24_vbr</a></td>
897 <td class="td_left">Type ID for packed structure types (19)</td>
898 </tr>
899 <tr>
900 <td><a href="#zlist">zlist</a>(<a href="#uint24_vbr">uint24_vbr</a>)</td>
901 <td class="td_left">Slot number of each of the element's fields.</td>
902 </tr>
903 </tbody>
904</table>
Reid Spencer1ab929c2004-07-05 08:18:07 +0000905</div>
906<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +0000907<div class="doc_subsection"><a name="globalinfo">Module Global Info</a>
908</div>
Reid Spencer50026612004-05-22 02:28:36 +0000909<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +0000910<p>The module global info block contains the definitions of all global
911variables including their initializers and the <em>declaration</em> of
912all functions. The format is shown in the table below:</p>
913<table>
914 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +0000915 <tr>
916 <th><b>Type</b></th>
917 <th class="td_left"><b>Field Description</b></th>
Reid Spencer2cc36152004-07-05 19:04:27 +0000918 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000919 <tr>
920 <td><a href="#block">block</a></td>
Chris Lattner45f50512006-01-23 23:38:09 +0000921 <td class="td_left">Module global info identifier (0x05) + size</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000922 </tr>
923 <tr>
924 <td><a href="#zlist">zlist</a>(<a href="#globalvar">globalvar</a>)</td>
925 <td class="td_left">A zero terminated list of global var
Reid Spencer09daa632004-08-18 20:06:19 +0000926definitions occurring in the module.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000927 </tr>
928 <tr>
Chris Lattnera1dad812004-10-16 18:03:55 +0000929 <td><a href="#zlist">zlist</a>(<a href="#funcfield">funcfield</a>)</td>
930 <td class="td_left">A zero terminated list of function definitions
Reid Spencer09daa632004-08-18 20:06:19 +0000931occurring in the module.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000932 </tr>
933 <tr>
Chris Lattner45f50512006-01-23 23:38:09 +0000934 <td><a href="#llist">llist</a>(<a href="#string">string</a>)</td>
Chris Lattner767d6ba2005-11-12 01:46:21 +0000935 <td class="td_left">A length list
Reid Spencer9bd2be22004-07-29 00:13:04 +0000936of strings that specify the names of the libraries that this module
Chris Lattner45f50512006-01-23 23:38:09 +0000937depends upon.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000938 </tr>
939 <tr>
Chris Lattner45f50512006-01-23 23:38:09 +0000940 <td><a href="#string">string</a></td>
Chris Lattner767d6ba2005-11-12 01:46:21 +0000941 <td class="td_left">The target
Reid Spencer9bd2be22004-07-29 00:13:04 +0000942triple for the module (blank means no target triple specified, i.e. a
Owen Andersond753fbc2007-01-27 19:23:25 +0000943platform-independent module).</td>
944 </tr>
945 <tr>
946 <td><a href="#string">string</a></td>
947 <td class="td_left">The data layout string describing the endianness, pointer size, and
948type alignments for which the module was written (blank means no data layout specified, i.e. a platform-independent module).</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000949 </tr>
Chris Lattner767d6ba2005-11-12 01:46:21 +0000950 <tr>
Chris Lattner45f50512006-01-23 23:38:09 +0000951 <td><a href="#llist">llist</a>(<a href="#string">string</a>)</td>
Chris Lattner767d6ba2005-11-12 01:46:21 +0000952 <td class="td_left">A length list
953of strings that defines a table of section strings for globals. A global's
Chris Lattner45f50512006-01-23 23:38:09 +0000954SectionID is an index into this table.</td>
955 </tr>
956 <tr>
957 <td><a href="#string">string</a></td>
958 <td class="td_left">The inline asm block for this module.</td>
Chris Lattner767d6ba2005-11-12 01:46:21 +0000959 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000960 </tbody>
961</table>
Reid Spencer50026612004-05-22 02:28:36 +0000962</div>
Chris Lattnera1dad812004-10-16 18:03:55 +0000963
Reid Spencer2cc36152004-07-05 19:04:27 +0000964<!-- _______________________________________________________________________ -->
965<div class="doc_subsubsection"><a name="globalvar">Global Variable Field</a>
966</div>
Chris Lattnerafedcb32005-11-06 07:20:25 +0000967
Reid Spencer2cc36152004-07-05 19:04:27 +0000968<div class="doc_text">
Chris Lattnerafedcb32005-11-06 07:20:25 +0000969
Reid Spencer9bd2be22004-07-29 00:13:04 +0000970<p>Global variables are written using an <a href="#uint32_vbr">uint32_vbr</a>
Chris Lattnerafedcb32005-11-06 07:20:25 +0000971that encodes information about the global variable, an optional extension vbr,
972and a an optional initializers for the global var.</p>
973
Reid Spencer9bd2be22004-07-29 00:13:04 +0000974<p>The table below provides the bit layout of the first <a
975 href="#uint32_vbr">uint32_vbr</a> that describes the global variable.</p>
Chris Lattnerafedcb32005-11-06 07:20:25 +0000976
Reid Spencer9bd2be22004-07-29 00:13:04 +0000977<table>
978 <tbody>
Reid Spencer82c46712004-07-07 13:34:26 +0000979 <tr>
980 <th><b>Type</b></th>
981 <th class="td_left"><b>Description</b></th>
Reid Spencer82c46712004-07-07 13:34:26 +0000982 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000983 <tr>
984 <td><a href="#bit">bit(0)</a></td>
985 <td class="td_left">Is constant?</td>
986 </tr>
987 <tr>
988 <td><a href="#bit">bit(1)</a></td>
989 <td class="td_left">Has initializer? Note that this bit
990determines whether the constant initializer field (described below)
991follows. </td>
992 </tr>
993 <tr>
994 <td><a href="#bit">bit(2-4)</a></td>
995 <td class="td_left">Linkage type: 0=External, 1=Weak,
Anton Korobeynikovb74ed072006-09-14 18:23:27 +0000996 2=Appending, 3=Internal, 4=LinkOnce, 5=DllImport,
997 6=DllExport, 7=ExternWeak</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +0000998 </tr>
999 <tr>
1000 <td><a href="#bit">bit(5-31)</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +00001001 <td class="td_left">Type slot number of type for the global variable.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001002 </tr>
1003 </tbody>
1004</table>
Chris Lattnerafedcb32005-11-06 07:20:25 +00001005
1006<p>When the Linkage type is set to 3 (internal) and the initializer field is set
1007to 0 (an invalid combination), an extension word follows the first <a
1008href="#uint32_vbr">uint32_vbr</a> which encodes the real linkage and init flag,
1009and can includes more information:</p>
1010
1011<table>
1012 <tbody>
1013 <tr>
1014 <th><b>Type</b></th>
1015 <th class="td_left"><b>Description</b></th>
1016 </tr>
1017 <tr>
1018 <td><a href="#bit">bit(0)</a></td>
1019 <td class="td_left">Has initializer? Indicates the real value of the "Has
1020 initializer" field for the global. </td>
1021 </tr>
1022 <tr>
1023 <td><a href="#bit">bit(2-4)</a></td>
1024 <td class="td_left">Linkage type: Indicates the real value of the "linkage
1025 type" field for the global.</td>
1026 </tr>
1027 <tr>
1028 <td><a href="#bit">bit(4-8)</a></td>
1029 <td class="td_left">The log-base-2 of the alignment for the global.</td>
1030 </tr>
1031 <tr>
Chris Lattner767d6ba2005-11-12 01:46:21 +00001032 <td><a href="#bit">bit(9)</a></td>
1033 <td class="td_left">If this bit is set, a SectionID follows this vbr.</td>
1034 </tr>
1035 <tr>
Anton Korobeynikov7f705592007-01-12 19:20:47 +00001036 <td><a href="#bit">bit(10-12)</a></td>
1037 <td class="td_left">Visibility style: 0=Default, 1=Hidden.</td>
1038 </tr>
1039 <tr>
1040 <td><a href="#bit">bit(13-31)</a></td>
Chris Lattnerafedcb32005-11-06 07:20:25 +00001041 <td class="td_left">Currently unassigned.</td>
1042 </tr>
1043 </tbody>
1044</table>
1045
Chris Lattner767d6ba2005-11-12 01:46:21 +00001046<p>If the SectionID bit is set above, the following field is included:</p>
Chris Lattnerafedcb32005-11-06 07:20:25 +00001047
Reid Spencer9bd2be22004-07-29 00:13:04 +00001048<table>
1049 <tbody>
1050 <tr>
1051 <th><b>Type</b></th>
1052 <th class="td_left"><b>Description</b></th>
1053 </tr>
1054 <tr>
Chris Lattner767d6ba2005-11-12 01:46:21 +00001055 <td><a href="#uint32_vbr">uint32_vbr</a>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001056 </td>
Chris Lattner767d6ba2005-11-12 01:46:21 +00001057 <td class="td_left">An optional section ID number, specifying the string
1058 to use for the section of the global. This an index (+1) of an entry
1059 into the SectionID llist in the <a href="#globalinfo">Module Global
1060 Info</a> block. If this value is 0 or not present, the global has an
1061 empty section string.</td>
1062 </tr>
1063 </tbody>
1064</table>
1065
1066<p>If the "Has initializer" field is set, the following field is included:</p>
1067
1068<table>
1069 <tbody>
1070 <tr>
1071 <th><b>Type</b></th>
1072 <th class="td_left"><b>Description</b></th>
1073 </tr>
1074 <tr>
1075 <td><a href="#uint32_vbr">uint32_vbr</a>
1076 </td>
1077 <td class="td_left">An optional value slot number for the global
1078 variable's constant initializer.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001079 </tr>
1080 </tbody>
1081</table>
Reid Spencer2cc36152004-07-05 19:04:27 +00001082</div>
Chris Lattnera1dad812004-10-16 18:03:55 +00001083
1084<!-- _______________________________________________________________________ -->
1085<div class="doc_subsubsection"><a name="funcfield">Function Field</a>
1086</div>
1087<div class="doc_text">
1088<p>Functions are written using an <a href="#uint32_vbr">uint32_vbr</a>
Chris Lattnerafedcb32005-11-06 07:20:25 +00001089that encodes information about the function and a set of flags. If needed,
1090an extension word may follow this first field.</p>
Chris Lattnera1dad812004-10-16 18:03:55 +00001091
1092<p>The table below provides the bit layout of the <a
1093href="#uint32_vbr">uint32_vbr</a> that describes the function.</p>
1094
1095<table>
1096 <tbody>
1097 <tr>
1098 <th><b>Type</b></th>
1099 <th class="td_left"><b>Description</b></th>
1100 </tr>
1101 <tr>
Chris Lattnerced62222004-11-15 22:54:50 +00001102 <td><a href="#bit">bit(0-3)</a></td>
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001103 <td class="td_left">
Chris Lattner52942b52005-11-06 07:48:11 +00001104 Encodes the calling convention number of the function. The
Chris Lattnerbad10ee2005-05-06 22:57:40 +00001105 CC number of the function is the value of this field minus one.
1106 </td>
Chris Lattnerced62222004-11-15 22:54:50 +00001107 </tr>
1108 <tr>
1109 <td><a href="#bit">bit(4)</a></td>
1110 <td class="td_left">If this bit is set to 1, the indicated function is
1111 external, and there is no <a href="#functiondefs">Function Definiton
Anton Korobeynikovb74ed072006-09-14 18:23:27 +00001112 Block</a> in the bytecode file for the function. If the function is
1113 external and has <tt>dllimport or extern_weak</tt> linkage additional
1114 field in the extension word is used to indicate the actual linkage
1115 type.</td>
Chris Lattnera1dad812004-10-16 18:03:55 +00001116 </tr>
1117 <tr>
Chris Lattnerafedcb32005-11-06 07:20:25 +00001118 <td><a href="#bit">bit(5-30)</a></td>
Chris Lattnera1dad812004-10-16 18:03:55 +00001119 <td class="td_left">Type slot number of type for the function.</td>
1120 </tr>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +00001121 <tr>
Chris Lattnerafedcb32005-11-06 07:20:25 +00001122 <td><a href="#bit">bit(31)</a></td>
1123 <td class="td_left">Indicates whether an extension word follows.</td>
1124 </tr>
1125 </tbody>
1126</table>
1127
1128<p>If bit(31) is set, an additional <a href="#uint32_vbr">uint32_vbr</a> word
1129follows with the following fields:</p>
1130
1131<table>
1132 <tbody>
1133 <tr>
1134 <th><b>Type</b></th>
1135 <th class="td_left"><b>Description</b></th>
1136 </tr>
1137 <tr>
1138 <td><a href="#bit">bit(0-4)</a></td>
1139 <td class="td_left">The log-base-2 of the alignment for the function.</td>
1140 </tr>
1141 <tr>
Chris Lattner52942b52005-11-06 07:48:11 +00001142 <td><a href="#bit">bit(5-9)</a></td>
1143 <td class="td_left">The top nibble of the calling convention.</td>
1144 </tr>
1145 <tr>
Chris Lattner767d6ba2005-11-12 01:46:21 +00001146 <td><a href="#bit">bit(10)</a></td>
1147 <td class="td_left">If this bit is set, a SectionID follows this vbr.</td>
1148 </tr>
1149 <tr>
Anton Korobeynikovb74ed072006-09-14 18:23:27 +00001150 <td><a href="#bit">bit(11-12)</a></td>
1151 <td class="td_left">Linkage type for external functions. 0 - External
1152 linkage, 1 - DLLImport linkage, 2 - External weak linkage.</td>
1153 </tr>
1154 <tr>
1155 <td><a href="#bit">bit(13-31)</a></td>
Chris Lattnerafedcb32005-11-06 07:20:25 +00001156 <td class="td_left">Currently unassigned.</td>
1157 </tr>
Chris Lattnera1dad812004-10-16 18:03:55 +00001158 </tbody>
1159</table>
1160
Chris Lattner767d6ba2005-11-12 01:46:21 +00001161<p>If the SectionID bit is set above, the following field is included:</p>
1162
1163<table>
1164 <tbody>
1165 <tr>
1166 <th><b>Type</b></th>
1167 <th class="td_left"><b>Description</b></th>
1168 </tr>
1169 <tr>
1170 <td><a href="#uint32_vbr">uint32_vbr</a>
1171 </td>
1172 <td class="td_left">An optional section ID number, specifying the string
1173 to use for the section of the function. This an index (+1) of an entry
1174 into the SectionID llist in the <a href="#globalinfo">Module Global
1175 Info</a> block. If this value is 0 or not present, the function has an
1176 empty section string.</td>
1177 </tr>
1178 </tbody>
1179</table>
1180
Chris Lattnera1dad812004-10-16 18:03:55 +00001181</div>
1182
Reid Spencer50026612004-05-22 02:28:36 +00001183<!-- _______________________________________________________________________ -->
Reid Spencer1ab929c2004-07-05 08:18:07 +00001184<div class="doc_subsection"><a name="constantpool">Constant Pool</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +00001185<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +00001186<p>A constant pool defines as set of constant values. There are
1187actually two types of constant pool blocks: one for modules and one for
1188functions. For modules, the block begins with the constant strings
1189encountered anywhere in the module. For functions, the block begins
1190with types only encountered in the function. In both cases the header
1191is identical. The tables that follow, show the header, module constant
1192pool preamble, function constant pool preamble, and the part common to
1193both function and module constant pools.</p>
1194<p><b>Common Block Header</b></p>
1195<table>
1196 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +00001197 <tr>
1198 <th><b>Type</b></th>
1199 <th class="td_left"><b>Field Description</b></th>
Reid Spencer2cc36152004-07-05 19:04:27 +00001200 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001201 <tr>
1202 <td><a href="#block">block</a></td>
1203 <td class="td_left">Constant pool identifier (0x03) + size<br>
1204 </td>
1205 </tr>
1206 </tbody>
1207</table>
1208<p><b>Module Constant Pool Preamble (constant strings)</b></p>
1209<table>
1210 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +00001211 <tr>
1212 <th><b>Type</b></th>
1213 <th class="td_left"><b>Field Description</b></th>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001214 </tr>
1215 <tr>
Reid Spencer2cc36152004-07-05 19:04:27 +00001216 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1217 <td class="td_left">The number of constant strings that follow.</td>
Reid Spencer2cc36152004-07-05 19:04:27 +00001218 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001219 <tr>
1220 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1221 <td class="td_left">Zero. This identifies the following "plane"
1222as containing the constant strings. This is needed to identify it
1223uniquely from other constant planes that follow. </td>
1224 </tr>
1225 <tr>
1226 <td><a href="#uint24_vbr">uint24_vbr</a>+</td>
Reid Spencer09daa632004-08-18 20:06:19 +00001227 <td class="td_left">Type slot number of the constant string's type.
Reid Spencer9bd2be22004-07-29 00:13:04 +00001228Note that the constant string's type implicitly defines the length of
1229the string. </td>
1230 </tr>
1231 </tbody>
1232</table>
1233<p><b>Function Constant Pool Preamble (function types)</b></p>
1234<p>The structure of the types for functions is identical to the <a
1235 href="#globaltypes">Global Type Pool</a>. Please refer to that section
1236for the details. </p>
1237<p><b>Common Part (other constants)</b></p>
1238<table>
1239 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +00001240 <tr>
1241 <th><b>Type</b></th>
1242 <th class="td_left"><b>Field Description</b></th>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001243 </tr>
1244 <tr>
Reid Spencer2cc36152004-07-05 19:04:27 +00001245 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1246 <td class="td_left">Number of entries in this type plane.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001247 </tr>
1248 <tr>
1249 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer2cc36152004-07-05 19:04:27 +00001250 <td class="td_left">Type slot number of this plane.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001251 </tr>
1252 <tr>
Reid Spencer82c46712004-07-07 13:34:26 +00001253 <td><a href="#constant">constant</a>+</td>
Reid Spencer2cc36152004-07-05 19:04:27 +00001254 <td class="td_left">The definition of a constant (see below).</td>
1255 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001256 </tbody>
1257</table>
Reid Spencer2cc36152004-07-05 19:04:27 +00001258</div>
Chris Lattner591d0492006-01-25 23:31:53 +00001259
Reid Spencer2cc36152004-07-05 19:04:27 +00001260<!-- _______________________________________________________________________ -->
Chris Lattner591d0492006-01-25 23:31:53 +00001261<div class="doc_subsubsection"><a name="constant">Simple Constant Pool
1262Entries</a></div>
1263
Reid Spencer2cc36152004-07-05 19:04:27 +00001264<div class="doc_text">
Chris Lattner591d0492006-01-25 23:31:53 +00001265
1266<p>Constant pool entries come in many shapes and flavors. The sections that
1267follow define the format for each of them. All constants start with a <a
Reid Spencer9bd2be22004-07-29 00:13:04 +00001268 href="#uint32_vbr">uint32_vbr</a> encoded integer that provides the
1269number of operands for the constant. For primitive, structure, and
Chris Lattner591d0492006-01-25 23:31:53 +00001270array constants, this will always be zero to indicate that the form of the
1271constant is solely determined by its type. In this case, we have the following
1272field definitions, based on type:</p>
1273
Reid Spencer9bd2be22004-07-29 00:13:04 +00001274<ul>
1275 <li><b>Bool</b>. This is written as an <a href="#uint32_vbr">uint32_vbr</a>
1276of value 1U or 0U.</li>
1277 <li><b>Signed Integers (sbyte,short,int,long)</b>. These are written
1278as an <a href="#int64_vbr">int64_vbr</a> with the corresponding value.</li>
1279 <li><b>Unsigned Integers (ubyte,ushort,uint,ulong)</b>. These are
1280written as an <a href="#uint64_vbr">uint64_vbr</a> with the
1281corresponding value. </li>
1282 <li><b>Floating Point</b>. Both the float and double types are
1283written literally in binary format.</li>
1284 <li><b>Arrays</b>. Arrays are written simply as a list of <a
Reid Spencer09daa632004-08-18 20:06:19 +00001285 href="#uint32_vbr">uint32_vbr</a> encoded value slot numbers to the constant
Reid Spencer9bd2be22004-07-29 00:13:04 +00001286element values.</li>
1287 <li><b>Structures</b>. Structures are written simply as a list of <a
Reid Spencer09daa632004-08-18 20:06:19 +00001288 href="#uint32_vbr">uint32_vbr</a> encoded value slot numbers to the constant
Reid Spencer9bd2be22004-07-29 00:13:04 +00001289field values of the structure.</li>
1290</ul>
Chris Lattnera1dad812004-10-16 18:03:55 +00001291
Chris Lattner591d0492006-01-25 23:31:53 +00001292</div>
1293
1294<!-- _______________________________________________________________________ -->
Anton Korobeynikov7f705592007-01-12 19:20:47 +00001295<div class="doc_subsubsection">Undef Entries</div>
Chris Lattner591d0492006-01-25 23:31:53 +00001296
1297<div class="doc_text">
Chris Lattnera1dad812004-10-16 18:03:55 +00001298<p>When the number of operands to the constant is one, we have an 'undef' value
1299of the specified type.</p>
Chris Lattner591d0492006-01-25 23:31:53 +00001300</div>
Chris Lattnera1dad812004-10-16 18:03:55 +00001301
Chris Lattner591d0492006-01-25 23:31:53 +00001302<!-- _______________________________________________________________________ -->
Anton Korobeynikov7f705592007-01-12 19:20:47 +00001303<div class="doc_subsubsection">Inline Assembler Entries</div>
Chris Lattner591d0492006-01-25 23:31:53 +00001304
1305<div class="doc_text">
1306<p>Inline Assembler entries are stored in the constant pool, though they are not
1307 officially LLVM constants. These entries are marked with a value of
1308 "4294967295" (all ones) for the number of operands. They are encoded as
1309 follows:</p>
1310
1311<table>
1312 <tbody>
1313 <tr>
1314 <th><b>Type</b></th>
1315 <th class="td_left"><b>Field Description</b></th>
1316 </tr>
1317 <tr>
1318 <td><a href="#string">string</a></td>
1319 <td class="td_left">The asm string.</td>
1320 </tr>
1321 <tr>
1322 <td><a href="#string">string</a></td>
1323 <td class="td_left">The constraints string.</td>
1324 </tr>
1325 <tr>
1326 <td><a href="#uint32_vbr">uint32_vbr</a></td>
Anton Korobeynikov7f705592007-01-12 19:20:47 +00001327 <td class="td_left">Flags</td>
Chris Lattner591d0492006-01-25 23:31:53 +00001328 </tr>
1329 </tbody>
1330</table>
1331
1332<p>Currently, the only defined flag, the low bit, indicates whether or not the
1333 inline assembler has side effects.</p>
1334
1335</div>
1336
1337<!-- _______________________________________________________________________ -->
Anton Korobeynikov7f705592007-01-12 19:20:47 +00001338<div class="doc_subsubsection">Constant Expression Entries</div>
Chris Lattner591d0492006-01-25 23:31:53 +00001339
1340<div class="doc_text">
1341
1342<p>Otherwise, we have a constant expression. The format of the constant
1343expression is specified in the table below, and the number is equal to the
1344number of operands+1.</p>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001345<table>
1346 <tbody>
Reid Spencer2cc36152004-07-05 19:04:27 +00001347 <tr>
1348 <th><b>Type</b></th>
1349 <th class="td_left"><b>Field Description</b></th>
Reid Spencer2cc36152004-07-05 19:04:27 +00001350 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001351 <tr>
1352 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1353 <td class="td_left">Op code of the instruction for the constant
1354expression.</td>
1355 </tr>
1356 <tr>
1357 <td><a href="#uint32_vbr">uint32_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +00001358 <td class="td_left">The value slot number of the constant value for an
Reid Spencer9bd2be22004-07-29 00:13:04 +00001359operand.<sup>1</sup></td>
1360 </tr>
1361 <tr>
1362 <td><a href="#uint24_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +00001363 <td class="td_left">The type slot number for the type of the constant
Reid Spencer9bd2be22004-07-29 00:13:04 +00001364value for an operand.<sup>1</sup></td>
1365 </tr>
1366 </tbody>
1367</table>
1368Notes:
1369<ol>
1370 <li>Both these fields are repeatable but only in pairs.</li>
1371</ol>
Reid Spencer50026612004-05-22 02:28:36 +00001372</div>
1373<!-- _______________________________________________________________________ -->
Reid Spencer51f31e02004-07-05 22:28:02 +00001374<div class="doc_subsection"><a name="functiondefs">Function Definition</a></div>
Reid Spencer1ab929c2004-07-05 08:18:07 +00001375<div class="doc_text">
Reid Spencerc4a8d392007-01-31 00:26:08 +00001376<p>Function definitions contain the linkage, constant pool, instruction list,
1377and symbol table for a function. The following table shows the structure of
1378a function definition.</p>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001379<table>
1380 <tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +00001381 <tr>
1382 <th><b>Type</b></th>
1383 <th class="td_left"><b>Field Description</b></th>
Reid Spencer51f31e02004-07-05 22:28:02 +00001384 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001385 <tr>
1386 <td><a href="#block">block</a><br>
1387 </td>
1388 <td class="td_left">Function definition block identifier (0x02) +
1389size<br>
1390 </td>
1391 </tr>
1392 <tr>
1393 <td><a href="#uint32_vbr">uint32_vbr</a></td>
Anton Korobeynikov7f705592007-01-12 19:20:47 +00001394 <td class="td_left"><a href="#funclinkage_and_visibility">The linkage and
1395 visibility</a> style field</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001396 </tr>
1397 <tr>
1398 <td><a href="#block">block</a></td>
1399 <td class="td_left">The <a href="#constantpool">constant pool</a>
1400block for this function.<sup>2</sup></td>
1401 </tr>
1402 <tr>
1403 <td><a href="#block">block</a></td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001404 <td class="td_left">The <a href="#instructionlist">instruction
1405list</a> for the function.</td>
1406 </tr>
1407 <tr>
1408 <td><a href="#block">block</a></td>
Reid Spencer8996e552004-08-17 00:49:03 +00001409 <td class="td_left">The function's <a href="#symtab">symbol
Reid Spencer9bd2be22004-07-29 00:13:04 +00001410table</a> containing only those symbols pertinent to the function
1411(mostly block labels).</td>
1412 </tr>
1413 </tbody>
1414</table>
1415Notes:
1416<ol>
1417 <li>Note that if the linkage type is "External" then none of the
1418other fields will be present as the function is defined elsewhere.</li>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001419</ol>
Anton Korobeynikov7f705592007-01-12 19:20:47 +00001420
1421<!-- _______________________________________________________________________ -->
1422<div class="doc_subsubsection"><a name="funclinkage_and_visibility">Linkage and
1423 visibility word</a>
Reid Spencer1ab929c2004-07-05 08:18:07 +00001424</div>
Anton Korobeynikov7f705592007-01-12 19:20:47 +00001425<div class="doc_text">
1426
1427<table>
1428 <tbody>
1429 <tr>
1430 <th><b>Type</b></th>
1431 <th class="td_left"><b>Field Description</b></th>
1432 </tr>
1433 <tr>
1434 <td><a href="#bit">bit(0-15)</a></td>
1435 <td class="td_left">The linkage type of the function: 0=External, 1=Weak,
14362=Appending, 3=Internal, 4=LinkOnce, 5=DllImport, 6=DllExport<sup>1</sup></td>
1437 </tr>
1438 <tr>
Anton Korobeynikovdaa31892007-01-23 12:43:53 +00001439 <td><a href="#bit">bit(16-18)</a></td>
Anton Korobeynikov7f705592007-01-12 19:20:47 +00001440 <td class="td_left">Visibility style: 0=Default, 1=Hidden.</td>
1441 </tr>
Anton Korobeynikovdaa31892007-01-23 12:43:53 +00001442 <tr>
1443 <td><a href="#bit">bit(19-31)</a></td>
Reid Spencerc4a8d392007-01-31 00:26:08 +00001444 <td class="td_left">Currently unassigned.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001445 </tr>
1446 </tbody>
1447</table>
Reid Spencer1ab929c2004-07-05 08:18:07 +00001448</div>
Reid Spencerc4a8d392007-01-31 00:26:08 +00001449
Reid Spencer1ab929c2004-07-05 08:18:07 +00001450<!-- _______________________________________________________________________ -->
Reid Spencer09daa632004-08-18 20:06:19 +00001451<div class="doc_subsection"><a name="instructionlist">Instruction List</a></div>
Reid Spencer50026612004-05-22 02:28:36 +00001452<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +00001453<p>The instructions in a function are written as a simple list. Basic
1454blocks are inferred by the terminating instruction types. The format of
1455the block is given in the following table.</p>
1456<table>
1457 <tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +00001458 <tr>
1459 <th><b>Type</b></th>
1460 <th class="td_left"><b>Field Description</b></th>
Reid Spencer51f31e02004-07-05 22:28:02 +00001461 </tr>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001462 <tr>
1463 <td><a href="#block">block</a><br>
1464 </td>
1465 <td class="td_left">Instruction list identifier (0x07) + size<br>
1466 </td>
1467 </tr>
1468 <tr>
1469 <td><a href="#instruction">instruction</a>+</td>
1470 <td class="td_left">An instruction. Instructions have a variety
1471of formats. See <a href="#instruction">Instructions</a> for details.</td>
1472 </tr>
1473 </tbody>
1474</table>
Reid Spencer50026612004-05-22 02:28:36 +00001475</div>
Chris Lattnerf97fa592005-11-05 22:20:06 +00001476
Reid Spencer51f31e02004-07-05 22:28:02 +00001477<!-- _______________________________________________________________________ -->
Chris Lattnerf97fa592005-11-05 22:20:06 +00001478<div class="doc_subsection"><a name="instructions">Instructions</a></div>
1479
Reid Spencer51f31e02004-07-05 22:28:02 +00001480<div class="doc_text">
Chris Lattnerf97fa592005-11-05 22:20:06 +00001481<p>Instructions are written out one at a time as distinct units. Each
1482instruction
1483record contains at least an <a href="#opcodes">opcode</a> and a type field,
Chris Lattner49b16022005-11-05 22:32:06 +00001484and may contain a <a href="#instoperands">list of operands</a> (whose
1485interpretation depends on the opcode). Based on the number of operands, the
Chris Lattnerf97fa592005-11-05 22:20:06 +00001486<a href="#instencode">instruction is encoded</a> in a
1487dense format that tries to encoded each instruction into 32-bits if
1488possible. </p>
Reid Spencer51f31e02004-07-05 22:28:02 +00001489</div>
Reid Spencer09daa632004-08-18 20:06:19 +00001490
1491<!-- _______________________________________________________________________ -->
Chris Lattnerf97fa592005-11-05 22:20:06 +00001492<div class="doc_subsubsection"><a name="opcodes">Instruction Opcodes</a></div>
Reid Spencer09daa632004-08-18 20:06:19 +00001493<div class="doc_text">
1494 <p>Instructions encode an opcode that identifies the kind of instruction.
1495 Opcodes are an enumerated integer value. The specific values used depend on
1496 the version of LLVM you're using. The opcode values are defined in the
Reid Spencer05fe4b02006-03-14 05:39:39 +00001497 <a href="http://llvm.org/cvsweb/cvsweb.cgi/llvm/include/llvm/Instruction.def">
Reid Spencer09daa632004-08-18 20:06:19 +00001498 <tt>include/llvm/Instruction.def</tt></a> file. You should check there for the
1499 most recent definitions. The table below provides the opcodes defined as of
Nate Begeman52ca9e42004-08-27 07:59:37 +00001500 the writing of this document. The table associates each opcode mnemonic with
Reid Spencer09daa632004-08-18 20:06:19 +00001501 its enumeration value and the bytecode and LLVM version numbers in which the
1502 opcode was introduced.</p>
1503 <table>
1504 <tbody>
1505 <tr>
1506 <th>Opcode</th>
1507 <th>Number</th>
1508 <th>Bytecode Version</th>
1509 <th>LLVM Version</th>
1510 </tr>
1511 <tr><td colspan="4"><b>Terminator Instructions</b></td></tr>
1512 <tr><td>Ret</td><td>1</td><td>1</td><td>1.0</td></tr>
1513 <tr><td>Br</td><td>2</td><td>1</td><td>1.0</td></tr>
1514 <tr><td>Switch</td><td>3</td><td>1</td><td>1.0</td></tr>
1515 <tr><td>Invoke</td><td>4</td><td>1</td><td>1.0</td></tr>
1516 <tr><td>Unwind</td><td>5</td><td>1</td><td>1.0</td></tr>
Chris Lattnera1dad812004-10-16 18:03:55 +00001517 <tr><td>Unreachable</td><td>6</td><td>1</td><td>1.4</td></tr>
Reid Spencer09daa632004-08-18 20:06:19 +00001518 <tr><td colspan="4"><b>Binary Operators</b></td></tr>
Chris Lattnera1dad812004-10-16 18:03:55 +00001519 <tr><td>Add</td><td>7</td><td>1</td><td>1.0</td></tr>
1520 <tr><td>Sub</td><td>8</td><td>1</td><td>1.0</td></tr>
1521 <tr><td>Mul</td><td>9</td><td>1</td><td>1.0</td></tr>
Reid Spencer3727fd32006-11-08 20:06:36 +00001522 <tr><td>UDiv</td><td>10</td><td>1</td><td>1.9</td></tr>
1523 <tr><td>SDiv</td><td>11</td><td>1</td><td>1.9</td></tr>
1524 <tr><td>FDiv</td><td>12</td><td>1</td><td>1.9</td></tr>
1525 <tr><td>URem</td><td>13</td><td>1</td><td>1.9</td></tr>
1526 <tr><td>SRem</td><td>14</td><td>1</td><td>1.9</td></tr>
1527 <tr><td>FRem</td><td>15</td><td>1</td><td>1.9</td></tr>
Reid Spencer09daa632004-08-18 20:06:19 +00001528 <tr><td colspan="4"><b>Logical Operators</b></td></tr>
Reid Spencer3727fd32006-11-08 20:06:36 +00001529 <tr><td>And</td><td>16</td><td>1</td><td>1.0</td></tr>
1530 <tr><td>Or</td><td>17</td><td>1</td><td>1.0</td></tr>
1531 <tr><td>Xor</td><td>18</td><td>1</td><td>1.0</td></tr>
Reid Spencer09daa632004-08-18 20:06:19 +00001532 <tr><td colspan="4"><b>Binary Comparison Operators</b></td></tr>
Reid Spencer3727fd32006-11-08 20:06:36 +00001533 <tr><td>SetEQ</td><td>19</td><td>1</td><td>1.0</td></tr>
1534 <tr><td>SetNE</td><td>20</td><td>1</td><td>1.0</td></tr>
1535 <tr><td>SetLE</td><td>21</td><td>1</td><td>1.0</td></tr>
1536 <tr><td>SetGE</td><td>22</td><td>1</td><td>1.0</td></tr>
1537 <tr><td>SetLT</td><td>23</td><td>1</td><td>1.0</td></tr>
1538 <tr><td>SetGT</td><td>24</td><td>1</td><td>1.0</td></tr>
Reid Spencer09daa632004-08-18 20:06:19 +00001539 <tr><td colspan="4"><b>Memory Operators</b></td></tr>
Reid Spencer3727fd32006-11-08 20:06:36 +00001540 <tr><td>Malloc</td><td>25</td><td>1</td><td>1.0</td></tr>
1541 <tr><td>Free</td><td>26</td><td>1</td><td>1.0</td></tr>
1542 <tr><td>Alloca</td><td>27</td><td>1</td><td>1.0</td></tr>
1543 <tr><td>Load</td><td>28</td><td>1</td><td>1.0</td></tr>
1544 <tr><td>Store</td><td>29</td><td>1</td><td>1.0</td></tr>
1545 <tr><td>GetElementPtr</td><td>30</td><td>1</td><td>1.0</td></tr>
Reid Spencer09daa632004-08-18 20:06:19 +00001546 <tr><td colspan="4"><b>Other Operators</b></td></tr>
Reid Spencer3727fd32006-11-08 20:06:36 +00001547 <tr><td>PHI</td><td>31</td><td>1</td><td>1.0</td></tr>
1548 <tr><td>Cast</td><td>32</td><td>1</td><td>1.0</td></tr>
1549 <tr><td>Call</td><td>33</td><td>1</td><td>1.0</td></tr>
1550 <tr><td>Shl</td><td>34</td><td>1</td><td>1.0</td></tr>
Reid Spencer3727fd32006-11-08 20:06:36 +00001551 <tr><td>LShr</td><td>35</td><td>6</td><td>2.0</td></tr>
1552 <tr><td>AShr</td><td>36</td><td>6</td><td>2.0</td></tr>
Reid Spencer145b2532006-11-08 20:07:40 +00001553 <tr><td>Select</td><td>37</td><td>2</td><td>1.2</td></tr>
1554 <tr><td>UserOp1</td><td>38</td><td>1</td><td>1.0</td></tr>
1555 <tr><td>UserOp2</td><td>39</td><td>1</td><td>1.0</td></tr>
1556 <tr><td>VAArg</td><td>40</td><td>5</td><td>1.5</td></tr>
1557 <tr><td>ExtractElement</td><td>41</td><td>5</td><td>1.5</td></tr>
1558 <tr><td>InsertElement</td><td>42</td><td>5</td><td>1.5</td></tr>
1559 <tr><td>ShuffleElement</td><td>43</td><td>5</td><td>1.5</td></tr>
Reid Spencera5681772005-05-14 00:06:06 +00001560 <tr><td colspan="4">
1561 <b>Pseudo Instructions<a href="#pi_note">*</a></b>
1562 </td></tr>
1563 <tr><td>Invoke+CC </td><td>56</td><td>5</td><td>1.5</td></tr>
1564 <tr><td>Invoke+FastCC</td><td>57</td><td>5</td><td>1.5</td></tr>
1565 <tr><td>Call+CC</td><td>58</td><td>5</td><td>1.5</td></tr>
1566 <tr><td>Call+FastCC+TailCall</td><td>59</td><td>5</td><td>1.5</td></tr>
1567 <tr><td>Call+FastCC</td><td>60</td><td>5</td><td>1.5</td></tr>
1568 <tr><td>Call+CCC+TailCall</td><td>61</td><td>5</td><td>1.5</td></tr>
1569 <tr><td>Load+Volatile</td><td>62</td><td>3</td><td>1.3</td></tr>
1570 <tr><td>Store+Volatile</td><td>63</td><td>3</td><td>1.3</td></tr>
Reid Spencer09daa632004-08-18 20:06:19 +00001571 </tbody>
1572 </table>
Reid Spencer09daa632004-08-18 20:06:19 +00001573
Reid Spencera5681772005-05-14 00:06:06 +00001574<p><b><a name="pi_note">* Note: </a></b>
Chris Lattnerf97fa592005-11-05 22:20:06 +00001575These aren't really opcodes from an LLVM language perspective. They encode
Reid Spencera5681772005-05-14 00:06:06 +00001576information into other opcodes without reserving space for that information.
1577For example, opcode=63 is a Volatile Store. The opcode for this
1578instruction is 25 (Store) but we encode it as 63 to indicate that is a Volatile
1579Store. The same is done for the calling conventions and tail calls.
1580In each of these entries in range 56-63, the opcode is documented as the base
1581opcode (Invoke, Call, Store) plus some set of modifiers, as follows:</p>
1582<dl>
1583 <dt>CC</dt>
Chris Lattner120bc6d2005-05-14 01:30:15 +00001584 <dd>This means an arbitrary calling convention is specified
1585 in a VBR that follows the opcode. This is used when the instruction cannot
1586 be encoded with one of the more compact forms.
Reid Spencera5681772005-05-14 00:06:06 +00001587 </dd>
1588 <dt>FastCC</dt>
1589 <dd>This indicates that the Call or Invoke is using the FastCC calling
Chris Lattner120bc6d2005-05-14 01:30:15 +00001590 convention.</dd>
Reid Spencera5681772005-05-14 00:06:06 +00001591 <dt>CCC</dt>
Chris Lattner120bc6d2005-05-14 01:30:15 +00001592 <dd>This indicates that the Call or Invoke is using the native "C" calling
1593 convention.</dd>
Reid Spencera5681772005-05-14 00:06:06 +00001594 <dt>TailCall</dt>
Chris Lattner120bc6d2005-05-14 01:30:15 +00001595 <dd>This indicates that the Call has the 'tail' modifier.</dd>
Reid Spencera5681772005-05-14 00:06:06 +00001596</dl>
Chris Lattnerf97fa592005-11-05 22:20:06 +00001597</div>
1598
Chris Lattner49b16022005-11-05 22:32:06 +00001599<!-- _______________________________________________________________________ -->
1600<div class="doc_subsubsection"><a name="instoperands">Instruction
1601Operands</a></div>
1602
1603<div class="doc_text">
1604<p>
1605Based on the instruction opcode and type, the bytecode format implicitly (to
1606save space) specifies the interpretation of the operand list. For most
1607instructions, the type of each operand is implicit from the type of the
1608instruction itself (e.g. the type of operands of a binary operator must match
1609the type of the instruction). As such, the bytecode format generally only
1610encodes the value number of the operand, not the type.</p>
1611
1612<p>In some cases, however, this is not sufficient. This section enumerates
1613those cases:</p>
1614
1615<ul>
1616<li>getelementptr: the slot numbers for sequential type indexes are shifted up
1617two bits. This allows the low order bits will encode the type of index used,
1618as follows: 0=uint, 1=int, 2=ulong, 3=long.</li>
1619<li>cast: the result type number is encoded as the second operand.</li>
1620<li>alloca/malloc: If the allocation has an explicit alignment, the log2 of the
1621 alignment is encoded as the second operand.</li>
1622<li>call: If the tail marker and calling convention cannot be <a
1623 href="#pi_note">encoded into the opcode</a> of the call, it is passed as an
1624 additional operand. The low bit of the operand is a flag indicating whether
1625 the call is a tail call. The rest of the bits contain the calling
1626 convention number (shifted left by one bit).</li>
1627</ul>
1628</div>
Chris Lattnerf97fa592005-11-05 22:20:06 +00001629
1630<!-- _______________________________________________________________________ -->
1631<div class="doc_subsubsection"><a name="instencode">Instruction
1632Encoding</a></div>
1633
1634<div class="doc_text">
1635<p>For brevity, instructions are written in one of four formats,
1636depending on the number of operands to the instruction. Each
1637instruction begins with a <a href="#uint32_vbr">uint32_vbr</a> that
1638encodes the type of the instruction as well as other things. The tables
1639that follow describe the format of this first part of each instruction.</p>
1640<p><b>Instruction Format 0</b></p>
1641<p>This format is used for a few instructions that can't easily be
1642shortened because they have large numbers of operands (e.g. PHI Node or
1643getelementptr). Each of the opcode, type, and operand fields is found in
1644successive fields.</p>
1645<table>
1646 <tbody>
1647 <tr>
1648 <th><b>Type</b></th>
1649 <th class="td_left"><b>Field Description</b></th>
1650 </tr>
1651 <tr>
1652 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1653 <td class="td_left">Specifies the opcode of the instruction. Note
1654that for compatibility with the other instruction formats, the opcode
1655is shifted left by 2 bits. Bits 0 and 1 must have value zero for this
1656format.</td>
1657 </tr>
1658 <tr>
1659 <td><a href="#uint24_vbr">uint24_vbr</a></td>
1660 <td class="td_left">Provides the type slot number of the result type of
1661 the instruction.</td>
1662 </tr>
1663 <tr>
1664 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1665 <td class="td_left">The number of operands that follow.</td>
1666 </tr>
1667 <tr>
1668 <td><a href="#uint32_vbr">uint32_vbr</a>+</td>
1669 <td class="td_left">The slot number of the value(s) for the operand(s).
Chris Lattner49b16022005-11-05 22:32:06 +00001670 </td>
Chris Lattnerf97fa592005-11-05 22:20:06 +00001671 </tr>
1672 </tbody>
1673</table>
Chris Lattner49b16022005-11-05 22:32:06 +00001674
Chris Lattnerf97fa592005-11-05 22:20:06 +00001675<p><b>Instruction Format 1</b></p>
1676<p>This format encodes the opcode, type and a single operand into a
1677single <a href="#uint32_vbr">uint32_vbr</a> as follows:</p>
1678<table>
1679 <tbody>
1680 <tr>
1681 <th><b>Bits</b></th>
1682 <th><b>Type</b></th>
1683 <th class="td_left"><b>Field Description</b></th>
1684 </tr>
1685 <tr>
1686 <td>0-1</td>
1687 <td>constant "1"</td>
1688 <td class="td_left">These two bits must be the value 1 which identifies
1689 this as an instruction of format 1.</td>
1690 </tr>
1691 <tr>
1692 <td>2-7</td>
1693 <td><a href="#instructions">opcode</a></td>
1694 <td class="td_left">Specifies the opcode of the instruction. Note that
1695 the maximum opcode value is 63.</td>
1696 </tr>
1697 <tr>
1698 <td>8-19</td>
1699 <td><a href="#unsigned">unsigned</a></td>
1700 <td class="td_left">Specifies the slot number of the type for this
1701 instruction. Maximum slot number is 2<sup>12</sup>-1=4095.</td>
1702 </tr>
1703 <tr>
1704 <td>20-31</td>
1705 <td><a href="#unsigned">unsigned</a></td>
1706 <td class="td_left">Specifies the slot number of the value for the
1707 first operand. Maximum slot number is 2<sup>12</sup>-1=4095. Note that
1708 the value 2<sup>12</sup>-1 denotes zero operands.</td>
1709 </tr>
1710 </tbody>
1711</table>
1712<p><b>Instruction Format 2</b></p>
1713<p>This format encodes the opcode, type and two operands into a single <a
1714 href="#uint32_vbr">uint32_vbr</a> as follows:</p>
1715<table>
1716 <tbody>
1717 <tr>
1718 <th><b>Bits</b></th>
1719 <th><b>Type</b></th>
1720 <th class="td_left"><b>Field Description</b></th>
1721 </tr>
1722 <tr>
1723 <td>0-1</td>
1724 <td>constant "2"</td>
1725 <td class="td_left">These two bits must be the value 2 which identifies
1726 this as an instruction of format 2.</td>
1727 </tr>
1728 <tr>
1729 <td>2-7</td>
1730 <td><a href="#instructions">opcode</a></td>
1731 <td class="td_left">Specifies the opcode of the instruction. Note that
1732 the maximum opcode value is 63.</td>
1733 </tr>
1734 <tr>
1735 <td>8-15</td>
1736 <td><a href="#unsigned">unsigned</a></td>
1737 <td class="td_left">Specifies the slot number of the type for this
1738 instruction. Maximum slot number is 2<sup>8</sup>-1=255.</td>
1739 </tr>
1740 <tr>
1741 <td>16-23</td>
1742 <td><a href="#unsigned">unsigned</a></td>
1743 <td class="td_left">Specifies the slot number of the value for the first
1744 operand. Maximum slot number is 2<sup>8</sup>-1=255.</td>
1745 </tr>
1746 <tr>
1747 <td>24-31</td>
1748 <td><a href="#unsigned">unsigned</a></td>
1749 <td class="td_left">Specifies the slot number of the value for the second
1750 operand. Maximum slot number is 2<sup>8</sup>-1=255.</td>
1751 </tr>
1752 </tbody>
1753</table>
1754<p><b>Instruction Format 3</b></p>
1755<p>This format encodes the opcode, type and three operands into a
1756single <a href="#uint32_vbr">uint32_vbr</a> as follows:</p>
1757<table>
1758 <tbody>
1759 <tr>
1760 <th><b>Bits</b></th>
1761 <th><b>Type</b></th>
1762 <th class="td_left"><b>Field Description</b></th>
1763 </tr>
1764 <tr>
1765 <td>0-1</td>
1766 <td>constant "3"</td>
1767 <td class="td_left">These two bits must be the value 3 which identifies
1768 this as an instruction of format 3.</td>
1769 </tr>
1770 <tr>
1771 <td>2-7</td>
1772 <td><a href="#instructions">opcode</a></td>
1773 <td class="td_left">Specifies the opcode of the instruction. Note that
1774 the maximum opcode value is 63.</td>
1775 </tr>
1776 <tr>
1777 <td>8-13</td>
1778 <td><a href="#unsigned">unsigned</a></td>
1779 <td class="td_left">Specifies the slot number of the type for this
1780 instruction. Maximum slot number is 2<sup>6</sup>-1=63.</td>
1781 </tr>
1782 <tr>
1783 <td>14-19</td>
1784 <td><a href="#unsigned">unsigned</a></td>
1785 <td class="td_left">Specifies the slot number of the value for the first
1786 operand. Maximum slot number is 2<sup>6</sup>-1=63.</td>
1787 </tr>
1788 <tr>
1789 <td>20-25</td>
1790 <td><a href="#unsigned">unsigned</a></td>
1791 <td class="td_left">Specifies the slot number of the value for the second
1792 operand. Maximum slot number is 2<sup>6</sup>-1=63.</td>
1793 </tr>
1794 <tr>
1795 <td>26-31</td>
1796 <td><a href="#unsigned">unsigned</a></td>
1797 <td class="td_left">Specifies the slot number of the value for the third
1798 operand. Maximum slot number is 2<sup>6</sup>-1=63.</td>
1799 </tr>
1800 </tbody>
1801</table>
1802</div>
Reid Spencera5681772005-05-14 00:06:06 +00001803
Reid Spencer50026612004-05-22 02:28:36 +00001804<!-- _______________________________________________________________________ -->
Reid Spencerb39021b2004-05-23 17:05:09 +00001805<div class="doc_subsection"><a name="symtab">Symbol Table</a> </div>
Reid Spencer50026612004-05-22 02:28:36 +00001806<div class="doc_text">
Reid Spencer09daa632004-08-18 20:06:19 +00001807<p>A symbol table can be put out in conjunction with a module or a function. A
1808symbol table has a list of name/type associations followed by a list of
1809name/value associations. The name/value associations are organized into "type
1810planes" so that all values of a common type are listed together. Each type
1811plane starts with the number of entries in the plane and the type slot number
1812for all the values in that plane (so the type can be looked up in the global
1813type pool). For each entry in a type plane, the slot number of the value and
1814the name associated with that value are written. The format is given in the
1815table below. </p>
Reid Spencer2cc36152004-07-05 19:04:27 +00001816<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001817 <tbody>
1818 <tr>
1819 <th><b>Type</b></th>
1820 <th class="td_left"><b>Field Description</b></th>
1821 </tr>
1822 <tr>
1823 <td><a href="#block">block</a><br>
1824 </td>
1825 <td class="td_left">Symbol Table Identifier (0x04)</td>
1826 </tr>
1827 <tr>
Reid Spencer09daa632004-08-18 20:06:19 +00001828 <td><a href="#llist">llist</a>(<a href="#symtab_entry">type_entry</a>)</td>
Reid Spencer8996e552004-08-17 00:49:03 +00001829 <td class="td_left">A length list of symbol table entries for
1830 <tt>Type</tt>s
1831 </td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001832 </tr>
1833 <tr>
Reid Spencer8996e552004-08-17 00:49:03 +00001834 <td><a href="#zlist">llist</a>(<a href="#symtab_plane">symtab_plane</a>)</td>
Reid Spencer09daa632004-08-18 20:06:19 +00001835 <td class="td_left">A length list of "type planes" of symbol table
Reid Spencer8996e552004-08-17 00:49:03 +00001836 entries for <tt>Value</tt>s</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001837 </tr>
1838 </tbody>
Reid Spencerb39021b2004-05-23 17:05:09 +00001839</table>
Reid Spencer50026612004-05-22 02:28:36 +00001840</div>
Reid Spencer09daa632004-08-18 20:06:19 +00001841
1842<!-- _______________________________________________________________________ -->
1843<div class="doc_subsubsection"> <a name="type_entry">Symbol Table Type
1844Entry</a>
1845</div>
1846<div class="doc_text">
1847<p>A symbol table type entry associates a name with a type. The name is provided
1848simply as an array of chars. The type is provided as a type slot number (index)
1849into the global type pool. The format is given in the following table:</p>
1850<table>
1851 <tbody>
1852 <tr>
1853 <th><b>Type</b></th>
1854 <th class="td_left"><b>Field Description</b></th>
1855 </tr>
1856 <tr>
1857 <td><a href="#uint32_vbr">uint24_vbr</a></td>
1858 <td class="td_left">Type slot number of the type being given a
1859 name relative to the global type pool.
1860 </td>
1861 </tr>
1862 <tr>
1863 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1864 <td class="td_left">Length of the character array that follows.</td>
1865 </tr>
1866 <tr>
1867 <td><a href="#char">char</a>+</td>
1868 <td class="td_left">The characters of the name.</td>
1869 </tr>
1870 </tbody>
1871</table>
1872</div>
Reid Spencer51f31e02004-07-05 22:28:02 +00001873<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +00001874<div class="doc_subsubsection"> <a name="symtab_plane">Symbol Table
1875Plane</a>
Reid Spencer51f31e02004-07-05 22:28:02 +00001876</div>
1877<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +00001878<p>A symbol table plane provides the symbol table entries for all
1879values of a common type. The encoding is given in the following table:</p>
Reid Spencer51f31e02004-07-05 22:28:02 +00001880<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001881 <tbody>
1882 <tr>
1883 <th><b>Type</b></th>
1884 <th class="td_left"><b>Field Description</b></th>
1885 </tr>
1886 <tr>
1887 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1888 <td class="td_left">Number of entries in this plane.</td>
1889 </tr>
1890 <tr>
1891 <td><a href="#uint32_vbr">uint32_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +00001892 <td class="td_left">Type slot number of type for all values in this plane..</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001893 </tr>
1894 <tr>
Reid Spencer09daa632004-08-18 20:06:19 +00001895 <td><a href="#value_entry">value_entry</a>+</td>
1896 <td class="td_left">The symbol table entries for to associate values with
1897 names.</td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001898 </tr>
1899 </tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +00001900</table>
1901</div>
Reid Spencer51f31e02004-07-05 22:28:02 +00001902<!-- _______________________________________________________________________ -->
Reid Spencer09daa632004-08-18 20:06:19 +00001903<div class="doc_subsubsection"><a name="value_entry">Symbol Table Value
Reid Spencer9bd2be22004-07-29 00:13:04 +00001904Entry</a>
Reid Spencer51f31e02004-07-05 22:28:02 +00001905</div>
1906<div class="doc_text">
Reid Spencer09daa632004-08-18 20:06:19 +00001907<p>A symbol table value entry provides the assocation between a value and the
1908name given to the value. The value is referenced by its slot number. The
Reid Spencer9bd2be22004-07-29 00:13:04 +00001909format is given in the following table:</p>
Reid Spencer51f31e02004-07-05 22:28:02 +00001910<table>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001911 <tbody>
1912 <tr>
1913 <th><b>Type</b></th>
1914 <th class="td_left"><b>Field Description</b></th>
1915 </tr>
1916 <tr>
1917 <td><a href="#uint32_vbr">uint24_vbr</a></td>
Reid Spencer09daa632004-08-18 20:06:19 +00001918 <td class="td_left">Value slot number of the value being given a name.
Reid Spencer8996e552004-08-17 00:49:03 +00001919 </td>
Reid Spencer9bd2be22004-07-29 00:13:04 +00001920 </tr>
1921 <tr>
1922 <td><a href="#uint32_vbr">uint32_vbr</a></td>
1923 <td class="td_left">Length of the character array that follows.</td>
1924 </tr>
1925 <tr>
1926 <td><a href="#char">char</a>+</td>
1927 <td class="td_left">The characters of the name.</td>
1928 </tr>
1929 </tbody>
Reid Spencer51f31e02004-07-05 22:28:02 +00001930</table>
1931</div>
Reid Spencer09daa632004-08-18 20:06:19 +00001932
Reid Spencer7c76d332004-06-08 07:41:41 +00001933<!-- *********************************************************************** -->
Reid Spencer9bd2be22004-07-29 00:13:04 +00001934<div class="doc_section"> <a name="versiondiffs">Version Differences</a>
1935</div>
Reid Spencer7c76d332004-06-08 07:41:41 +00001936<!-- *********************************************************************** -->
1937<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +00001938<p>This section describes the differences in the Bytecode Format across
1939LLVM
1940versions. The versions are listed in reverse order because it assumes
1941the current version is as documented in the previous sections. Each
1942section here
Chris Lattner1cc070c2004-07-05 18:05:48 +00001943describes the differences between that version and the one that <i>follows</i>.
Reid Spencer7c76d332004-06-08 07:41:41 +00001944</p>
1945</div>
Chris Lattnera1dad812004-10-16 18:03:55 +00001946
Reid Spencer7c76d332004-06-08 07:41:41 +00001947<!-- _______________________________________________________________________ -->
Reid Spencer43dfdb72004-08-18 20:17:05 +00001948<div class="doc_subsection"><a name="vers13">Version 1.3 Differences From
Chris Lattnera1dad812004-10-16 18:03:55 +00001949 1.4</a></div>
Reid Spencercf549e12004-08-17 07:43:43 +00001950<!-- _______________________________________________________________________ -->
Chris Lattnera1dad812004-10-16 18:03:55 +00001951
1952<div class="doc_subsubsection">Unreachable Instruction</div>
1953<div class="doc_text">
1954 <p>The LLVM <a href="LangRef.html#i_unreachable">Unreachable</a> instruction
1955 was added in version 1.4 of LLVM. This caused all instruction numbers after
1956 it to shift down by one.</p>
1957</div>
1958
1959<div class="doc_subsubsection">Function Flags</div>
1960<div class="doc_text">
1961 <p>LLVM bytecode versions prior to 1.4 did not include the 5 bit offset
1962 in <a href="#funcfield">the function list</a> in the <a
1963 href="#globalinfo">Module Global Info</a> block.</p>
1964</div>
1965
1966<div class="doc_subsubsection">Function Flags</div>
1967<div class="doc_text">
1968 <p>LLVM bytecode versions prior to 1.4 did not include the 'undef' constant
1969 value, which affects the encoding of <a href="#constant">Constant
1970 Fields</a>.</p>
1971</div>
1972
Chris Lattner1dfc8e52004-10-16 00:29:30 +00001973<!--
Reid Spencercf549e12004-08-17 07:43:43 +00001974<div class="doc_subsubsection">Aligned Data</div>
1975<div class="doc_text">
1976 <p>In version 1.3, certain data items were aligned to 32-bit boundaries. In
1977 version 1.4, alignment of data was done away with completely. The need for
1978 alignment has gone away and the only thing it adds is bytecode file size
1979 overhead. In most cases this overhead was small. However, in functions with
1980 large numbers of format 0 instructions (GEPs and PHIs with lots of parameters)
1981 or regular instructions with large valued operands (e.g. because there's just
1982 a lot of instructions in the function) the overhead can be extreme. In one
1983 test case, the overhead was 44,000 bytes (34% of the total file size).
1984 Consequently in release 1.4, the decision was made to eliminate alignment
1985 altogether.</p>
1986 <p>In version 1.3 format, the following bytecode constructs were aligned (i.e.
1987 they were followed by one to three bytes of padding):</p>
1988 <ul>
1989 <li>All blocks.</li>
1990 <li>Instructions using the long format (format 0).</li>
1991 <li>All call instructions that called a var args function.</li>
1992 <li>The target triple (a string field at the end of the module block).</li>
1993 <li>The version field (immediately following the signature).</li>
1994 </ul>
1995 <p>None of these constructs are aligned in version 1.4</p>
1996</div>
Chris Lattner1dfc8e52004-10-16 00:29:30 +00001997-->
Chris Lattnera1dad812004-10-16 18:03:55 +00001998
Reid Spencercf549e12004-08-17 07:43:43 +00001999<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +00002000<div class="doc_subsection"><a name="vers12">Version 1.2 Differences
2001From 1.3</a></div>
Reid Spencer1ab929c2004-07-05 08:18:07 +00002002<!-- _______________________________________________________________________ -->
Chris Lattnera1dad812004-10-16 18:03:55 +00002003
Reid Spencer1ab929c2004-07-05 08:18:07 +00002004<div class="doc_subsubsection">Type Derives From Value</div>
Reid Spencer7c76d332004-06-08 07:41:41 +00002005<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +00002006<p>In version 1.2, the Type class in the LLVM IR derives from the Value
2007class. This is not the case in version 1.3. Consequently, in version
20081.2 the notion of a "Type Type" was used to write out values that were
2009Types. The types always occuped plane 12 (corresponding to the
2010TypeTyID) of any type planed set of values. In 1.3 this representation
2011is not convenient because the TypeTyID (12) is not present and its
2012value is now used for LabelTyID. Consequently, the data structures
2013written that involve types do so by writing all the types first and
2014then each of the value planes according to those types. In version 1.2,
2015the types would have been written intermingled with the values.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +00002016</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +00002017<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +00002018<div class="doc_subsubsection">Restricted getelementptr Types</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +00002019<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +00002020<p>In version 1.2, the getelementptr instruction required a ubyte type
2021index for accessing a structure field and a long type index for
2022accessing an array element. Consequently, it was only possible to
2023access structures of 255 or fewer elements. Starting in version 1.3,
2024this restriction was lifted. Structures must now be indexed with uint
2025constants. Arrays may now be indexed with int, uint, long, or ulong
2026typed values. The consequence of this was that the bytecode format had
2027to change in order to accommodate the larger range of structure indices.</p>
Reid Spencer7c76d332004-06-08 07:41:41 +00002028</div>
Reid Spencer7c76d332004-06-08 07:41:41 +00002029<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +00002030<div class="doc_subsubsection">Short Block Headers</div>
2031<div class="doc_text">
2032<p>In version 1.2, block headers were always 8 bytes being comprised of
2033both an unsigned integer type and an unsigned integer size. For very
2034small modules, these block headers turn out to be a large fraction of
2035the total bytecode file size. In an attempt to make these small files
2036smaller, the type and size information was encoded into a single
2037unsigned integer (4 bytes) comprised of 5 bits for the block type
2038(maximum 31 block types) and 27 bits for the block size (max
2039~134MBytes). These limits seemed sufficient for any blocks or sizes
2040forseen in the future. Note that the module block, which encloses all
2041the other blocks is still written as 8 bytes since bytecode files
2042larger than 134MBytes might be possible.</p>
2043</div>
2044<!-- _______________________________________________________________________ -->
2045<div class="doc_subsubsection">Dependent Libraries and Target Triples</div>
2046<div class="doc_text">
2047<p>In version 1.2, the bytecode format does not store module's target
2048triple or dependent. These fields have been added to the end of the <a
2049 href="#globalinfo">module global info block</a>. The purpose of these
2050fields is to allow a front end compiler to specifiy that the generated
2051module is specific to a particular target triple (operating
2052system/manufacturer/processor) which makes it non-portable; and to
2053allow front end compilers to specify the list of libraries that the
2054module depends on for successful linking.</p>
2055</div>
2056<!-- _______________________________________________________________________ -->
2057<div class="doc_subsubsection">Types Restricted to 24-bits</div>
2058<div class="doc_text">
2059<p>In version 1.2, type slot identifiers were written as 32-bit VBR
2060quantities. In 1.3 this has been reduced to 24-bits in order to ensure
2061that it is not possible to overflow the type field of a global variable
2062definition. 24-bits for type slot numbers is deemed sufficient for any
2063practical use of LLVM.</p>
2064</div>
2065<!-- _______________________________________________________________________ -->
2066<!-- _______________________________________________________________________ -->
2067<div class="doc_subsection"><a name="vers11">Version 1.1 Differences
2068From 1.2 </a></div>
Reid Spencer1ab929c2004-07-05 08:18:07 +00002069<!-- _______________________________________________________________________ -->
2070<div class="doc_subsubsection">Explicit Primitive Zeros</div>
Reid Spencer7c76d332004-06-08 07:41:41 +00002071<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +00002072<p>In version 1.1, the zero value for primitives was explicitly encoded
2073into the bytecode format. Since these zero values are constant values
2074in the LLVM IR and never change, there is no reason to explicitly
2075encode them. This explicit encoding was removed in version 1.2.</p>
Reid Spencer1ab929c2004-07-05 08:18:07 +00002076</div>
Reid Spencer1ab929c2004-07-05 08:18:07 +00002077<!-- _______________________________________________________________________ -->
2078<div class="doc_subsubsection">Inconsistent Module Global Info</div>
2079<div class="doc_text">
Reid Spencer9bd2be22004-07-29 00:13:04 +00002080<p>In version 1.1, the Module Global Info block was not aligned causing
2081the next block to be read in on an unaligned boundary. This problem was
2082corrected in version 1.2.<br>
2083<br>
2084</p>
Reid Spencer7c76d332004-06-08 07:41:41 +00002085</div>
Reid Spencer7c76d332004-06-08 07:41:41 +00002086<!-- _______________________________________________________________________ -->
Reid Spencer9bd2be22004-07-29 00:13:04 +00002087<div class="doc_subsection"><a name="vers10">Version 1.0 Differences
2088From 1.1</a></div>
Reid Spencer7c76d332004-06-08 07:41:41 +00002089<div class="doc_text">
Reid Spencer1ab929c2004-07-05 08:18:07 +00002090<p>None. Version 1.0 and 1.1 bytecode formats are identical.</p>
Reid Spencer7c76d332004-06-08 07:41:41 +00002091</div>
Reid Spencer50026612004-05-22 02:28:36 +00002092<!-- *********************************************************************** -->
2093<hr>
Reid Spencer9bd2be22004-07-29 00:13:04 +00002094<address> <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2095 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
2096<a href="http://validator.w3.org/check/referer"><img
2097 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!"></a>
2098<a href="mailto:rspencer@x10sys.com">Reid Spencer</a> and <a
2099 href="mailto:sabre@nondot.org">Chris Lattner</a><br>
Reid Spencer05fe4b02006-03-14 05:39:39 +00002100<a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
Reid Spencer9bd2be22004-07-29 00:13:04 +00002101Last modified: $Date$
Reid Spencer50026612004-05-22 02:28:36 +00002102</address>
Reid Spencer9bd2be22004-07-29 00:13:04 +00002103</body>
2104</html>