blob: 795b91a76ba92db5f56c5e6d927198cf3dcf99ac [file] [log] [blame]
Misha Brukman13fd15c2004-01-15 00:14:41 +00001<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
Chris Lattner261efe92003-11-25 01:02:51 +00003<html>
4<head>
5 <title>LLVM Programmer's Manual</title>
Misha Brukman13fd15c2004-01-15 00:14:41 +00006 <link rel="stylesheet" href="llvm.css" type="text/css">
Chris Lattner261efe92003-11-25 01:02:51 +00007</head>
Misha Brukman13fd15c2004-01-15 00:14:41 +00008<body>
9
10<div class="doc_title">
11 LLVM Programmer's Manual
12</div>
13
Chris Lattner9355b472002-09-06 02:50:58 +000014<ol>
Misha Brukman13fd15c2004-01-15 00:14:41 +000015 <li><a href="#introduction">Introduction</a></li>
Chris Lattner9355b472002-09-06 02:50:58 +000016 <li><a href="#general">General Information</a>
Chris Lattner261efe92003-11-25 01:02:51 +000017 <ul>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +000018 <li><a href="#stl">The C++ Standard Template Library</a></li>
19<!--
20 <li>The <tt>-time-passes</tt> option</li>
21 <li>How to use the LLVM Makefile system</li>
22 <li>How to write a regression test</li>
Chris Lattner61db4652004-12-08 19:05:44 +000023
Reid Spencerfe8f4ff2004-11-01 09:02:53 +000024-->
Chris Lattner84b7f8d2003-08-01 22:20:59 +000025 </ul>
Chris Lattner261efe92003-11-25 01:02:51 +000026 </li>
27 <li><a href="#apis">Important and useful LLVM APIs</a>
28 <ul>
29 <li><a href="#isa">The <tt>isa&lt;&gt;</tt>, <tt>cast&lt;&gt;</tt>
30and <tt>dyn_cast&lt;&gt;</tt> templates</a> </li>
Misha Brukman2c122ce2005-11-01 21:12:49 +000031 <li><a href="#DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt>
Chris Lattner261efe92003-11-25 01:02:51 +000032option</a>
33 <ul>
34 <li><a href="#DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt>
35and the <tt>-debug-only</tt> option</a> </li>
36 </ul>
37 </li>
Chris Lattner0be6fdf2006-12-19 21:46:21 +000038 <li><a href="#Statistic">The <tt>Statistic</tt> class &amp; <tt>-stats</tt>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +000039option</a></li>
40<!--
41 <li>The <tt>InstVisitor</tt> template
42 <li>The general graph API
43-->
Chris Lattnerf623a082005-10-17 01:36:23 +000044 <li><a href="#ViewGraph">Viewing graphs while debugging code</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +000045 </ul>
46 </li>
Chris Lattner098129a2007-02-03 03:04:03 +000047 <li><a href="#datastructure">Picking the Right Data Structure for a Task</a>
48 <ul>
Chris Lattner74c4ca12007-02-03 07:59:07 +000049 <li><a href="#ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
50 <ul>
51 <li><a href="#dss_fixedarrays">Fixed Size Arrays</a></li>
52 <li><a href="#dss_heaparrays">Heap Allocated Arrays</a></li>
53 <li><a href="#dss_smallvector">"llvm/ADT/SmallVector.h"</a></li>
54 <li><a href="#dss_vector">&lt;vector&gt;</a></li>
55 <li><a href="#dss_deque">&lt;deque&gt;</a></li>
56 <li><a href="#dss_list">&lt;list&gt;</a></li>
57 <li><a href="#dss_ilist">llvm/ADT/ilist</a></li>
Chris Lattnerc5722432007-02-03 19:49:31 +000058 <li><a href="#dss_other">Other Sequential Container Options</a></li>
Chris Lattner098129a2007-02-03 03:04:03 +000059 </ul></li>
Chris Lattner74c4ca12007-02-03 07:59:07 +000060 <li><a href="#ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
61 <ul>
62 <li><a href="#dss_sortedvectorset">A sorted 'vector'</a></li>
63 <li><a href="#dss_smallset">"llvm/ADT/SmallSet.h"</a></li>
64 <li><a href="#dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a></li>
65 <li><a href="#dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a></li>
66 <li><a href="#dss_set">&lt;set&gt;</a></li>
67 <li><a href="#dss_setvector">"llvm/ADT/SetVector.h"</a></li>
Chris Lattnerc5722432007-02-03 19:49:31 +000068 <li><a href="#dss_uniquevector">"llvm/ADT/UniqueVector.h"</a></li>
69 <li><a href="#dss_otherset">Other Set-Like ContainerOptions</a></li>
Chris Lattner74c4ca12007-02-03 07:59:07 +000070 </ul></li>
Chris Lattnerf3692522007-02-03 19:51:56 +000071 <li><a href="#ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
72 <ul>
73 <li><a href="#dss_sortedvectormap">A sorted 'vector'</a></li>
Chris Lattner796f9fa2007-02-08 19:14:21 +000074 <li><a href="#dss_stringmap">"llvm/ADT/StringMap.h"</a></li>
Chris Lattnerf3692522007-02-03 19:51:56 +000075 <li><a href="#dss_indexedmap">"llvm/ADT/IndexedMap.h"</a></li>
76 <li><a href="#dss_densemap">"llvm/ADT/DenseMap.h"</a></li>
77 <li><a href="#dss_map">&lt;map&gt;</a></li>
78 <li><a href="#dss_othermap">Other Map-Like Container Options</a></li>
79 </ul></li>
Daniel Berlin1939ace2007-09-24 17:52:25 +000080 <li><a href="#ds_bit">BitVector-like containers</a>
81 <ul>
82 <li><a href="#dss_bitvector">A dense bitvector</a></li>
83 <li><a href="#dss_sparsebitvector">A sparse bitvector</a></li>
84 </ul></li>
Chris Lattner74c4ca12007-02-03 07:59:07 +000085 </ul>
Chris Lattner098129a2007-02-03 03:04:03 +000086 </li>
Chris Lattnerae7f7592002-09-06 18:31:18 +000087 <li><a href="#common">Helpful Hints for Common Operations</a>
Chris Lattnerae7f7592002-09-06 18:31:18 +000088 <ul>
Chris Lattner261efe92003-11-25 01:02:51 +000089 <li><a href="#inspection">Basic Inspection and Traversal Routines</a>
90 <ul>
91 <li><a href="#iterate_function">Iterating over the <tt>BasicBlock</tt>s
92in a <tt>Function</tt></a> </li>
93 <li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
94in a <tt>BasicBlock</tt></a> </li>
95 <li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
96in a <tt>Function</tt></a> </li>
97 <li><a href="#iterate_convert">Turning an iterator into a
98class pointer</a> </li>
99 <li><a href="#iterate_complex">Finding call sites: a more
100complex example</a> </li>
101 <li><a href="#calls_and_invokes">Treating calls and invokes
102the same way</a> </li>
103 <li><a href="#iterate_chains">Iterating over def-use &amp;
104use-def chains</a> </li>
105 </ul>
106 </li>
107 <li><a href="#simplechanges">Making simple changes</a>
108 <ul>
109 <li><a href="#schanges_creating">Creating and inserting new
110 <tt>Instruction</tt>s</a> </li>
111 <li><a href="#schanges_deleting">Deleting <tt>Instruction</tt>s</a> </li>
112 <li><a href="#schanges_replacing">Replacing an <tt>Instruction</tt>
113with another <tt>Value</tt></a> </li>
Tanya Lattnerb011c662007-06-20 18:33:15 +0000114 <li><a href="#schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a> </li>
Chris Lattner261efe92003-11-25 01:02:51 +0000115 </ul>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000116 </li>
Chris Lattnerae7f7592002-09-06 18:31:18 +0000117<!--
118 <li>Working with the Control Flow Graph
119 <ul>
120 <li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
121 <li>
122 <li>
123 </ul>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000124-->
Chris Lattner261efe92003-11-25 01:02:51 +0000125 </ul>
126 </li>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000127
128 <li><a href="#advanced">Advanced Topics</a>
129 <ul>
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000130 <li><a href="#TypeResolve">LLVM Type Resolution</a>
131 <ul>
132 <li><a href="#BuildRecType">Basic Recursive Type Construction</a></li>
133 <li><a href="#refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a></li>
134 <li><a href="#PATypeHolder">The PATypeHolder Class</a></li>
135 <li><a href="#AbstractTypeUser">The AbstractTypeUser Class</a></li>
136 </ul></li>
137
Chris Lattner263a98e2007-02-16 04:37:31 +0000138 <li><a href="#SymbolTable">The <tt>ValueSymbolTable</tt> and <tt>TypeSymbolTable</tt> classes </a></li>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000139 </ul></li>
140
Joel Stanley9b96c442002-09-06 21:55:13 +0000141 <li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
Chris Lattner9355b472002-09-06 02:50:58 +0000142 <ul>
Reid Spencer303c4b42007-01-12 17:26:25 +0000143 <li><a href="#Type">The <tt>Type</tt> class</a> </li>
Chris Lattner2b78d962007-02-03 20:02:25 +0000144 <li><a href="#Module">The <tt>Module</tt> class</a></li>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000145 <li><a href="#Value">The <tt>Value</tt> class</a>
Chris Lattner2b78d962007-02-03 20:02:25 +0000146 <ul>
147 <li><a href="#User">The <tt>User</tt> class</a>
Chris Lattner9355b472002-09-06 02:50:58 +0000148 <ul>
Chris Lattner2b78d962007-02-03 20:02:25 +0000149 <li><a href="#Instruction">The <tt>Instruction</tt> class</a></li>
150 <li><a href="#Constant">The <tt>Constant</tt> class</a>
151 <ul>
152 <li><a href="#GlobalValue">The <tt>GlobalValue</tt> class</a>
Chris Lattner261efe92003-11-25 01:02:51 +0000153 <ul>
Chris Lattner2b78d962007-02-03 20:02:25 +0000154 <li><a href="#Function">The <tt>Function</tt> class</a></li>
155 <li><a href="#GlobalVariable">The <tt>GlobalVariable</tt> class</a></li>
156 </ul>
157 </li>
158 </ul>
159 </li>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000160 </ul>
Chris Lattner2b78d962007-02-03 20:02:25 +0000161 </li>
162 <li><a href="#BasicBlock">The <tt>BasicBlock</tt> class</a></li>
163 <li><a href="#Argument">The <tt>Argument</tt> class</a></li>
164 </ul>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000165 </li>
166 </ul>
Chris Lattner261efe92003-11-25 01:02:51 +0000167 </li>
Chris Lattner9355b472002-09-06 02:50:58 +0000168</ol>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000169
Chris Lattner69bf8a92004-05-23 21:06:58 +0000170<div class="doc_author">
171 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
Chris Lattner94c43592004-05-26 16:52:55 +0000172 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>,
173 <a href="mailto:jstanley@cs.uiuc.edu">Joel Stanley</a>, and
174 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000175</div>
176
Chris Lattner9355b472002-09-06 02:50:58 +0000177<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000178<div class="doc_section">
179 <a name="introduction">Introduction </a>
180</div>
Chris Lattner9355b472002-09-06 02:50:58 +0000181<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000182
183<div class="doc_text">
184
185<p>This document is meant to highlight some of the important classes and
Chris Lattner261efe92003-11-25 01:02:51 +0000186interfaces available in the LLVM source-base. This manual is not
187intended to explain what LLVM is, how it works, and what LLVM code looks
188like. It assumes that you know the basics of LLVM and are interested
189in writing transformations or otherwise analyzing or manipulating the
Misha Brukman13fd15c2004-01-15 00:14:41 +0000190code.</p>
191
192<p>This document should get you oriented so that you can find your
Chris Lattner261efe92003-11-25 01:02:51 +0000193way in the continuously growing source code that makes up the LLVM
194infrastructure. Note that this manual is not intended to serve as a
195replacement for reading the source code, so if you think there should be
196a method in one of these classes to do something, but it's not listed,
197check the source. Links to the <a href="/doxygen/">doxygen</a> sources
198are provided to make this as easy as possible.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000199
200<p>The first section of this document describes general information that is
201useful to know when working in the LLVM infrastructure, and the second describes
202the Core LLVM classes. In the future this manual will be extended with
203information describing how to use extension libraries, such as dominator
204information, CFG traversal routines, and useful utilities like the <tt><a
205href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.</p>
206
207</div>
208
Chris Lattner9355b472002-09-06 02:50:58 +0000209<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000210<div class="doc_section">
211 <a name="general">General Information</a>
212</div>
213<!-- *********************************************************************** -->
214
215<div class="doc_text">
216
217<p>This section contains general information that is useful if you are working
218in the LLVM source-base, but that isn't specific to any particular API.</p>
219
220</div>
221
222<!-- ======================================================================= -->
223<div class="doc_subsection">
224 <a name="stl">The C++ Standard Template Library</a>
225</div>
226
227<div class="doc_text">
228
229<p>LLVM makes heavy use of the C++ Standard Template Library (STL),
Chris Lattner261efe92003-11-25 01:02:51 +0000230perhaps much more than you are used to, or have seen before. Because of
231this, you might want to do a little background reading in the
232techniques used and capabilities of the library. There are many good
233pages that discuss the STL, and several books on the subject that you
Misha Brukman13fd15c2004-01-15 00:14:41 +0000234can get, so it will not be discussed in this document.</p>
235
236<p>Here are some useful links:</p>
237
238<ol>
239
240<li><a href="http://www.dinkumware.com/refxcpp.html">Dinkumware C++ Library
241reference</a> - an excellent reference for the STL and other parts of the
242standard C++ library.</li>
243
244<li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
Tanya Lattner09cf73c2004-06-22 04:24:55 +0000245O'Reilly book in the making. It has a decent
246Standard Library
247Reference that rivals Dinkumware's, and is unfortunately no longer free since the book has been
Misha Brukman13fd15c2004-01-15 00:14:41 +0000248published.</li>
249
250<li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
251Questions</a></li>
252
253<li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
254Contains a useful <a
255href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
256STL</a>.</li>
257
258<li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
259Page</a></li>
260
Tanya Lattner79445ba2004-12-08 18:34:56 +0000261<li><a href="http://64.78.49.204/">
Reid Spencer096603a2004-05-26 08:41:35 +0000262Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
263the book).</a></li>
264
Misha Brukman13fd15c2004-01-15 00:14:41 +0000265</ol>
266
267<p>You are also encouraged to take a look at the <a
268href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
269to write maintainable code more than where to put your curly braces.</p>
270
271</div>
272
273<!-- ======================================================================= -->
274<div class="doc_subsection">
275 <a name="stl">Other useful references</a>
276</div>
277
278<div class="doc_text">
279
Misha Brukman13fd15c2004-01-15 00:14:41 +0000280<ol>
281<li><a href="http://www.psc.edu/%7Esemke/cvs_branches.html">CVS
Chris Lattner261efe92003-11-25 01:02:51 +0000282Branch and Tag Primer</a></li>
Misha Brukmana0f71e42004-06-18 18:39:00 +0000283<li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
284static and shared libraries across platforms</a></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000285</ol>
286
287</div>
288
Chris Lattner9355b472002-09-06 02:50:58 +0000289<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000290<div class="doc_section">
291 <a name="apis">Important and useful LLVM APIs</a>
292</div>
293<!-- *********************************************************************** -->
294
295<div class="doc_text">
296
297<p>Here we highlight some LLVM APIs that are generally useful and good to
298know about when writing transformations.</p>
299
300</div>
301
302<!-- ======================================================================= -->
303<div class="doc_subsection">
Misha Brukman2c122ce2005-11-01 21:12:49 +0000304 <a name="isa">The <tt>isa&lt;&gt;</tt>, <tt>cast&lt;&gt;</tt> and
305 <tt>dyn_cast&lt;&gt;</tt> templates</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000306</div>
307
308<div class="doc_text">
309
310<p>The LLVM source-base makes extensive use of a custom form of RTTI.
Chris Lattner261efe92003-11-25 01:02:51 +0000311These templates have many similarities to the C++ <tt>dynamic_cast&lt;&gt;</tt>
312operator, but they don't have some drawbacks (primarily stemming from
313the fact that <tt>dynamic_cast&lt;&gt;</tt> only works on classes that
314have a v-table). Because they are used so often, you must know what they
315do and how they work. All of these templates are defined in the <a
Chris Lattner695b78b2005-04-26 22:56:16 +0000316 href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000317file (note that you very rarely have to include this file directly).</p>
318
319<dl>
320 <dt><tt>isa&lt;&gt;</tt>: </dt>
321
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000322 <dd><p>The <tt>isa&lt;&gt;</tt> operator works exactly like the Java
Misha Brukman13fd15c2004-01-15 00:14:41 +0000323 "<tt>instanceof</tt>" operator. It returns true or false depending on whether
324 a reference or pointer points to an instance of the specified class. This can
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000325 be very useful for constraint checking of various sorts (example below).</p>
326 </dd>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000327
328 <dt><tt>cast&lt;&gt;</tt>: </dt>
329
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000330 <dd><p>The <tt>cast&lt;&gt;</tt> operator is a "checked cast" operation. It
Misha Brukman13fd15c2004-01-15 00:14:41 +0000331 converts a pointer or reference from a base class to a derived cast, causing
332 an assertion failure if it is not really an instance of the right type. This
333 should be used in cases where you have some information that makes you believe
334 that something is of the right type. An example of the <tt>isa&lt;&gt;</tt>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000335 and <tt>cast&lt;&gt;</tt> template is:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000336
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000337<div class="doc_code">
338<pre>
339static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
340 if (isa&lt;<a href="#Constant">Constant</a>&gt;(V) || isa&lt;<a href="#Argument">Argument</a>&gt;(V) || isa&lt;<a href="#GlobalValue">GlobalValue</a>&gt;(V))
341 return true;
Chris Lattner69bf8a92004-05-23 21:06:58 +0000342
Bill Wendling82e2eea2006-10-11 18:00:22 +0000343 // <i>Otherwise, it must be an instruction...</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000344 return !L-&gt;contains(cast&lt;<a href="#Instruction">Instruction</a>&gt;(V)-&gt;getParent());
345}
346</pre>
347</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000348
349 <p>Note that you should <b>not</b> use an <tt>isa&lt;&gt;</tt> test followed
350 by a <tt>cast&lt;&gt;</tt>, for that use the <tt>dyn_cast&lt;&gt;</tt>
351 operator.</p>
352
353 </dd>
354
355 <dt><tt>dyn_cast&lt;&gt;</tt>:</dt>
356
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000357 <dd><p>The <tt>dyn_cast&lt;&gt;</tt> operator is a "checking cast" operation.
358 It checks to see if the operand is of the specified type, and if so, returns a
Misha Brukman13fd15c2004-01-15 00:14:41 +0000359 pointer to it (this operator does not work with references). If the operand is
360 not of the correct type, a null pointer is returned. Thus, this works very
Misha Brukman2c122ce2005-11-01 21:12:49 +0000361 much like the <tt>dynamic_cast&lt;&gt;</tt> operator in C++, and should be
362 used in the same circumstances. Typically, the <tt>dyn_cast&lt;&gt;</tt>
363 operator is used in an <tt>if</tt> statement or some other flow control
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000364 statement like this:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000365
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000366<div class="doc_code">
367<pre>
368if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast&lt;<a href="#AllocationInst">AllocationInst</a>&gt;(Val)) {
Bill Wendling82e2eea2006-10-11 18:00:22 +0000369 // <i>...</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000370}
371</pre>
372</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000373
Misha Brukman2c122ce2005-11-01 21:12:49 +0000374 <p>This form of the <tt>if</tt> statement effectively combines together a call
375 to <tt>isa&lt;&gt;</tt> and a call to <tt>cast&lt;&gt;</tt> into one
376 statement, which is very convenient.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000377
Misha Brukman2c122ce2005-11-01 21:12:49 +0000378 <p>Note that the <tt>dyn_cast&lt;&gt;</tt> operator, like C++'s
379 <tt>dynamic_cast&lt;&gt;</tt> or Java's <tt>instanceof</tt> operator, can be
380 abused. In particular, you should not use big chained <tt>if/then/else</tt>
381 blocks to check for lots of different variants of classes. If you find
382 yourself wanting to do this, it is much cleaner and more efficient to use the
383 <tt>InstVisitor</tt> class to dispatch over the instruction type directly.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000384
Misha Brukman2c122ce2005-11-01 21:12:49 +0000385 </dd>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000386
Misha Brukman2c122ce2005-11-01 21:12:49 +0000387 <dt><tt>cast_or_null&lt;&gt;</tt>: </dt>
388
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000389 <dd><p>The <tt>cast_or_null&lt;&gt;</tt> operator works just like the
Misha Brukman2c122ce2005-11-01 21:12:49 +0000390 <tt>cast&lt;&gt;</tt> operator, except that it allows for a null pointer as an
391 argument (which it then propagates). This can sometimes be useful, allowing
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000392 you to combine several null checks into one.</p></dd>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000393
Misha Brukman2c122ce2005-11-01 21:12:49 +0000394 <dt><tt>dyn_cast_or_null&lt;&gt;</tt>: </dt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000395
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000396 <dd><p>The <tt>dyn_cast_or_null&lt;&gt;</tt> operator works just like the
Misha Brukman2c122ce2005-11-01 21:12:49 +0000397 <tt>dyn_cast&lt;&gt;</tt> operator, except that it allows for a null pointer
398 as an argument (which it then propagates). This can sometimes be useful,
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000399 allowing you to combine several null checks into one.</p></dd>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000400
Misha Brukman2c122ce2005-11-01 21:12:49 +0000401</dl>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000402
403<p>These five templates can be used with any classes, whether they have a
404v-table or not. To add support for these templates, you simply need to add
405<tt>classof</tt> static methods to the class you are interested casting
406to. Describing this is currently outside the scope of this document, but there
407are lots of examples in the LLVM source base.</p>
408
409</div>
410
411<!-- ======================================================================= -->
412<div class="doc_subsection">
Misha Brukman2c122ce2005-11-01 21:12:49 +0000413 <a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000414</div>
415
416<div class="doc_text">
417
418<p>Often when working on your pass you will put a bunch of debugging printouts
419and other code into your pass. After you get it working, you want to remove
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000420it, but you may need it again in the future (to work out new bugs that you run
Misha Brukman13fd15c2004-01-15 00:14:41 +0000421across).</p>
422
423<p> Naturally, because of this, you don't want to delete the debug printouts,
424but you don't want them to always be noisy. A standard compromise is to comment
425them out, allowing you to enable them if you need them in the future.</p>
426
Chris Lattner695b78b2005-04-26 22:56:16 +0000427<p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
Misha Brukman13fd15c2004-01-15 00:14:41 +0000428file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
429this problem. Basically, you can put arbitrary code into the argument of the
430<tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
431tool) is run with the '<tt>-debug</tt>' command line argument:</p>
432
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000433<div class="doc_code">
434<pre>
Bill Wendling832171c2006-12-07 20:04:42 +0000435DOUT &lt;&lt; "I am here!\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000436</pre>
437</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000438
439<p>Then you can run your pass like this:</p>
440
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000441<div class="doc_code">
442<pre>
443$ opt &lt; a.bc &gt; /dev/null -mypass
Bill Wendling82e2eea2006-10-11 18:00:22 +0000444<i>&lt;no output&gt;</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000445$ opt &lt; a.bc &gt; /dev/null -mypass -debug
446I am here!
447</pre>
448</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000449
450<p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
451to not have to create "yet another" command line option for the debug output for
452your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
453so they do not cause a performance impact at all (for the same reason, they
454should also not contain side-effects!).</p>
455
456<p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
457enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
458"<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
459program hasn't been started yet, you can always just run it with
460<tt>-debug</tt>.</p>
461
462</div>
463
464<!-- _______________________________________________________________________ -->
465<div class="doc_subsubsection">
Chris Lattnerc9151082005-04-26 22:57:07 +0000466 <a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
Misha Brukman13fd15c2004-01-15 00:14:41 +0000467 the <tt>-debug-only</tt> option</a>
468</div>
469
470<div class="doc_text">
471
472<p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
473just turns on <b>too much</b> information (such as when working on the code
474generator). If you want to enable debug information with more fine-grained
475control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
476option as follows:</p>
477
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000478<div class="doc_code">
479<pre>
Bill Wendling832171c2006-12-07 20:04:42 +0000480DOUT &lt;&lt; "No debug type\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000481#undef DEBUG_TYPE
482#define DEBUG_TYPE "foo"
Bill Wendling832171c2006-12-07 20:04:42 +0000483DOUT &lt;&lt; "'foo' debug type\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000484#undef DEBUG_TYPE
485#define DEBUG_TYPE "bar"
Bill Wendling832171c2006-12-07 20:04:42 +0000486DOUT &lt;&lt; "'bar' debug type\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000487#undef DEBUG_TYPE
488#define DEBUG_TYPE ""
Bill Wendling832171c2006-12-07 20:04:42 +0000489DOUT &lt;&lt; "No debug type (2)\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000490</pre>
491</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000492
493<p>Then you can run your pass like this:</p>
494
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000495<div class="doc_code">
496<pre>
497$ opt &lt; a.bc &gt; /dev/null -mypass
Bill Wendling82e2eea2006-10-11 18:00:22 +0000498<i>&lt;no output&gt;</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000499$ opt &lt; a.bc &gt; /dev/null -mypass -debug
500No debug type
501'foo' debug type
502'bar' debug type
503No debug type (2)
504$ opt &lt; a.bc &gt; /dev/null -mypass -debug-only=foo
505'foo' debug type
506$ opt &lt; a.bc &gt; /dev/null -mypass -debug-only=bar
507'bar' debug type
508</pre>
509</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000510
511<p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
512a file, to specify the debug type for the entire module (if you do this before
Chris Lattner695b78b2005-04-26 22:56:16 +0000513you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
Misha Brukman13fd15c2004-01-15 00:14:41 +0000514<tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
515"bar", because there is no system in place to ensure that names do not
516conflict. If two different modules use the same string, they will all be turned
517on when the name is specified. This allows, for example, all debug information
518for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
Chris Lattner261efe92003-11-25 01:02:51 +0000519even if the source lives in multiple files.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000520
521</div>
522
523<!-- ======================================================================= -->
524<div class="doc_subsection">
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000525 <a name="Statistic">The <tt>Statistic</tt> class &amp; <tt>-stats</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000526 option</a>
527</div>
528
529<div class="doc_text">
530
531<p>The "<tt><a
Chris Lattner695b78b2005-04-26 22:56:16 +0000532href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000533provides a class named <tt>Statistic</tt> that is used as a unified way to
Misha Brukman13fd15c2004-01-15 00:14:41 +0000534keep track of what the LLVM compiler is doing and how effective various
535optimizations are. It is useful to see what optimizations are contributing to
536making a particular program run faster.</p>
537
538<p>Often you may run your pass on some big program, and you're interested to see
539how many times it makes a certain transformation. Although you can do this with
540hand inspection, or some ad-hoc method, this is a real pain and not very useful
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000541for big programs. Using the <tt>Statistic</tt> class makes it very easy to
Misha Brukman13fd15c2004-01-15 00:14:41 +0000542keep track of this information, and the calculated information is presented in a
543uniform manner with the rest of the passes being executed.</p>
544
545<p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
546it are as follows:</p>
547
548<ol>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000549 <li><p>Define your statistic like this:</p>
550
551<div class="doc_code">
552<pre>
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000553#define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
554STATISTIC(NumXForms, "The # of times I did stuff");
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000555</pre>
556</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000557
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000558 <p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
559 specified by the first argument. The pass name is taken from the DEBUG_TYPE
560 macro, and the description is taken from the second argument. The variable
Reid Spencer06565dc2007-01-12 17:11:23 +0000561 defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000562
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000563 <li><p>Whenever you make a transformation, bump the counter:</p>
564
565<div class="doc_code">
566<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +0000567++NumXForms; // <i>I did stuff!</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000568</pre>
569</div>
570
Chris Lattner261efe92003-11-25 01:02:51 +0000571 </li>
572 </ol>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000573
574 <p>That's all you have to do. To get '<tt>opt</tt>' to print out the
575 statistics gathered, use the '<tt>-stats</tt>' option:</p>
576
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000577<div class="doc_code">
578<pre>
579$ opt -stats -mypassname &lt; program.bc &gt; /dev/null
Bill Wendling82e2eea2006-10-11 18:00:22 +0000580<i>... statistics output ...</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000581</pre>
582</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000583
Reid Spencer6b6c73e2007-02-09 16:00:28 +0000584 <p> When running <tt>opt</tt> on a C file from the SPEC benchmark
Chris Lattner261efe92003-11-25 01:02:51 +0000585suite, it gives a report that looks like this:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000586
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000587<div class="doc_code">
588<pre>
Gabor Greif04367bf2007-07-06 22:07:22 +0000589 7646 bitcodewriter - Number of normal instructions
590 725 bitcodewriter - Number of oversized instructions
591 129996 bitcodewriter - Number of bitcode bytes written
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000592 2817 raise - Number of insts DCEd or constprop'd
593 3213 raise - Number of cast-of-self removed
594 5046 raise - Number of expression trees converted
595 75 raise - Number of other getelementptr's formed
596 138 raise - Number of load/store peepholes
597 42 deadtypeelim - Number of unused typenames removed from symtab
598 392 funcresolve - Number of varargs functions resolved
599 27 globaldce - Number of global variables removed
600 2 adce - Number of basic blocks removed
601 134 cee - Number of branches revectored
602 49 cee - Number of setcc instruction eliminated
603 532 gcse - Number of loads removed
604 2919 gcse - Number of instructions removed
605 86 indvars - Number of canonical indvars added
606 87 indvars - Number of aux indvars removed
607 25 instcombine - Number of dead inst eliminate
608 434 instcombine - Number of insts combined
609 248 licm - Number of load insts hoisted
610 1298 licm - Number of insts hoisted to a loop pre-header
611 3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
612 75 mem2reg - Number of alloca's promoted
613 1444 cfgsimplify - Number of blocks simplified
614</pre>
615</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000616
617<p>Obviously, with so many optimizations, having a unified framework for this
618stuff is very nice. Making your pass fit well into the framework makes it more
619maintainable and useful.</p>
620
621</div>
622
Chris Lattnerf623a082005-10-17 01:36:23 +0000623<!-- ======================================================================= -->
624<div class="doc_subsection">
625 <a name="ViewGraph">Viewing graphs while debugging code</a>
626</div>
627
628<div class="doc_text">
629
630<p>Several of the important data structures in LLVM are graphs: for example
631CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
632LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
633<a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
634DAGs</a>. In many cases, while debugging various parts of the compiler, it is
635nice to instantly visualize these graphs.</p>
636
637<p>LLVM provides several callbacks that are available in a debug build to do
638exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
639the current LLVM tool will pop up a window containing the CFG for the function
640where each basic block is a node in the graph, and each node contains the
641instructions in the block. Similarly, there also exists
642<tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
643<tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
644and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
Jim Laskey543a0ee2006-10-02 12:28:07 +0000645you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
Chris Lattnerf623a082005-10-17 01:36:23 +0000646up a window. Alternatively, you can sprinkle calls to these functions in your
647code in places you want to debug.</p>
648
649<p>Getting this to work requires a small amount of configuration. On Unix
650systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
651toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
652Mac OS/X, download and install the Mac OS/X <a
653href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
Reid Spencer128a7a72007-02-03 21:06:43 +0000654<tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
Chris Lattnerf623a082005-10-17 01:36:23 +0000655it) to your path. Once in your system and path are set up, rerun the LLVM
656configure script and rebuild LLVM to enable this functionality.</p>
657
Jim Laskey543a0ee2006-10-02 12:28:07 +0000658<p><tt>SelectionDAG</tt> has been extended to make it easier to locate
659<i>interesting</i> nodes in large complex graphs. From gdb, if you
660<tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
Reid Spencer128a7a72007-02-03 21:06:43 +0000661next <tt>call DAG.viewGraph()</tt> would highlight the node in the
Jim Laskey543a0ee2006-10-02 12:28:07 +0000662specified color (choices of colors can be found at <a
Chris Lattner302da1e2007-02-03 03:05:57 +0000663href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
Jim Laskey543a0ee2006-10-02 12:28:07 +0000664complex node attributes can be provided with <tt>call
665DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
666found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
667Attributes</a>.) If you want to restart and clear all the current graph
668attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
669
Chris Lattnerf623a082005-10-17 01:36:23 +0000670</div>
671
Chris Lattner098129a2007-02-03 03:04:03 +0000672<!-- *********************************************************************** -->
673<div class="doc_section">
674 <a name="datastructure">Picking the Right Data Structure for a Task</a>
675</div>
676<!-- *********************************************************************** -->
677
678<div class="doc_text">
679
Reid Spencer128a7a72007-02-03 21:06:43 +0000680<p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
681 and we commonly use STL data structures. This section describes the trade-offs
Chris Lattner098129a2007-02-03 03:04:03 +0000682 you should consider when you pick one.</p>
683
684<p>
685The first step is a choose your own adventure: do you want a sequential
686container, a set-like container, or a map-like container? The most important
687thing when choosing a container is the algorithmic properties of how you plan to
688access the container. Based on that, you should use:</p>
689
690<ul>
Reid Spencer128a7a72007-02-03 21:06:43 +0000691<li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
Chris Lattner098129a2007-02-03 03:04:03 +0000692 of an value based on another value. Map-like containers also support
693 efficient queries for containment (whether a key is in the map). Map-like
694 containers generally do not support efficient reverse mapping (values to
695 keys). If you need that, use two maps. Some map-like containers also
696 support efficient iteration through the keys in sorted order. Map-like
697 containers are the most expensive sort, only use them if you need one of
698 these capabilities.</li>
699
700<li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
701 stuff into a container that automatically eliminates duplicates. Some
702 set-like containers support efficient iteration through the elements in
703 sorted order. Set-like containers are more expensive than sequential
704 containers.
705</li>
706
707<li>a <a href="#ds_sequential">sequential</a> container provides
708 the most efficient way to add elements and keeps track of the order they are
709 added to the collection. They permit duplicates and support efficient
Reid Spencer128a7a72007-02-03 21:06:43 +0000710 iteration, but do not support efficient look-up based on a key.
Chris Lattner098129a2007-02-03 03:04:03 +0000711</li>
712
Daniel Berlin1939ace2007-09-24 17:52:25 +0000713<li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
714 perform set operations on sets of numeric id's, while automatically
715 eliminating duplicates. Bit containers require a maximum of 1 bit for each
716 identifier you want to store.
717</li>
Chris Lattner098129a2007-02-03 03:04:03 +0000718</ul>
719
720<p>
Reid Spencer128a7a72007-02-03 21:06:43 +0000721Once the proper category of container is determined, you can fine tune the
Chris Lattner098129a2007-02-03 03:04:03 +0000722memory use, constant factors, and cache behaviors of access by intelligently
Reid Spencer128a7a72007-02-03 21:06:43 +0000723picking a member of the category. Note that constant factors and cache behavior
Chris Lattner098129a2007-02-03 03:04:03 +0000724can be a big deal. If you have a vector that usually only contains a few
725elements (but could contain many), for example, it's much better to use
726<a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
727. Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
728cost of adding the elements to the container. </p>
729
730</div>
731
732<!-- ======================================================================= -->
733<div class="doc_subsection">
734 <a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
735</div>
736
737<div class="doc_text">
738There are a variety of sequential containers available for you, based on your
739needs. Pick the first in this section that will do what you want.
740</div>
741
742<!-- _______________________________________________________________________ -->
743<div class="doc_subsubsection">
744 <a name="dss_fixedarrays">Fixed Size Arrays</a>
745</div>
746
747<div class="doc_text">
748<p>Fixed size arrays are very simple and very fast. They are good if you know
749exactly how many elements you have, or you have a (low) upper bound on how many
750you have.</p>
751</div>
752
753<!-- _______________________________________________________________________ -->
754<div class="doc_subsubsection">
755 <a name="dss_heaparrays">Heap Allocated Arrays</a>
756</div>
757
758<div class="doc_text">
759<p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
760the number of elements is variable, if you know how many elements you will need
761before the array is allocated, and if the array is usually large (if not,
762consider a <a href="#dss_smallvector">SmallVector</a>). The cost of a heap
763allocated array is the cost of the new/delete (aka malloc/free). Also note that
764if you are allocating an array of a type with a constructor, the constructor and
Reid Spencer128a7a72007-02-03 21:06:43 +0000765destructors will be run for every element in the array (re-sizable vectors only
Chris Lattner098129a2007-02-03 03:04:03 +0000766construct those elements actually used).</p>
767</div>
768
769<!-- _______________________________________________________________________ -->
770<div class="doc_subsubsection">
771 <a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
772</div>
773
774<div class="doc_text">
775<p><tt>SmallVector&lt;Type, N&gt;</tt> is a simple class that looks and smells
776just like <tt>vector&lt;Type&gt;</tt>:
777it supports efficient iteration, lays out elements in memory order (so you can
778do pointer arithmetic between elements), supports efficient push_back/pop_back
779operations, supports efficient random access to its elements, etc.</p>
780
781<p>The advantage of SmallVector is that it allocates space for
782some number of elements (N) <b>in the object itself</b>. Because of this, if
783the SmallVector is dynamically smaller than N, no malloc is performed. This can
784be a big win in cases where the malloc/free call is far more expensive than the
785code that fiddles around with the elements.</p>
786
787<p>This is good for vectors that are "usually small" (e.g. the number of
788predecessors/successors of a block is usually less than 8). On the other hand,
789this makes the size of the SmallVector itself large, so you don't want to
790allocate lots of them (doing so will waste a lot of space). As such,
791SmallVectors are most useful when on the stack.</p>
792
793<p>SmallVector also provides a nice portable and efficient replacement for
794<tt>alloca</tt>.</p>
795
796</div>
797
798<!-- _______________________________________________________________________ -->
799<div class="doc_subsubsection">
800 <a name="dss_vector">&lt;vector&gt;</a>
801</div>
802
803<div class="doc_text">
804<p>
805std::vector is well loved and respected. It is useful when SmallVector isn't:
806when the size of the vector is often large (thus the small optimization will
807rarely be a benefit) or if you will be allocating many instances of the vector
808itself (which would waste space for elements that aren't in the container).
809vector is also useful when interfacing with code that expects vectors :).
810</p>
Chris Lattner32d84762007-02-05 06:30:51 +0000811
812<p>One worthwhile note about std::vector: avoid code like this:</p>
813
814<div class="doc_code">
815<pre>
816for ( ... ) {
Chris Lattner9bb3dbb2007-03-28 18:27:57 +0000817 std::vector&lt;foo&gt; V;
Chris Lattner32d84762007-02-05 06:30:51 +0000818 use V;
819}
820</pre>
821</div>
822
823<p>Instead, write this as:</p>
824
825<div class="doc_code">
826<pre>
Chris Lattner9bb3dbb2007-03-28 18:27:57 +0000827std::vector&lt;foo&gt; V;
Chris Lattner32d84762007-02-05 06:30:51 +0000828for ( ... ) {
829 use V;
830 V.clear();
831}
832</pre>
833</div>
834
835<p>Doing so will save (at least) one heap allocation and free per iteration of
836the loop.</p>
837
Chris Lattner098129a2007-02-03 03:04:03 +0000838</div>
839
840<!-- _______________________________________________________________________ -->
841<div class="doc_subsubsection">
Chris Lattner74c4ca12007-02-03 07:59:07 +0000842 <a name="dss_deque">&lt;deque&gt;</a>
843</div>
844
845<div class="doc_text">
846<p>std::deque is, in some senses, a generalized version of std::vector. Like
847std::vector, it provides constant time random access and other similar
848properties, but it also provides efficient access to the front of the list. It
849does not guarantee continuity of elements within memory.</p>
850
851<p>In exchange for this extra flexibility, std::deque has significantly higher
852constant factor costs than std::vector. If possible, use std::vector or
853something cheaper.</p>
854</div>
855
856<!-- _______________________________________________________________________ -->
857<div class="doc_subsubsection">
Chris Lattner098129a2007-02-03 03:04:03 +0000858 <a name="dss_list">&lt;list&gt;</a>
859</div>
860
861<div class="doc_text">
862<p>std::list is an extremely inefficient class that is rarely useful.
863It performs a heap allocation for every element inserted into it, thus having an
864extremely high constant factor, particularly for small data types. std::list
865also only supports bidirectional iteration, not random access iteration.</p>
866
867<p>In exchange for this high cost, std::list supports efficient access to both
868ends of the list (like std::deque, but unlike std::vector or SmallVector). In
869addition, the iterator invalidation characteristics of std::list are stronger
870than that of a vector class: inserting or removing an element into the list does
871not invalidate iterator or pointers to other elements in the list.</p>
872</div>
873
874<!-- _______________________________________________________________________ -->
875<div class="doc_subsubsection">
876 <a name="dss_ilist">llvm/ADT/ilist</a>
877</div>
878
879<div class="doc_text">
880<p><tt>ilist&lt;T&gt;</tt> implements an 'intrusive' doubly-linked list. It is
881intrusive, because it requires the element to store and provide access to the
882prev/next pointers for the list.</p>
883
884<p>ilist has the same drawbacks as std::list, and additionally requires an
885ilist_traits implementation for the element type, but it provides some novel
886characteristics. In particular, it can efficiently store polymorphic objects,
887the traits class is informed when an element is inserted or removed from the
888list, and ilists are guaranteed to support a constant-time splice operation.
889</p>
890
891<p>These properties are exactly what we want for things like Instructions and
892basic blocks, which is why these are implemented with ilists.</p>
893</div>
894
895<!-- _______________________________________________________________________ -->
896<div class="doc_subsubsection">
Chris Lattnerc5722432007-02-03 19:49:31 +0000897 <a name="dss_other">Other Sequential Container options</a>
Chris Lattner098129a2007-02-03 03:04:03 +0000898</div>
899
900<div class="doc_text">
Chris Lattner74c4ca12007-02-03 07:59:07 +0000901<p>Other STL containers are available, such as std::string.</p>
Chris Lattner098129a2007-02-03 03:04:03 +0000902
903<p>There are also various STL adapter classes such as std::queue,
904std::priority_queue, std::stack, etc. These provide simplified access to an
905underlying container but don't affect the cost of the container itself.</p>
906
907</div>
908
909
910<!-- ======================================================================= -->
911<div class="doc_subsection">
912 <a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
913</div>
914
915<div class="doc_text">
916
Chris Lattner74c4ca12007-02-03 07:59:07 +0000917<p>Set-like containers are useful when you need to canonicalize multiple values
918into a single representation. There are several different choices for how to do
919this, providing various trade-offs.</p>
920
921</div>
922
923
924<!-- _______________________________________________________________________ -->
925<div class="doc_subsubsection">
926 <a name="dss_sortedvectorset">A sorted 'vector'</a>
927</div>
928
929<div class="doc_text">
930
Chris Lattner3b23a8c2007-02-03 08:10:45 +0000931<p>If you intend to insert a lot of elements, then do a lot of queries, a
932great approach is to use a vector (or other sequential container) with
Chris Lattner74c4ca12007-02-03 07:59:07 +0000933std::sort+std::unique to remove duplicates. This approach works really well if
Chris Lattner3b23a8c2007-02-03 08:10:45 +0000934your usage pattern has these two distinct phases (insert then query), and can be
935coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
936</p>
937
938<p>
939This combination provides the several nice properties: the result data is
940contiguous in memory (good for cache locality), has few allocations, is easy to
941address (iterators in the final vector are just indices or pointers), and can be
942efficiently queried with a standard binary or radix search.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000943
944</div>
945
946<!-- _______________________________________________________________________ -->
947<div class="doc_subsubsection">
948 <a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
949</div>
950
951<div class="doc_text">
952
Reid Spencer128a7a72007-02-03 21:06:43 +0000953<p>If you have a set-like data structure that is usually small and whose elements
Chris Lattner4ddfac12007-02-03 07:59:51 +0000954are reasonably small, a <tt>SmallSet&lt;Type, N&gt;</tt> is a good choice. This set
Chris Lattner74c4ca12007-02-03 07:59:07 +0000955has space for N elements in place (thus, if the set is dynamically smaller than
Chris Lattner14868db2007-02-03 08:20:15 +0000956N, no malloc traffic is required) and accesses them with a simple linear search.
957When the set grows beyond 'N' elements, it allocates a more expensive representation that
Chris Lattner74c4ca12007-02-03 07:59:07 +0000958guarantees efficient access (for most types, it falls back to std::set, but for
Chris Lattner14868db2007-02-03 08:20:15 +0000959pointers it uses something far better, <a
Chris Lattner74c4ca12007-02-03 07:59:07 +0000960href="#dss_smallptrset">SmallPtrSet</a>).</p>
961
962<p>The magic of this class is that it handles small sets extremely efficiently,
963but gracefully handles extremely large sets without loss of efficiency. The
964drawback is that the interface is quite small: it supports insertion, queries
965and erasing, but does not support iteration.</p>
966
967</div>
968
969<!-- _______________________________________________________________________ -->
970<div class="doc_subsubsection">
971 <a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
972</div>
973
974<div class="doc_text">
975
976<p>SmallPtrSet has all the advantages of SmallSet (and a SmallSet of pointers is
Reid Spencer128a7a72007-02-03 21:06:43 +0000977transparently implemented with a SmallPtrSet), but also supports iterators. If
Chris Lattner14868db2007-02-03 08:20:15 +0000978more than 'N' insertions are performed, a single quadratically
Chris Lattner74c4ca12007-02-03 07:59:07 +0000979probed hash table is allocated and grows as needed, providing extremely
980efficient access (constant time insertion/deleting/queries with low constant
981factors) and is very stingy with malloc traffic.</p>
982
983<p>Note that, unlike std::set, the iterators of SmallPtrSet are invalidated
984whenever an insertion occurs. Also, the values visited by the iterators are not
985visited in sorted order.</p>
986
987</div>
988
989<!-- _______________________________________________________________________ -->
990<div class="doc_subsubsection">
991 <a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
992</div>
993
994<div class="doc_text">
995
Chris Lattner098129a2007-02-03 03:04:03 +0000996<p>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000997FoldingSet is an aggregate class that is really good at uniquing
998expensive-to-create or polymorphic objects. It is a combination of a chained
999hash table with intrusive links (uniqued objects are required to inherit from
Chris Lattner14868db2007-02-03 08:20:15 +00001000FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
1001its ID process.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +00001002
Chris Lattner14868db2007-02-03 08:20:15 +00001003<p>Consider a case where you want to implement a "getOrCreateFoo" method for
Chris Lattner74c4ca12007-02-03 07:59:07 +00001004a complex object (for example, a node in the code generator). The client has a
1005description of *what* it wants to generate (it knows the opcode and all the
1006operands), but we don't want to 'new' a node, then try inserting it into a set
Chris Lattner14868db2007-02-03 08:20:15 +00001007only to find out it already exists, at which point we would have to delete it
1008and return the node that already exists.
Chris Lattner098129a2007-02-03 03:04:03 +00001009</p>
1010
Chris Lattner74c4ca12007-02-03 07:59:07 +00001011<p>To support this style of client, FoldingSet perform a query with a
1012FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
1013element that we want to query for. The query either returns the element
1014matching the ID or it returns an opaque ID that indicates where insertion should
Chris Lattner14868db2007-02-03 08:20:15 +00001015take place. Construction of the ID usually does not require heap traffic.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +00001016
1017<p>Because FoldingSet uses intrusive links, it can support polymorphic objects
1018in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
1019Because the elements are individually allocated, pointers to the elements are
1020stable: inserting or removing elements does not invalidate any pointers to other
1021elements.
1022</p>
1023
1024</div>
1025
1026<!-- _______________________________________________________________________ -->
1027<div class="doc_subsubsection">
1028 <a name="dss_set">&lt;set&gt;</a>
1029</div>
1030
1031<div class="doc_text">
1032
Chris Lattnerc5722432007-02-03 19:49:31 +00001033<p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
1034many things but great at nothing. std::set allocates memory for each element
Chris Lattner74c4ca12007-02-03 07:59:07 +00001035inserted (thus it is very malloc intensive) and typically stores three pointers
Chris Lattner14868db2007-02-03 08:20:15 +00001036per element in the set (thus adding a large amount of per-element space
1037overhead). It offers guaranteed log(n) performance, which is not particularly
Chris Lattnerc5722432007-02-03 19:49:31 +00001038fast from a complexity standpoint (particularly if the elements of the set are
1039expensive to compare, like strings), and has extremely high constant factors for
1040lookup, insertion and removal.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +00001041
Chris Lattner14868db2007-02-03 08:20:15 +00001042<p>The advantages of std::set are that its iterators are stable (deleting or
Chris Lattner74c4ca12007-02-03 07:59:07 +00001043inserting an element from the set does not affect iterators or pointers to other
1044elements) and that iteration over the set is guaranteed to be in sorted order.
1045If the elements in the set are large, then the relative overhead of the pointers
1046and malloc traffic is not a big deal, but if the elements of the set are small,
1047std::set is almost never a good choice.</p>
1048
1049</div>
1050
1051<!-- _______________________________________________________________________ -->
1052<div class="doc_subsubsection">
1053 <a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
1054</div>
1055
1056<div class="doc_text">
Chris Lattneredca3c52007-02-04 00:00:26 +00001057<p>LLVM's SetVector&lt;Type&gt; is an adapter class that combines your choice of
1058a set-like container along with a <a href="#ds_sequential">Sequential
1059Container</a>. The important property
Chris Lattner74c4ca12007-02-03 07:59:07 +00001060that this provides is efficient insertion with uniquing (duplicate elements are
1061ignored) with iteration support. It implements this by inserting elements into
1062both a set-like container and the sequential container, using the set-like
1063container for uniquing and the sequential container for iteration.
1064</p>
1065
1066<p>The difference between SetVector and other sets is that the order of
1067iteration is guaranteed to match the order of insertion into the SetVector.
1068This property is really important for things like sets of pointers. Because
1069pointer values are non-deterministic (e.g. vary across runs of the program on
Chris Lattneredca3c52007-02-04 00:00:26 +00001070different machines), iterating over the pointers in the set will
Chris Lattner74c4ca12007-02-03 07:59:07 +00001071not be in a well-defined order.</p>
1072
1073<p>
1074The drawback of SetVector is that it requires twice as much space as a normal
1075set and has the sum of constant factors from the set-like container and the
1076sequential container that it uses. Use it *only* if you need to iterate over
1077the elements in a deterministic order. SetVector is also expensive to delete
Chris Lattneredca3c52007-02-04 00:00:26 +00001078elements out of (linear time), unless you use it's "pop_back" method, which is
1079faster.
Chris Lattner74c4ca12007-02-03 07:59:07 +00001080</p>
1081
Chris Lattneredca3c52007-02-04 00:00:26 +00001082<p>SetVector is an adapter class that defaults to using std::vector and std::set
1083for the underlying containers, so it is quite expensive. However,
1084<tt>"llvm/ADT/SetVector.h"</tt> also provides a SmallSetVector class, which
1085defaults to using a SmallVector and SmallSet of a specified size. If you use
1086this, and if your sets are dynamically smaller than N, you will save a lot of
1087heap traffic.</p>
1088
Chris Lattner74c4ca12007-02-03 07:59:07 +00001089</div>
1090
1091<!-- _______________________________________________________________________ -->
1092<div class="doc_subsubsection">
Chris Lattnerc5722432007-02-03 19:49:31 +00001093 <a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
1094</div>
1095
1096<div class="doc_text">
1097
1098<p>
1099UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
1100retains a unique ID for each element inserted into the set. It internally
1101contains a map and a vector, and it assigns a unique ID for each value inserted
1102into the set.</p>
1103
1104<p>UniqueVector is very expensive: its cost is the sum of the cost of
1105maintaining both the map and vector, it has high complexity, high constant
1106factors, and produces a lot of malloc traffic. It should be avoided.</p>
1107
1108</div>
1109
1110
1111<!-- _______________________________________________________________________ -->
1112<div class="doc_subsubsection">
1113 <a name="dss_otherset">Other Set-Like Container Options</a>
Chris Lattner74c4ca12007-02-03 07:59:07 +00001114</div>
1115
1116<div class="doc_text">
1117
1118<p>
1119The STL provides several other options, such as std::multiset and the various
Chris Lattnerc5722432007-02-03 19:49:31 +00001120"hash_set" like containers (whether from C++ TR1 or from the SGI library).</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +00001121
1122<p>std::multiset is useful if you're not interested in elimination of
Chris Lattner14868db2007-02-03 08:20:15 +00001123duplicates, but has all the drawbacks of std::set. A sorted vector (where you
1124don't delete duplicate entries) or some other approach is almost always
1125better.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +00001126
1127<p>The various hash_set implementations (exposed portably by
Chris Lattner14868db2007-02-03 08:20:15 +00001128"llvm/ADT/hash_set") is a simple chained hashtable. This algorithm is as malloc
1129intensive as std::set (performing an allocation for each element inserted,
Chris Lattner74c4ca12007-02-03 07:59:07 +00001130thus having really high constant factors) but (usually) provides O(1)
1131insertion/deletion of elements. This can be useful if your elements are large
Chris Lattner14868db2007-02-03 08:20:15 +00001132(thus making the constant-factor cost relatively low) or if comparisons are
1133expensive. Element iteration does not visit elements in a useful order.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +00001134
Chris Lattner098129a2007-02-03 03:04:03 +00001135</div>
1136
1137<!-- ======================================================================= -->
1138<div class="doc_subsection">
1139 <a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
1140</div>
1141
1142<div class="doc_text">
Chris Lattnerc5722432007-02-03 19:49:31 +00001143Map-like containers are useful when you want to associate data to a key. As
1144usual, there are a lot of different ways to do this. :)
1145</div>
1146
1147<!-- _______________________________________________________________________ -->
1148<div class="doc_subsubsection">
1149 <a name="dss_sortedvectormap">A sorted 'vector'</a>
1150</div>
1151
1152<div class="doc_text">
1153
1154<p>
1155If your usage pattern follows a strict insert-then-query approach, you can
1156trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
1157for set-like containers</a>. The only difference is that your query function
1158(which uses std::lower_bound to get efficient log(n) lookup) should only compare
1159the key, not both the key and value. This yields the same advantages as sorted
1160vectors for sets.
1161</p>
1162</div>
1163
1164<!-- _______________________________________________________________________ -->
1165<div class="doc_subsubsection">
Chris Lattner796f9fa2007-02-08 19:14:21 +00001166 <a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
Chris Lattnerc5722432007-02-03 19:49:31 +00001167</div>
1168
1169<div class="doc_text">
1170
1171<p>
1172Strings are commonly used as keys in maps, and they are difficult to support
1173efficiently: they are variable length, inefficient to hash and compare when
Chris Lattner796f9fa2007-02-08 19:14:21 +00001174long, expensive to copy, etc. StringMap is a specialized container designed to
1175cope with these issues. It supports mapping an arbitrary range of bytes to an
1176arbitrary other object.</p>
Chris Lattnerc5722432007-02-03 19:49:31 +00001177
Chris Lattner796f9fa2007-02-08 19:14:21 +00001178<p>The StringMap implementation uses a quadratically-probed hash table, where
Chris Lattnerc5722432007-02-03 19:49:31 +00001179the buckets store a pointer to the heap allocated entries (and some other
1180stuff). The entries in the map must be heap allocated because the strings are
1181variable length. The string data (key) and the element object (value) are
1182stored in the same allocation with the string data immediately after the element
1183object. This container guarantees the "<tt>(char*)(&amp;Value+1)</tt>" points
1184to the key string for a value.</p>
1185
Chris Lattner796f9fa2007-02-08 19:14:21 +00001186<p>The StringMap is very fast for several reasons: quadratic probing is very
Chris Lattnerc5722432007-02-03 19:49:31 +00001187cache efficient for lookups, the hash value of strings in buckets is not
Chris Lattner796f9fa2007-02-08 19:14:21 +00001188recomputed when lookup up an element, StringMap rarely has to touch the
Chris Lattnerc5722432007-02-03 19:49:31 +00001189memory for unrelated objects when looking up a value (even when hash collisions
1190happen), hash table growth does not recompute the hash values for strings
1191already in the table, and each pair in the map is store in a single allocation
1192(the string data is stored in the same allocation as the Value of a pair).</p>
1193
Chris Lattner796f9fa2007-02-08 19:14:21 +00001194<p>StringMap also provides query methods that take byte ranges, so it only ever
Chris Lattnerc5722432007-02-03 19:49:31 +00001195copies a string if a value is inserted into the table.</p>
1196</div>
1197
1198<!-- _______________________________________________________________________ -->
1199<div class="doc_subsubsection">
1200 <a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
1201</div>
1202
1203<div class="doc_text">
1204<p>
1205IndexedMap is a specialized container for mapping small dense integers (or
1206values that can be mapped to small dense integers) to some other type. It is
1207internally implemented as a vector with a mapping function that maps the keys to
1208the dense integer range.
1209</p>
1210
1211<p>
1212This is useful for cases like virtual registers in the LLVM code generator: they
1213have a dense mapping that is offset by a compile-time constant (the first
1214virtual register ID).</p>
1215
1216</div>
1217
1218<!-- _______________________________________________________________________ -->
1219<div class="doc_subsubsection">
1220 <a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
1221</div>
1222
1223<div class="doc_text">
1224
1225<p>
1226DenseMap is a simple quadratically probed hash table. It excels at supporting
1227small keys and values: it uses a single allocation to hold all of the pairs that
1228are currently inserted in the map. DenseMap is a great way to map pointers to
1229pointers, or map other small types to each other.
1230</p>
1231
1232<p>
1233There are several aspects of DenseMap that you should be aware of, however. The
1234iterators in a densemap are invalidated whenever an insertion occurs, unlike
1235map. Also, because DenseMap allocates space for a large number of key/value
Chris Lattnera4a264d2007-02-03 20:17:53 +00001236pairs (it starts with 64 by default), it will waste a lot of space if your keys
1237or values are large. Finally, you must implement a partial specialization of
Chris Lattner76c1b972007-09-17 18:34:04 +00001238DenseMapInfo for the key that you want, if it isn't already supported. This
Chris Lattnerc5722432007-02-03 19:49:31 +00001239is required to tell DenseMap about two special marker values (which can never be
Chris Lattnera4a264d2007-02-03 20:17:53 +00001240inserted into the map) that it needs internally.</p>
Chris Lattnerc5722432007-02-03 19:49:31 +00001241
1242</div>
1243
1244<!-- _______________________________________________________________________ -->
1245<div class="doc_subsubsection">
1246 <a name="dss_map">&lt;map&gt;</a>
1247</div>
1248
1249<div class="doc_text">
1250
1251<p>
1252std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
1253a single allocation per pair inserted into the map, it offers log(n) lookup with
1254an extremely large constant factor, imposes a space penalty of 3 pointers per
1255pair in the map, etc.</p>
1256
1257<p>std::map is most useful when your keys or values are very large, if you need
1258to iterate over the collection in sorted order, or if you need stable iterators
1259into the map (i.e. they don't get invalidated if an insertion or deletion of
1260another element takes place).</p>
1261
1262</div>
1263
1264<!-- _______________________________________________________________________ -->
1265<div class="doc_subsubsection">
1266 <a name="dss_othermap">Other Map-Like Container Options</a>
1267</div>
1268
1269<div class="doc_text">
1270
1271<p>
1272The STL provides several other options, such as std::multimap and the various
1273"hash_map" like containers (whether from C++ TR1 or from the SGI library).</p>
1274
1275<p>std::multimap is useful if you want to map a key to multiple values, but has
1276all the drawbacks of std::map. A sorted vector or some other approach is almost
1277always better.</p>
1278
1279<p>The various hash_map implementations (exposed portably by
1280"llvm/ADT/hash_map") are simple chained hash tables. This algorithm is as
1281malloc intensive as std::map (performing an allocation for each element
1282inserted, thus having really high constant factors) but (usually) provides O(1)
1283insertion/deletion of elements. This can be useful if your elements are large
1284(thus making the constant-factor cost relatively low) or if comparisons are
1285expensive. Element iteration does not visit elements in a useful order.</p>
1286
Chris Lattner098129a2007-02-03 03:04:03 +00001287</div>
1288
Daniel Berlin1939ace2007-09-24 17:52:25 +00001289<!-- ======================================================================= -->
1290<div class="doc_subsection">
1291 <a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
1292</div>
1293
1294<div class="doc_text">
1295Unlike the other containers, there are only two bit storage containers, and when
1296to use each is relatively straightforward.
1297</div>
1298
1299<!-- _______________________________________________________________________ -->
1300<div class="doc_subsubsection">
1301 <a name="dss_bitvector">BitVector</a>
1302</div>
1303
1304<div class="doc_text">
1305<p> The BitVector container provides a fixed size set of bits for manipulation.
1306It supports individual bit setting/testing, as well as set operations. The set
1307operations take time O(size of bitvector), but operations are performed one word
1308at a time, instead of one bit at a time. This makes the BitVector very fast for
1309set operations compared to other containers. Use the BitVector when you expect
1310the number of set bits to be high (IE a dense set).
1311</p>
1312</div>
1313
1314<!-- _______________________________________________________________________ -->
1315<div class="doc_subsubsection">
1316 <a name="dss_sparsebitvector">SparseBitVector</a>
1317</div>
1318
1319<div class="doc_text">
1320<p> The SparseBitVector container is much like BitVector, with one major
1321difference: Only the bits that are set, are stored. This makes the
1322SparseBitVector much more space efficient than BitVector when the set is sparse,
1323as well as making set operations O(number of set bits) instead of O(size of
1324universe). The downside to the SparseBitVector is that setting and testing of random bits is O(N), and on large SparseBitVectors, this can be slower than BitVector. In our implementation, setting or testing bits in sorted order
1325(either forwards or reverse) is O(1) worst case. Testing and setting bits within 128 bits (depends on size) of the current bit is also O(1). As a general statement, testing/setting bits in a SparseBitVector is O(distance away from last set bit).
1326</p>
1327</div>
Chris Lattnerf623a082005-10-17 01:36:23 +00001328
Misha Brukman13fd15c2004-01-15 00:14:41 +00001329<!-- *********************************************************************** -->
1330<div class="doc_section">
1331 <a name="common">Helpful Hints for Common Operations</a>
1332</div>
1333<!-- *********************************************************************** -->
1334
1335<div class="doc_text">
1336
1337<p>This section describes how to perform some very simple transformations of
1338LLVM code. This is meant to give examples of common idioms used, showing the
1339practical side of LLVM transformations. <p> Because this is a "how-to" section,
1340you should also read about the main classes that you will be working with. The
1341<a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
1342and descriptions of the main classes that you should know about.</p>
1343
1344</div>
1345
1346<!-- NOTE: this section should be heavy on example code -->
1347<!-- ======================================================================= -->
1348<div class="doc_subsection">
1349 <a name="inspection">Basic Inspection and Traversal Routines</a>
1350</div>
1351
1352<div class="doc_text">
1353
1354<p>The LLVM compiler infrastructure have many different data structures that may
1355be traversed. Following the example of the C++ standard template library, the
1356techniques used to traverse these various data structures are all basically the
1357same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
1358method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
1359function returns an iterator pointing to one past the last valid element of the
1360sequence, and there is some <tt>XXXiterator</tt> data type that is common
1361between the two operations.</p>
1362
1363<p>Because the pattern for iteration is common across many different aspects of
1364the program representation, the standard template library algorithms may be used
1365on them, and it is easier to remember how to iterate. First we show a few common
1366examples of the data structures that need to be traversed. Other data
1367structures are traversed in very similar ways.</p>
1368
1369</div>
1370
1371<!-- _______________________________________________________________________ -->
Chris Lattner69bf8a92004-05-23 21:06:58 +00001372<div class="doc_subsubsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +00001373 <a name="iterate_function">Iterating over the </a><a
1374 href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
1375 href="#Function"><tt>Function</tt></a>
1376</div>
1377
1378<div class="doc_text">
1379
1380<p>It's quite common to have a <tt>Function</tt> instance that you'd like to
1381transform in some way; in particular, you'd like to manipulate its
1382<tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
1383the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
1384an example that prints the name of a <tt>BasicBlock</tt> and the number of
1385<tt>Instruction</tt>s it contains:</p>
1386
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001387<div class="doc_code">
1388<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +00001389// <i>func is a pointer to a Function instance</i>
1390for (Function::iterator i = func-&gt;begin(), e = func-&gt;end(); i != e; ++i)
1391 // <i>Print out the name of the basic block if it has one, and then the</i>
1392 // <i>number of instructions that it contains</i>
Bill Wendling832171c2006-12-07 20:04:42 +00001393 llvm::cerr &lt;&lt; "Basic block (name=" &lt;&lt; i-&gt;getName() &lt;&lt; ") has "
1394 &lt;&lt; i-&gt;size() &lt;&lt; " instructions.\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001395</pre>
1396</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001397
1398<p>Note that i can be used as if it were a pointer for the purposes of
Joel Stanley9b96c442002-09-06 21:55:13 +00001399invoking member functions of the <tt>Instruction</tt> class. This is
1400because the indirection operator is overloaded for the iterator
Chris Lattner7496ec52003-08-05 22:54:23 +00001401classes. In the above code, the expression <tt>i-&gt;size()</tt> is
Misha Brukman13fd15c2004-01-15 00:14:41 +00001402exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
1403
1404</div>
1405
1406<!-- _______________________________________________________________________ -->
Chris Lattner69bf8a92004-05-23 21:06:58 +00001407<div class="doc_subsubsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +00001408 <a name="iterate_basicblock">Iterating over the </a><a
1409 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1410 href="#BasicBlock"><tt>BasicBlock</tt></a>
1411</div>
1412
1413<div class="doc_text">
1414
1415<p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
1416easy to iterate over the individual instructions that make up
1417<tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
1418a <tt>BasicBlock</tt>:</p>
1419
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001420<div class="doc_code">
Chris Lattner55c04612005-03-06 06:00:13 +00001421<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +00001422// <i>blk is a pointer to a BasicBlock instance</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001423for (BasicBlock::iterator i = blk-&gt;begin(), e = blk-&gt;end(); i != e; ++i)
Bill Wendling82e2eea2006-10-11 18:00:22 +00001424 // <i>The next statement works since operator&lt;&lt;(ostream&amp;,...)</i>
1425 // <i>is overloaded for Instruction&amp;</i>
Bill Wendling832171c2006-12-07 20:04:42 +00001426 llvm::cerr &lt;&lt; *i &lt;&lt; "\n";
Chris Lattner55c04612005-03-06 06:00:13 +00001427</pre>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001428</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001429
1430<p>However, this isn't really the best way to print out the contents of a
1431<tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
1432anything you'll care about, you could have just invoked the print routine on the
Bill Wendling832171c2006-12-07 20:04:42 +00001433basic block itself: <tt>llvm::cerr &lt;&lt; *blk &lt;&lt; "\n";</tt>.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001434
1435</div>
1436
1437<!-- _______________________________________________________________________ -->
Chris Lattner69bf8a92004-05-23 21:06:58 +00001438<div class="doc_subsubsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +00001439 <a name="iterate_institer">Iterating over the </a><a
1440 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1441 href="#Function"><tt>Function</tt></a>
1442</div>
1443
1444<div class="doc_text">
1445
1446<p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
1447<tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
1448<tt>InstIterator</tt> should be used instead. You'll need to include <a
1449href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
1450and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
Chris Lattner69bf8a92004-05-23 21:06:58 +00001451small example that shows how to dump all instructions in a function to the standard error stream:<p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001452
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001453<div class="doc_code">
1454<pre>
1455#include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
1456
Reid Spencer128a7a72007-02-03 21:06:43 +00001457// <i>F is a pointer to a Function instance</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001458for (inst_iterator i = inst_begin(F), e = inst_end(F); i != e; ++i)
Bill Wendling832171c2006-12-07 20:04:42 +00001459 llvm::cerr &lt;&lt; *i &lt;&lt; "\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001460</pre>
1461</div>
1462
1463<p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
Reid Spencer128a7a72007-02-03 21:06:43 +00001464work list with its initial contents. For example, if you wanted to
1465initialize a work list to contain all instructions in a <tt>Function</tt>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001466F, all you would need to do is something like:</p>
1467
1468<div class="doc_code">
1469<pre>
1470std::set&lt;Instruction*&gt; worklist;
1471worklist.insert(inst_begin(F), inst_end(F));
1472</pre>
1473</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001474
1475<p>The STL set <tt>worklist</tt> would now contain all instructions in the
1476<tt>Function</tt> pointed to by F.</p>
1477
1478</div>
1479
1480<!-- _______________________________________________________________________ -->
1481<div class="doc_subsubsection">
1482 <a name="iterate_convert">Turning an iterator into a class pointer (and
1483 vice-versa)</a>
1484</div>
1485
1486<div class="doc_text">
1487
1488<p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
Joel Stanley9b96c442002-09-06 21:55:13 +00001489instance when all you've got at hand is an iterator. Well, extracting
Chris Lattner69bf8a92004-05-23 21:06:58 +00001490a reference or a pointer from an iterator is very straight-forward.
Chris Lattner261efe92003-11-25 01:02:51 +00001491Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001492is a <tt>BasicBlock::const_iterator</tt>:</p>
1493
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001494<div class="doc_code">
1495<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +00001496Instruction&amp; inst = *i; // <i>Grab reference to instruction reference</i>
1497Instruction* pinst = &amp;*i; // <i>Grab pointer to instruction reference</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001498const Instruction&amp; inst = *j;
1499</pre>
1500</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001501
1502<p>However, the iterators you'll be working with in the LLVM framework are
1503special: they will automatically convert to a ptr-to-instance type whenever they
1504need to. Instead of dereferencing the iterator and then taking the address of
1505the result, you can simply assign the iterator to the proper pointer type and
1506you get the dereference and address-of operation as a result of the assignment
1507(behind the scenes, this is a result of overloading casting mechanisms). Thus
1508the last line of the last example,</p>
1509
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001510<div class="doc_code">
1511<pre>
1512Instruction* pinst = &amp;*i;
1513</pre>
1514</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001515
1516<p>is semantically equivalent to</p>
1517
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001518<div class="doc_code">
1519<pre>
1520Instruction* pinst = i;
1521</pre>
1522</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001523
Chris Lattner69bf8a92004-05-23 21:06:58 +00001524<p>It's also possible to turn a class pointer into the corresponding iterator,
1525and this is a constant time operation (very efficient). The following code
1526snippet illustrates use of the conversion constructors provided by LLVM
1527iterators. By using these, you can explicitly grab the iterator of something
1528without actually obtaining it via iteration over some structure:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001529
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001530<div class="doc_code">
1531<pre>
1532void printNextInstruction(Instruction* inst) {
1533 BasicBlock::iterator it(inst);
Bill Wendling82e2eea2006-10-11 18:00:22 +00001534 ++it; // <i>After this line, it refers to the instruction after *inst</i>
Bill Wendling832171c2006-12-07 20:04:42 +00001535 if (it != inst-&gt;getParent()-&gt;end()) llvm::cerr &lt;&lt; *it &lt;&lt; "\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001536}
1537</pre>
1538</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001539
Misha Brukman13fd15c2004-01-15 00:14:41 +00001540</div>
1541
1542<!--_______________________________________________________________________-->
1543<div class="doc_subsubsection">
1544 <a name="iterate_complex">Finding call sites: a slightly more complex
1545 example</a>
1546</div>
1547
1548<div class="doc_text">
1549
1550<p>Say that you're writing a FunctionPass and would like to count all the
1551locations in the entire module (that is, across every <tt>Function</tt>) where a
1552certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
1553learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
Chris Lattner69bf8a92004-05-23 21:06:58 +00001554much more straight-forward manner, but this example will allow us to explore how
Reid Spencer128a7a72007-02-03 21:06:43 +00001555you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
Misha Brukman13fd15c2004-01-15 00:14:41 +00001556is what we want to do:</p>
1557
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001558<div class="doc_code">
1559<pre>
1560initialize callCounter to zero
1561for each Function f in the Module
1562 for each BasicBlock b in f
1563 for each Instruction i in b
1564 if (i is a CallInst and calls the given function)
1565 increment callCounter
1566</pre>
1567</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001568
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001569<p>And the actual code is (remember, because we're writing a
Misha Brukman13fd15c2004-01-15 00:14:41 +00001570<tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001571override the <tt>runOnFunction</tt> method):</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001572
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001573<div class="doc_code">
1574<pre>
1575Function* targetFunc = ...;
1576
1577class OurFunctionPass : public FunctionPass {
1578 public:
1579 OurFunctionPass(): callCounter(0) { }
1580
1581 virtual runOnFunction(Function&amp; F) {
1582 for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
1583 for (BasicBlock::iterator i = b-&gt;begin(); ie = b-&gt;end(); i != ie; ++i) {
1584 if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a>&lt;<a
1585 href="#CallInst">CallInst</a>&gt;(&amp;*i)) {
Bill Wendling82e2eea2006-10-11 18:00:22 +00001586 // <i>We know we've encountered a call instruction, so we</i>
1587 // <i>need to determine if it's a call to the</i>
1588 // <i>function pointed to by m_func or not</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001589
1590 if (callInst-&gt;getCalledFunction() == targetFunc)
1591 ++callCounter;
1592 }
1593 }
1594 }
Bill Wendling82e2eea2006-10-11 18:00:22 +00001595 }
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001596
1597 private:
1598 unsigned callCounter;
1599};
1600</pre>
1601</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001602
1603</div>
1604
Brian Gaekef1972c62003-11-07 19:25:45 +00001605<!--_______________________________________________________________________-->
Misha Brukman13fd15c2004-01-15 00:14:41 +00001606<div class="doc_subsubsection">
1607 <a name="calls_and_invokes">Treating calls and invokes the same way</a>
1608</div>
1609
1610<div class="doc_text">
1611
1612<p>You may have noticed that the previous example was a bit oversimplified in
1613that it did not deal with call sites generated by 'invoke' instructions. In
1614this, and in other situations, you may find that you want to treat
1615<tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
1616most-specific common base class is <tt>Instruction</tt>, which includes lots of
1617less closely-related things. For these cases, LLVM provides a handy wrapper
1618class called <a
Reid Spencer05fe4b02006-03-14 05:39:39 +00001619href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
Chris Lattner69bf8a92004-05-23 21:06:58 +00001620It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
1621methods that provide functionality common to <tt>CallInst</tt>s and
Misha Brukman13fd15c2004-01-15 00:14:41 +00001622<tt>InvokeInst</tt>s.</p>
1623
Chris Lattner69bf8a92004-05-23 21:06:58 +00001624<p>This class has "value semantics": it should be passed by value, not by
1625reference and it should not be dynamically allocated or deallocated using
1626<tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
1627assignable and constructable, with costs equivalents to that of a bare pointer.
1628If you look at its definition, it has only a single pointer member.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001629
1630</div>
1631
Chris Lattner1a3105b2002-09-09 05:49:39 +00001632<!--_______________________________________________________________________-->
Misha Brukman13fd15c2004-01-15 00:14:41 +00001633<div class="doc_subsubsection">
1634 <a name="iterate_chains">Iterating over def-use &amp; use-def chains</a>
1635</div>
1636
1637<div class="doc_text">
1638
1639<p>Frequently, we might have an instance of the <a
Chris Lattner00815172007-01-04 22:01:45 +00001640href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
Misha Brukman384047f2004-06-03 23:29:12 +00001641determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
1642<tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
1643For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
1644particular function <tt>foo</tt>. Finding all of the instructions that
1645<i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
1646of <tt>F</tt>:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001647
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001648<div class="doc_code">
1649<pre>
1650Function* F = ...;
1651
Bill Wendling82e2eea2006-10-11 18:00:22 +00001652for (Value::use_iterator i = F-&gt;use_begin(), e = F-&gt;use_end(); i != e; ++i)
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001653 if (Instruction *Inst = dyn_cast&lt;Instruction&gt;(*i)) {
Bill Wendling832171c2006-12-07 20:04:42 +00001654 llvm::cerr &lt;&lt; "F is used in instruction:\n";
1655 llvm::cerr &lt;&lt; *Inst &lt;&lt; "\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001656 }
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001657</pre>
1658</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001659
1660<p>Alternately, it's common to have an instance of the <a
Misha Brukman384047f2004-06-03 23:29:12 +00001661href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
Misha Brukman13fd15c2004-01-15 00:14:41 +00001662<tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
1663<tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
1664<tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
1665all of the values that a particular instruction uses (that is, the operands of
1666the particular <tt>Instruction</tt>):</p>
1667
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001668<div class="doc_code">
1669<pre>
1670Instruction* pi = ...;
1671
1672for (User::op_iterator i = pi-&gt;op_begin(), e = pi-&gt;op_end(); i != e; ++i) {
1673 Value* v = *i;
Bill Wendling82e2eea2006-10-11 18:00:22 +00001674 // <i>...</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001675}
1676</pre>
1677</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001678
Chris Lattner1a3105b2002-09-09 05:49:39 +00001679<!--
1680 def-use chains ("finding all users of"): Value::use_begin/use_end
1681 use-def chains ("finding all values used"): User::op_begin/op_end [op=operand]
Misha Brukman13fd15c2004-01-15 00:14:41 +00001682-->
1683
1684</div>
1685
1686<!-- ======================================================================= -->
1687<div class="doc_subsection">
1688 <a name="simplechanges">Making simple changes</a>
1689</div>
1690
1691<div class="doc_text">
1692
1693<p>There are some primitive transformation operations present in the LLVM
Joel Stanley753eb712002-09-11 22:32:24 +00001694infrastructure that are worth knowing about. When performing
Chris Lattner261efe92003-11-25 01:02:51 +00001695transformations, it's fairly common to manipulate the contents of basic
1696blocks. This section describes some of the common methods for doing so
Misha Brukman13fd15c2004-01-15 00:14:41 +00001697and gives example code.</p>
1698
1699</div>
1700
Chris Lattner261efe92003-11-25 01:02:51 +00001701<!--_______________________________________________________________________-->
Misha Brukman13fd15c2004-01-15 00:14:41 +00001702<div class="doc_subsubsection">
1703 <a name="schanges_creating">Creating and inserting new
1704 <tt>Instruction</tt>s</a>
1705</div>
1706
1707<div class="doc_text">
1708
1709<p><i>Instantiating Instructions</i></p>
1710
Chris Lattner69bf8a92004-05-23 21:06:58 +00001711<p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
Misha Brukman13fd15c2004-01-15 00:14:41 +00001712constructor for the kind of instruction to instantiate and provide the necessary
1713parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
1714(const-ptr-to) <tt>Type</tt>. Thus:</p>
1715
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001716<div class="doc_code">
1717<pre>
1718AllocaInst* ai = new AllocaInst(Type::IntTy);
1719</pre>
1720</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001721
1722<p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
Reid Spencer128a7a72007-02-03 21:06:43 +00001723one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001724subclass is likely to have varying default parameters which change the semantics
1725of the instruction, so refer to the <a
Misha Brukman31ca1de2004-06-03 23:35:54 +00001726href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
Misha Brukman13fd15c2004-01-15 00:14:41 +00001727Instruction</a> that you're interested in instantiating.</p>
1728
1729<p><i>Naming values</i></p>
1730
1731<p>It is very useful to name the values of instructions when you're able to, as
1732this facilitates the debugging of your transformations. If you end up looking
1733at generated LLVM machine code, you definitely want to have logical names
1734associated with the results of instructions! By supplying a value for the
1735<tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
1736associate a logical name with the result of the instruction's execution at
Reid Spencer128a7a72007-02-03 21:06:43 +00001737run time. For example, say that I'm writing a transformation that dynamically
Misha Brukman13fd15c2004-01-15 00:14:41 +00001738allocates space for an integer on the stack, and that integer is going to be
1739used as some kind of index by some other code. To accomplish this, I place an
1740<tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
1741<tt>Function</tt>, and I'm intending to use it within the same
1742<tt>Function</tt>. I might do:</p>
1743
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001744<div class="doc_code">
1745<pre>
1746AllocaInst* pa = new AllocaInst(Type::IntTy, 0, "indexLoc");
1747</pre>
1748</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001749
1750<p>where <tt>indexLoc</tt> is now the logical name of the instruction's
Reid Spencer128a7a72007-02-03 21:06:43 +00001751execution value, which is a pointer to an integer on the run time stack.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001752
1753<p><i>Inserting instructions</i></p>
1754
1755<p>There are essentially two ways to insert an <tt>Instruction</tt>
1756into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
1757
Joel Stanley9dd1ad62002-09-18 03:17:23 +00001758<ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001759 <li>Insertion into an explicit instruction list
1760
1761 <p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
1762 <tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
1763 before <tt>*pi</tt>, we do the following: </p>
1764
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001765<div class="doc_code">
1766<pre>
1767BasicBlock *pb = ...;
1768Instruction *pi = ...;
1769Instruction *newInst = new Instruction(...);
1770
Bill Wendling82e2eea2006-10-11 18:00:22 +00001771pb-&gt;getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001772</pre>
1773</div>
Alkis Evlogimenos9a5dc4f2004-05-27 00:57:51 +00001774
1775 <p>Appending to the end of a <tt>BasicBlock</tt> is so common that
1776 the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
1777 classes provide constructors which take a pointer to a
1778 <tt>BasicBlock</tt> to be appended to. For example code that
1779 looked like: </p>
1780
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001781<div class="doc_code">
1782<pre>
1783BasicBlock *pb = ...;
1784Instruction *newInst = new Instruction(...);
1785
Bill Wendling82e2eea2006-10-11 18:00:22 +00001786pb-&gt;getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001787</pre>
1788</div>
Alkis Evlogimenos9a5dc4f2004-05-27 00:57:51 +00001789
1790 <p>becomes: </p>
1791
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001792<div class="doc_code">
1793<pre>
1794BasicBlock *pb = ...;
1795Instruction *newInst = new Instruction(..., pb);
1796</pre>
1797</div>
Alkis Evlogimenos9a5dc4f2004-05-27 00:57:51 +00001798
1799 <p>which is much cleaner, especially if you are creating
1800 long instruction streams.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001801
1802 <li>Insertion into an implicit instruction list
1803
1804 <p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
1805 are implicitly associated with an existing instruction list: the instruction
1806 list of the enclosing basic block. Thus, we could have accomplished the same
1807 thing as the above code without being given a <tt>BasicBlock</tt> by doing:
1808 </p>
1809
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001810<div class="doc_code">
1811<pre>
1812Instruction *pi = ...;
1813Instruction *newInst = new Instruction(...);
1814
1815pi-&gt;getParent()-&gt;getInstList().insert(pi, newInst);
1816</pre>
1817</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001818
1819 <p>In fact, this sequence of steps occurs so frequently that the
1820 <tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
1821 constructors which take (as a default parameter) a pointer to an
1822 <tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
1823 precede. That is, <tt>Instruction</tt> constructors are capable of
1824 inserting the newly-created instance into the <tt>BasicBlock</tt> of a
1825 provided instruction, immediately before that instruction. Using an
1826 <tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
1827 parameter, the above code becomes:</p>
1828
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001829<div class="doc_code">
1830<pre>
1831Instruction* pi = ...;
1832Instruction* newInst = new Instruction(..., pi);
1833</pre>
1834</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001835
1836 <p>which is much cleaner, especially if you're creating a lot of
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001837 instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001838</ul>
1839
1840</div>
1841
1842<!--_______________________________________________________________________-->
1843<div class="doc_subsubsection">
1844 <a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
1845</div>
1846
1847<div class="doc_text">
1848
1849<p>Deleting an instruction from an existing sequence of instructions that form a
Chris Lattner69bf8a92004-05-23 21:06:58 +00001850<a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward. First,
Misha Brukman13fd15c2004-01-15 00:14:41 +00001851you must have a pointer to the instruction that you wish to delete. Second, you
1852need to obtain the pointer to that instruction's basic block. You use the
1853pointer to the basic block to get its list of instructions and then use the
1854erase function to remove your instruction. For example:</p>
1855
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001856<div class="doc_code">
1857<pre>
1858<a href="#Instruction">Instruction</a> *I = .. ;
1859<a href="#BasicBlock">BasicBlock</a> *BB = I-&gt;getParent();
1860
1861BB-&gt;getInstList().erase(I);
1862</pre>
1863</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001864
1865</div>
1866
1867<!--_______________________________________________________________________-->
1868<div class="doc_subsubsection">
1869 <a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
1870 <tt>Value</tt></a>
1871</div>
1872
1873<div class="doc_text">
1874
1875<p><i>Replacing individual instructions</i></p>
1876
1877<p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
Chris Lattner261efe92003-11-25 01:02:51 +00001878permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001879and <tt>ReplaceInstWithInst</tt>.</p>
1880
Chris Lattner261efe92003-11-25 01:02:51 +00001881<h4><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h4>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001882
Chris Lattner261efe92003-11-25 01:02:51 +00001883<ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001884 <li><tt>ReplaceInstWithValue</tt>
1885
1886 <p>This function replaces all uses (within a basic block) of a given
1887 instruction with a value, and then removes the original instruction. The
1888 following example illustrates the replacement of the result of a particular
Chris Lattner58360822005-01-17 00:12:04 +00001889 <tt>AllocaInst</tt> that allocates memory for a single integer with a null
Misha Brukman13fd15c2004-01-15 00:14:41 +00001890 pointer to an integer.</p>
1891
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001892<div class="doc_code">
1893<pre>
1894AllocaInst* instToReplace = ...;
1895BasicBlock::iterator ii(instToReplace);
1896
1897ReplaceInstWithValue(instToReplace-&gt;getParent()-&gt;getInstList(), ii,
1898 Constant::getNullValue(PointerType::get(Type::IntTy)));
1899</pre></div></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001900
1901 <li><tt>ReplaceInstWithInst</tt>
1902
1903 <p>This function replaces a particular instruction with another
1904 instruction. The following example illustrates the replacement of one
1905 <tt>AllocaInst</tt> with another.</p>
1906
Bill Wendling3cd5ca62006-10-11 06:30:10 +00001907<div class="doc_code">
1908<pre>
1909AllocaInst* instToReplace = ...;
1910BasicBlock::iterator ii(instToReplace);
1911
1912ReplaceInstWithInst(instToReplace-&gt;getParent()-&gt;getInstList(), ii,
1913 new AllocaInst(Type::IntTy, 0, "ptrToReplacedInt"));
1914</pre></div></li>
Chris Lattner261efe92003-11-25 01:02:51 +00001915</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00001916
1917<p><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></p>
1918
1919<p>You can use <tt>Value::replaceAllUsesWith</tt> and
1920<tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
Chris Lattner00815172007-01-04 22:01:45 +00001921doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
Misha Brukman384047f2004-06-03 23:29:12 +00001922and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
Misha Brukman13fd15c2004-01-15 00:14:41 +00001923information.</p>
1924
1925<!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
1926include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
1927ReplaceInstWithValue, ReplaceInstWithInst -->
1928
1929</div>
1930
Tanya Lattnerb011c662007-06-20 18:33:15 +00001931<!--_______________________________________________________________________-->
1932<div class="doc_subsubsection">
1933 <a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
1934</div>
1935
1936<div class="doc_text">
1937
Tanya Lattnerc5dfcdb2007-06-20 20:46:37 +00001938<p>Deleting a global variable from a module is just as easy as deleting an
1939Instruction. First, you must have a pointer to the global variable that you wish
1940 to delete. You use this pointer to erase it from its parent, the module.
Tanya Lattnerb011c662007-06-20 18:33:15 +00001941 For example:</p>
1942
1943<div class="doc_code">
1944<pre>
1945<a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
Tanya Lattnerb011c662007-06-20 18:33:15 +00001946
Tanya Lattnerc5dfcdb2007-06-20 20:46:37 +00001947GV-&gt;eraseFromParent();
Tanya Lattnerb011c662007-06-20 18:33:15 +00001948</pre>
1949</div>
1950
1951</div>
1952
Chris Lattner9355b472002-09-06 02:50:58 +00001953<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +00001954<div class="doc_section">
Chris Lattnerd9d6e102005-04-23 16:10:52 +00001955 <a name="advanced">Advanced Topics</a>
1956</div>
1957<!-- *********************************************************************** -->
1958
1959<div class="doc_text">
Chris Lattnerf1b200b2005-04-23 17:27:36 +00001960<p>
1961This section describes some of the advanced or obscure API's that most clients
1962do not need to be aware of. These API's tend manage the inner workings of the
1963LLVM system, and only need to be accessed in unusual circumstances.
1964</p>
1965</div>
Chris Lattnerd9d6e102005-04-23 16:10:52 +00001966
Chris Lattnerf1b200b2005-04-23 17:27:36 +00001967<!-- ======================================================================= -->
1968<div class="doc_subsection">
1969 <a name="TypeResolve">LLVM Type Resolution</a>
1970</div>
Chris Lattnerd9d6e102005-04-23 16:10:52 +00001971
Chris Lattnerf1b200b2005-04-23 17:27:36 +00001972<div class="doc_text">
1973
1974<p>
1975The LLVM type system has a very simple goal: allow clients to compare types for
1976structural equality with a simple pointer comparison (aka a shallow compare).
1977This goal makes clients much simpler and faster, and is used throughout the LLVM
1978system.
1979</p>
1980
1981<p>
1982Unfortunately achieving this goal is not a simple matter. In particular,
1983recursive types and late resolution of opaque types makes the situation very
1984difficult to handle. Fortunately, for the most part, our implementation makes
1985most clients able to be completely unaware of the nasty internal details. The
1986primary case where clients are exposed to the inner workings of it are when
Gabor Greif04367bf2007-07-06 22:07:22 +00001987building a recursive type. In addition to this case, the LLVM bitcode reader,
Chris Lattnerf1b200b2005-04-23 17:27:36 +00001988assembly parser, and linker also have to be aware of the inner workings of this
1989system.
1990</p>
1991
Chris Lattner0f876db2005-04-25 15:47:57 +00001992<p>
1993For our purposes below, we need three concepts. First, an "Opaque Type" is
1994exactly as defined in the <a href="LangRef.html#t_opaque">language
1995reference</a>. Second an "Abstract Type" is any type which includes an
Reid Spencer06565dc2007-01-12 17:11:23 +00001996opaque type as part of its type graph (for example "<tt>{ opaque, i32 }</tt>").
1997Third, a concrete type is a type that is not an abstract type (e.g. "<tt>{ i32,
Chris Lattner0f876db2005-04-25 15:47:57 +00001998float }</tt>").
1999</p>
2000
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002001</div>
2002
2003<!-- ______________________________________________________________________ -->
2004<div class="doc_subsubsection">
2005 <a name="BuildRecType">Basic Recursive Type Construction</a>
2006</div>
2007
2008<div class="doc_text">
2009
2010<p>
2011Because the most common question is "how do I build a recursive type with LLVM",
2012we answer it now and explain it as we go. Here we include enough to cause this
2013to be emitted to an output .ll file:
2014</p>
2015
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002016<div class="doc_code">
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002017<pre>
Reid Spencer06565dc2007-01-12 17:11:23 +00002018%mylist = type { %mylist*, i32 }
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002019</pre>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002020</div>
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002021
2022<p>
2023To build this, use the following LLVM APIs:
2024</p>
2025
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002026<div class="doc_code">
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002027<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +00002028// <i>Create the initial outer struct</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002029<a href="#PATypeHolder">PATypeHolder</a> StructTy = OpaqueType::get();
2030std::vector&lt;const Type*&gt; Elts;
2031Elts.push_back(PointerType::get(StructTy));
2032Elts.push_back(Type::IntTy);
2033StructType *NewSTy = StructType::get(Elts);
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002034
Reid Spencer06565dc2007-01-12 17:11:23 +00002035// <i>At this point, NewSTy = "{ opaque*, i32 }". Tell VMCore that</i>
Bill Wendling82e2eea2006-10-11 18:00:22 +00002036// <i>the struct and the opaque type are actually the same.</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002037cast&lt;OpaqueType&gt;(StructTy.get())-&gt;<a href="#refineAbstractTypeTo">refineAbstractTypeTo</a>(NewSTy);
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002038
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002039// <i>NewSTy is potentially invalidated, but StructTy (a <a href="#PATypeHolder">PATypeHolder</a>) is</i>
Bill Wendling82e2eea2006-10-11 18:00:22 +00002040// <i>kept up-to-date</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002041NewSTy = cast&lt;StructType&gt;(StructTy.get());
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002042
Bill Wendling82e2eea2006-10-11 18:00:22 +00002043// <i>Add a name for the type to the module symbol table (optional)</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002044MyModule-&gt;addTypeName("mylist", NewSTy);
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002045</pre>
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002046</div>
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002047
2048<p>
2049This code shows the basic approach used to build recursive types: build a
2050non-recursive type using 'opaque', then use type unification to close the cycle.
2051The type unification step is performed by the <tt><a
Chris Lattneraff26d12007-02-03 03:06:52 +00002052href="#refineAbstractTypeTo">refineAbstractTypeTo</a></tt> method, which is
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002053described next. After that, we describe the <a
2054href="#PATypeHolder">PATypeHolder class</a>.
2055</p>
2056
2057</div>
2058
2059<!-- ______________________________________________________________________ -->
2060<div class="doc_subsubsection">
2061 <a name="refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a>
2062</div>
2063
2064<div class="doc_text">
2065<p>
2066The <tt>refineAbstractTypeTo</tt> method starts the type unification process.
2067While this method is actually a member of the DerivedType class, it is most
2068often used on OpaqueType instances. Type unification is actually a recursive
2069process. After unification, types can become structurally isomorphic to
2070existing types, and all duplicates are deleted (to preserve pointer equality).
2071</p>
2072
2073<p>
2074In the example above, the OpaqueType object is definitely deleted.
Reid Spencer06565dc2007-01-12 17:11:23 +00002075Additionally, if there is an "{ \2*, i32}" type already created in the system,
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002076the pointer and struct type created are <b>also</b> deleted. Obviously whenever
2077a type is deleted, any "Type*" pointers in the program are invalidated. As
2078such, it is safest to avoid having <i>any</i> "Type*" pointers to abstract types
2079live across a call to <tt>refineAbstractTypeTo</tt> (note that non-abstract
2080types can never move or be deleted). To deal with this, the <a
2081href="#PATypeHolder">PATypeHolder</a> class is used to maintain a stable
2082reference to a possibly refined type, and the <a
2083href="#AbstractTypeUser">AbstractTypeUser</a> class is used to update more
2084complex datastructures.
2085</p>
2086
2087</div>
2088
2089<!-- ______________________________________________________________________ -->
2090<div class="doc_subsubsection">
2091 <a name="PATypeHolder">The PATypeHolder Class</a>
2092</div>
2093
2094<div class="doc_text">
2095<p>
2096PATypeHolder is a form of a "smart pointer" for Type objects. When VMCore
2097happily goes about nuking types that become isomorphic to existing types, it
2098automatically updates all PATypeHolder objects to point to the new type. In the
2099example above, this allows the code to maintain a pointer to the resultant
2100resolved recursive type, even though the Type*'s are potentially invalidated.
2101</p>
2102
2103<p>
2104PATypeHolder is an extremely light-weight object that uses a lazy union-find
2105implementation to update pointers. For example the pointer from a Value to its
2106Type is maintained by PATypeHolder objects.
2107</p>
2108
2109</div>
2110
2111<!-- ______________________________________________________________________ -->
2112<div class="doc_subsubsection">
2113 <a name="AbstractTypeUser">The AbstractTypeUser Class</a>
2114</div>
2115
2116<div class="doc_text">
2117
2118<p>
2119Some data structures need more to perform more complex updates when types get
Chris Lattner263a98e2007-02-16 04:37:31 +00002120resolved. To support this, a class can derive from the AbstractTypeUser class.
2121This class
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002122allows it to get callbacks when certain types are resolved. To register to get
2123callbacks for a particular type, the DerivedType::{add/remove}AbstractTypeUser
Chris Lattner0f876db2005-04-25 15:47:57 +00002124methods can be called on a type. Note that these methods only work for <i>
Reid Spencer06565dc2007-01-12 17:11:23 +00002125 abstract</i> types. Concrete types (those that do not include any opaque
2126objects) can never be refined.
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002127</p>
Chris Lattnerd9d6e102005-04-23 16:10:52 +00002128</div>
2129
2130
2131<!-- ======================================================================= -->
2132<div class="doc_subsection">
Chris Lattner263a98e2007-02-16 04:37:31 +00002133 <a name="SymbolTable">The <tt>ValueSymbolTable</tt> and
2134 <tt>TypeSymbolTable</tt> classes</a>
Chris Lattnerd9d6e102005-04-23 16:10:52 +00002135</div>
Chris Lattnerf1b200b2005-04-23 17:27:36 +00002136
Chris Lattnerd9d6e102005-04-23 16:10:52 +00002137<div class="doc_text">
Chris Lattner263a98e2007-02-16 04:37:31 +00002138<p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
2139ValueSymbolTable</a></tt> class provides a symbol table that the <a
Chris Lattnerd9d6e102005-04-23 16:10:52 +00002140href="#Function"><tt>Function</tt></a> and <a href="#Module">
Chris Lattner263a98e2007-02-16 04:37:31 +00002141<tt>Module</tt></a> classes use for naming value definitions. The symbol table
2142can provide a name for any <a href="#Value"><tt>Value</tt></a>.
2143The <tt><a href="http://llvm.org/doxygen/classllvm_1_1TypeSymbolTable.html">
2144TypeSymbolTable</a></tt> class is used by the <tt>Module</tt> class to store
2145names for types.</p>
Chris Lattnerd9d6e102005-04-23 16:10:52 +00002146
Reid Spencera6362242007-01-07 00:41:39 +00002147<p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
2148by most clients. It should only be used when iteration over the symbol table
2149names themselves are required, which is very special purpose. Note that not
2150all LLVM
Chris Lattnerd9d6e102005-04-23 16:10:52 +00002151<a href="#Value">Value</a>s have names, and those without names (i.e. they have
2152an empty name) do not exist in the symbol table.
2153</p>
2154
Chris Lattner263a98e2007-02-16 04:37:31 +00002155<p>These symbol tables support iteration over the values/types in the symbol
2156table with <tt>begin/end/iterator</tt> and supports querying to see if a
2157specific name is in the symbol table (with <tt>lookup</tt>). The
2158<tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
2159simply call <tt>setName</tt> on a value, which will autoinsert it into the
2160appropriate symbol table. For types, use the Module::addTypeName method to
2161insert entries into the symbol table.</p>
Chris Lattnerd9d6e102005-04-23 16:10:52 +00002162
Chris Lattnerd9d6e102005-04-23 16:10:52 +00002163</div>
2164
2165
2166
2167<!-- *********************************************************************** -->
2168<div class="doc_section">
Misha Brukman13fd15c2004-01-15 00:14:41 +00002169 <a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
2170</div>
2171<!-- *********************************************************************** -->
2172
2173<div class="doc_text">
Reid Spencer303c4b42007-01-12 17:26:25 +00002174<p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
2175<br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002176
2177<p>The Core LLVM classes are the primary means of representing the program
Chris Lattner261efe92003-11-25 01:02:51 +00002178being inspected or transformed. The core LLVM classes are defined in
2179header files in the <tt>include/llvm/</tt> directory, and implemented in
Misha Brukman13fd15c2004-01-15 00:14:41 +00002180the <tt>lib/VMCore</tt> directory.</p>
2181
2182</div>
2183
2184<!-- ======================================================================= -->
2185<div class="doc_subsection">
Reid Spencer303c4b42007-01-12 17:26:25 +00002186 <a name="Type">The <tt>Type</tt> class and Derived Types</a>
2187</div>
2188
2189<div class="doc_text">
2190
2191 <p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
2192 a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
2193 through its subclasses. Certain primitive types (<tt>VoidType</tt>,
2194 <tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
2195 subclasses. They are hidden because they offer no useful functionality beyond
2196 what the <tt>Type</tt> class offers except to distinguish themselves from
2197 other subclasses of <tt>Type</tt>.</p>
2198 <p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
2199 named, but this is not a requirement. There exists exactly
2200 one instance of a given shape at any one time. This allows type equality to
2201 be performed with address equality of the Type Instance. That is, given two
2202 <tt>Type*</tt> values, the types are identical if the pointers are identical.
2203 </p>
2204</div>
2205
2206<!-- _______________________________________________________________________ -->
2207<div class="doc_subsubsection">
2208 <a name="m_Value">Important Public Methods</a>
2209</div>
2210
2211<div class="doc_text">
2212
2213<ul>
Chris Lattner8f79df32007-01-15 01:55:32 +00002214 <li><tt>bool isInteger() const</tt>: Returns true for any integer type.</li>
Reid Spencer303c4b42007-01-12 17:26:25 +00002215
2216 <li><tt>bool isFloatingPoint()</tt>: Return true if this is one of the two
2217 floating point types.</li>
2218
2219 <li><tt>bool isAbstract()</tt>: Return true if the type is abstract (contains
2220 an OpaqueType anywhere in its definition).</li>
2221
2222 <li><tt>bool isSized()</tt>: Return true if the type has known size. Things
2223 that don't have a size are abstract types, labels and void.</li>
2224
2225</ul>
2226</div>
2227
2228<!-- _______________________________________________________________________ -->
2229<div class="doc_subsubsection">
2230 <a name="m_Value">Important Derived Types</a>
2231</div>
2232<div class="doc_text">
2233<dl>
2234 <dt><tt>IntegerType</tt></dt>
2235 <dd>Subclass of DerivedType that represents integer types of any bit width.
2236 Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
2237 <tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
2238 <ul>
2239 <li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
2240 type of a specific bit width.</li>
2241 <li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
2242 type.</li>
2243 </ul>
2244 </dd>
2245 <dt><tt>SequentialType</tt></dt>
2246 <dd>This is subclassed by ArrayType and PointerType
2247 <ul>
2248 <li><tt>const Type * getElementType() const</tt>: Returns the type of each
2249 of the elements in the sequential type. </li>
2250 </ul>
2251 </dd>
2252 <dt><tt>ArrayType</tt></dt>
2253 <dd>This is a subclass of SequentialType and defines the interface for array
2254 types.
2255 <ul>
2256 <li><tt>unsigned getNumElements() const</tt>: Returns the number of
2257 elements in the array. </li>
2258 </ul>
2259 </dd>
2260 <dt><tt>PointerType</tt></dt>
Chris Lattner302da1e2007-02-03 03:05:57 +00002261 <dd>Subclass of SequentialType for pointer types.</dd>
Reid Spencer9d6565a2007-02-15 02:26:10 +00002262 <dt><tt>VectorType</tt></dt>
Reid Spencer485bad12007-02-15 03:07:05 +00002263 <dd>Subclass of SequentialType for vector types. A
2264 vector type is similar to an ArrayType but is distinguished because it is
2265 a first class type wherease ArrayType is not. Vector types are used for
Reid Spencer303c4b42007-01-12 17:26:25 +00002266 vector operations and are usually small vectors of of an integer or floating
2267 point type.</dd>
2268 <dt><tt>StructType</tt></dt>
2269 <dd>Subclass of DerivedTypes for struct types.</dd>
Duncan Sands8036ca42007-03-30 12:22:09 +00002270 <dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
Reid Spencer303c4b42007-01-12 17:26:25 +00002271 <dd>Subclass of DerivedTypes for function types.
2272 <ul>
2273 <li><tt>bool isVarArg() const</tt>: Returns true if its a vararg
2274 function</li>
2275 <li><tt> const Type * getReturnType() const</tt>: Returns the
2276 return type of the function.</li>
2277 <li><tt>const Type * getParamType (unsigned i)</tt>: Returns
2278 the type of the ith parameter.</li>
2279 <li><tt> const unsigned getNumParams() const</tt>: Returns the
2280 number of formal parameters.</li>
2281 </ul>
2282 </dd>
2283 <dt><tt>OpaqueType</tt></dt>
2284 <dd>Sublcass of DerivedType for abstract types. This class
2285 defines no content and is used as a placeholder for some other type. Note
2286 that OpaqueType is used (temporarily) during type resolution for forward
2287 references of types. Once the referenced type is resolved, the OpaqueType
2288 is replaced with the actual type. OpaqueType can also be used for data
2289 abstraction. At link time opaque types can be resolved to actual types
2290 of the same name.</dd>
2291</dl>
2292</div>
2293
Chris Lattner2b78d962007-02-03 20:02:25 +00002294
2295
2296<!-- ======================================================================= -->
2297<div class="doc_subsection">
2298 <a name="Module">The <tt>Module</tt> class</a>
2299</div>
2300
2301<div class="doc_text">
2302
2303<p><tt>#include "<a
2304href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
2305<a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
2306
2307<p>The <tt>Module</tt> class represents the top level structure present in LLVM
2308programs. An LLVM module is effectively either a translation unit of the
2309original program or a combination of several translation units merged by the
2310linker. The <tt>Module</tt> class keeps track of a list of <a
2311href="#Function"><tt>Function</tt></a>s, a list of <a
2312href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
2313href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
2314helpful member functions that try to make common operations easy.</p>
2315
2316</div>
2317
2318<!-- _______________________________________________________________________ -->
2319<div class="doc_subsubsection">
2320 <a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
2321</div>
2322
2323<div class="doc_text">
2324
2325<ul>
2326 <li><tt>Module::Module(std::string name = "")</tt></li>
2327</ul>
2328
2329<p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
2330provide a name for it (probably based on the name of the translation unit).</p>
2331
2332<ul>
2333 <li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
2334 <tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
2335
2336 <tt>begin()</tt>, <tt>end()</tt>
2337 <tt>size()</tt>, <tt>empty()</tt>
2338
2339 <p>These are forwarding methods that make it easy to access the contents of
2340 a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
2341 list.</p></li>
2342
2343 <li><tt>Module::FunctionListType &amp;getFunctionList()</tt>
2344
2345 <p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
2346 necessary to use when you need to update the list or perform a complex
2347 action that doesn't have a forwarding method.</p>
2348
2349 <p><!-- Global Variable --></p></li>
2350</ul>
2351
2352<hr>
2353
2354<ul>
2355 <li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
2356
2357 <tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
2358
2359 <tt>global_begin()</tt>, <tt>global_end()</tt>
2360 <tt>global_size()</tt>, <tt>global_empty()</tt>
2361
2362 <p> These are forwarding methods that make it easy to access the contents of
2363 a <tt>Module</tt> object's <a
2364 href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
2365
2366 <li><tt>Module::GlobalListType &amp;getGlobalList()</tt>
2367
2368 <p>Returns the list of <a
2369 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
2370 use when you need to update the list or perform a complex action that
2371 doesn't have a forwarding method.</p>
2372
2373 <p><!-- Symbol table stuff --> </p></li>
2374</ul>
2375
2376<hr>
2377
2378<ul>
2379 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
2380
2381 <p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
2382 for this <tt>Module</tt>.</p>
2383
2384 <p><!-- Convenience methods --></p></li>
2385</ul>
2386
2387<hr>
2388
2389<ul>
2390 <li><tt><a href="#Function">Function</a> *getFunction(const std::string
2391 &amp;Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt>
2392
2393 <p>Look up the specified function in the <tt>Module</tt> <a
2394 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
2395 <tt>null</tt>.</p></li>
2396
2397 <li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
2398 std::string &amp;Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
2399
2400 <p>Look up the specified function in the <tt>Module</tt> <a
2401 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
2402 external declaration for the function and return it.</p></li>
2403
2404 <li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
2405
2406 <p>If there is at least one entry in the <a
2407 href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
2408 href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
2409 string.</p></li>
2410
2411 <li><tt>bool addTypeName(const std::string &amp;Name, const <a
2412 href="#Type">Type</a> *Ty)</tt>
2413
2414 <p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
2415 mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
2416 name, true is returned and the <a
2417 href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
2418</ul>
2419
2420</div>
2421
2422
Reid Spencer303c4b42007-01-12 17:26:25 +00002423<!-- ======================================================================= -->
2424<div class="doc_subsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +00002425 <a name="Value">The <tt>Value</tt> class</a>
2426</div>
2427
Chris Lattner2b78d962007-02-03 20:02:25 +00002428<div class="doc_text">
Misha Brukman13fd15c2004-01-15 00:14:41 +00002429
2430<p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
2431<br>
Chris Lattner00815172007-01-04 22:01:45 +00002432doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002433
2434<p>The <tt>Value</tt> class is the most important class in the LLVM Source
2435base. It represents a typed value that may be used (among other things) as an
2436operand to an instruction. There are many different types of <tt>Value</tt>s,
2437such as <a href="#Constant"><tt>Constant</tt></a>s,<a
2438href="#Argument"><tt>Argument</tt></a>s. Even <a
2439href="#Instruction"><tt>Instruction</tt></a>s and <a
2440href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
2441
2442<p>A particular <tt>Value</tt> may be used many times in the LLVM representation
2443for a program. For example, an incoming argument to a function (represented
2444with an instance of the <a href="#Argument">Argument</a> class) is "used" by
2445every instruction in the function that references the argument. To keep track
2446of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
2447href="#User"><tt>User</tt></a>s that is using it (the <a
2448href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
2449graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
2450def-use information in the program, and is accessible through the <tt>use_</tt>*
2451methods, shown below.</p>
2452
2453<p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
2454and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
2455method. In addition, all LLVM values can be named. The "name" of the
2456<tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
2457
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002458<div class="doc_code">
2459<pre>
Reid Spencer06565dc2007-01-12 17:11:23 +00002460%<b>foo</b> = add i32 1, 2
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002461</pre>
2462</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002463
Duncan Sands8036ca42007-03-30 12:22:09 +00002464<p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002465that the name of any value may be missing (an empty string), so names should
2466<b>ONLY</b> be used for debugging (making the source code easier to read,
2467debugging printouts), they should not be used to keep track of values or map
2468between them. For this purpose, use a <tt>std::map</tt> of pointers to the
2469<tt>Value</tt> itself instead.</p>
2470
2471<p>One important aspect of LLVM is that there is no distinction between an SSA
2472variable and the operation that produces it. Because of this, any reference to
2473the value produced by an instruction (or the value available as an incoming
Chris Lattnerd5fc4fc2004-03-18 14:58:55 +00002474argument, for example) is represented as a direct pointer to the instance of
2475the class that
Misha Brukman13fd15c2004-01-15 00:14:41 +00002476represents this value. Although this may take some getting used to, it
2477simplifies the representation and makes it easier to manipulate.</p>
2478
2479</div>
2480
2481<!-- _______________________________________________________________________ -->
2482<div class="doc_subsubsection">
2483 <a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
2484</div>
2485
2486<div class="doc_text">
2487
Chris Lattner261efe92003-11-25 01:02:51 +00002488<ul>
2489 <li><tt>Value::use_iterator</tt> - Typedef for iterator over the
2490use-list<br>
2491 <tt>Value::use_const_iterator</tt> - Typedef for const_iterator over
2492the use-list<br>
2493 <tt>unsigned use_size()</tt> - Returns the number of users of the
2494value.<br>
Chris Lattner9355b472002-09-06 02:50:58 +00002495 <tt>bool use_empty()</tt> - Returns true if there are no users.<br>
Chris Lattner261efe92003-11-25 01:02:51 +00002496 <tt>use_iterator use_begin()</tt> - Get an iterator to the start of
2497the use-list.<br>
2498 <tt>use_iterator use_end()</tt> - Get an iterator to the end of the
2499use-list.<br>
2500 <tt><a href="#User">User</a> *use_back()</tt> - Returns the last
2501element in the list.
2502 <p> These methods are the interface to access the def-use
2503information in LLVM. As with all other iterators in LLVM, the naming
2504conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
Chris Lattner261efe92003-11-25 01:02:51 +00002505 </li>
2506 <li><tt><a href="#Type">Type</a> *getType() const</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002507 <p>This method returns the Type of the Value.</p>
Chris Lattner261efe92003-11-25 01:02:51 +00002508 </li>
2509 <li><tt>bool hasName() const</tt><br>
Chris Lattner9355b472002-09-06 02:50:58 +00002510 <tt>std::string getName() const</tt><br>
Chris Lattner261efe92003-11-25 01:02:51 +00002511 <tt>void setName(const std::string &amp;Name)</tt>
2512 <p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
2513be aware of the <a href="#nameWarning">precaution above</a>.</p>
Chris Lattner261efe92003-11-25 01:02:51 +00002514 </li>
2515 <li><tt>void replaceAllUsesWith(Value *V)</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002516
2517 <p>This method traverses the use list of a <tt>Value</tt> changing all <a
2518 href="#User"><tt>User</tt>s</a> of the current value to refer to
2519 "<tt>V</tt>" instead. For example, if you detect that an instruction always
2520 produces a constant value (for example through constant folding), you can
2521 replace all uses of the instruction with the constant like this:</p>
2522
Bill Wendling3cd5ca62006-10-11 06:30:10 +00002523<div class="doc_code">
2524<pre>
2525Inst-&gt;replaceAllUsesWith(ConstVal);
2526</pre>
2527</div>
2528
Chris Lattner261efe92003-11-25 01:02:51 +00002529</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002530
2531</div>
2532
2533<!-- ======================================================================= -->
2534<div class="doc_subsection">
2535 <a name="User">The <tt>User</tt> class</a>
2536</div>
2537
2538<div class="doc_text">
2539
2540<p>
2541<tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
Misha Brukman384047f2004-06-03 23:29:12 +00002542doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002543Superclass: <a href="#Value"><tt>Value</tt></a></p>
2544
2545<p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
2546refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
2547that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
2548referring to. The <tt>User</tt> class itself is a subclass of
2549<tt>Value</tt>.</p>
2550
2551<p>The operands of a <tt>User</tt> point directly to the LLVM <a
2552href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
2553Single Assignment (SSA) form, there can only be one definition referred to,
2554allowing this direct connection. This connection provides the use-def
2555information in LLVM.</p>
2556
2557</div>
2558
2559<!-- _______________________________________________________________________ -->
2560<div class="doc_subsubsection">
2561 <a name="m_User">Important Public Members of the <tt>User</tt> class</a>
2562</div>
2563
2564<div class="doc_text">
2565
2566<p>The <tt>User</tt> class exposes the operand list in two ways: through
2567an index access interface and through an iterator based interface.</p>
2568
Chris Lattner261efe92003-11-25 01:02:51 +00002569<ul>
Chris Lattner261efe92003-11-25 01:02:51 +00002570 <li><tt>Value *getOperand(unsigned i)</tt><br>
2571 <tt>unsigned getNumOperands()</tt>
2572 <p> These two methods expose the operands of the <tt>User</tt> in a
Misha Brukman13fd15c2004-01-15 00:14:41 +00002573convenient form for direct access.</p></li>
2574
Chris Lattner261efe92003-11-25 01:02:51 +00002575 <li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
2576list<br>
Chris Lattner58360822005-01-17 00:12:04 +00002577 <tt>op_iterator op_begin()</tt> - Get an iterator to the start of
2578the operand list.<br>
2579 <tt>op_iterator op_end()</tt> - Get an iterator to the end of the
Chris Lattner261efe92003-11-25 01:02:51 +00002580operand list.
2581 <p> Together, these methods make up the iterator based interface to
Misha Brukman13fd15c2004-01-15 00:14:41 +00002582the operands of a <tt>User</tt>.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +00002583</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002584
2585</div>
2586
2587<!-- ======================================================================= -->
2588<div class="doc_subsection">
2589 <a name="Instruction">The <tt>Instruction</tt> class</a>
2590</div>
2591
2592<div class="doc_text">
2593
2594<p><tt>#include "</tt><tt><a
2595href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
Misha Brukman31ca1de2004-06-03 23:35:54 +00002596doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002597Superclasses: <a href="#User"><tt>User</tt></a>, <a
2598href="#Value"><tt>Value</tt></a></p>
2599
2600<p>The <tt>Instruction</tt> class is the common base class for all LLVM
2601instructions. It provides only a few methods, but is a very commonly used
2602class. The primary data tracked by the <tt>Instruction</tt> class itself is the
2603opcode (instruction type) and the parent <a
2604href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
2605into. To represent a specific type of instruction, one of many subclasses of
2606<tt>Instruction</tt> are used.</p>
2607
2608<p> Because the <tt>Instruction</tt> class subclasses the <a
2609href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
2610way as for other <a href="#User"><tt>User</tt></a>s (with the
2611<tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
2612<tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
2613the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
2614file contains some meta-data about the various different types of instructions
2615in LLVM. It describes the enum values that are used as opcodes (for example
Reid Spencerc92d25d2006-12-19 19:47:19 +00002616<tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
Misha Brukman13fd15c2004-01-15 00:14:41 +00002617concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
2618example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
Reid Spencerc92d25d2006-12-19 19:47:19 +00002619href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
Misha Brukman13fd15c2004-01-15 00:14:41 +00002620this file confuses doxygen, so these enum values don't show up correctly in the
Misha Brukman31ca1de2004-06-03 23:35:54 +00002621<a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002622
2623</div>
2624
2625<!-- _______________________________________________________________________ -->
2626<div class="doc_subsubsection">
Reid Spencerc92d25d2006-12-19 19:47:19 +00002627 <a name="s_Instruction">Important Subclasses of the <tt>Instruction</tt>
2628 class</a>
2629</div>
2630<div class="doc_text">
2631 <ul>
2632 <li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
2633 <p>This subclasses represents all two operand instructions whose operands
2634 must be the same type, except for the comparison instructions.</p></li>
2635 <li><tt><a name="CastInst">CastInst</a></tt>
2636 <p>This subclass is the parent of the 12 casting instructions. It provides
2637 common operations on cast instructions.</p>
2638 <li><tt><a name="CmpInst">CmpInst</a></tt>
2639 <p>This subclass respresents the two comparison instructions,
2640 <a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
2641 <a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
2642 <li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
2643 <p>This subclass is the parent of all terminator instructions (those which
2644 can terminate a block).</p>
2645 </ul>
2646 </div>
2647
2648<!-- _______________________________________________________________________ -->
2649<div class="doc_subsubsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +00002650 <a name="m_Instruction">Important Public Members of the <tt>Instruction</tt>
2651 class</a>
2652</div>
2653
2654<div class="doc_text">
2655
Chris Lattner261efe92003-11-25 01:02:51 +00002656<ul>
2657 <li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002658 <p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
2659this <tt>Instruction</tt> is embedded into.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +00002660 <li><tt>bool mayWriteToMemory()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002661 <p>Returns true if the instruction writes to memory, i.e. it is a
2662 <tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +00002663 <li><tt>unsigned getOpcode()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002664 <p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +00002665 <li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002666 <p>Returns another instance of the specified instruction, identical
Chris Lattner261efe92003-11-25 01:02:51 +00002667in all ways to the original except that the instruction has no parent
2668(ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
Misha Brukman13fd15c2004-01-15 00:14:41 +00002669and it has no name</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +00002670</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002671
2672</div>
2673
2674<!-- ======================================================================= -->
2675<div class="doc_subsection">
Chris Lattner2b78d962007-02-03 20:02:25 +00002676 <a name="Constant">The <tt>Constant</tt> class and subclasses</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002677</div>
2678
2679<div class="doc_text">
2680
Chris Lattner2b78d962007-02-03 20:02:25 +00002681<p>Constant represents a base class for different types of constants. It
2682is subclassed by ConstantInt, ConstantArray, etc. for representing
2683the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
2684a subclass, which represents the address of a global variable or function.
2685</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002686
2687</div>
2688
2689<!-- _______________________________________________________________________ -->
Chris Lattner2b78d962007-02-03 20:02:25 +00002690<div class="doc_subsubsection">Important Subclasses of Constant </div>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002691<div class="doc_text">
Chris Lattner261efe92003-11-25 01:02:51 +00002692<ul>
Chris Lattner2b78d962007-02-03 20:02:25 +00002693 <li>ConstantInt : This subclass of Constant represents an integer constant of
2694 any width.
2695 <ul>
Reid Spencer97b4ee32007-03-01 21:05:33 +00002696 <li><tt>const APInt&amp; getValue() const</tt>: Returns the underlying
2697 value of this constant, an APInt value.</li>
2698 <li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
2699 value to an int64_t via sign extension. If the value (not the bit width)
2700 of the APInt is too large to fit in an int64_t, an assertion will result.
2701 For this reason, use of this method is discouraged.</li>
2702 <li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
2703 value to a uint64_t via zero extension. IF the value (not the bit width)
2704 of the APInt is too large to fit in a uint64_t, an assertion will result.
Reid Spencer4474d872007-03-02 01:31:31 +00002705 For this reason, use of this method is discouraged.</li>
Reid Spencer97b4ee32007-03-01 21:05:33 +00002706 <li><tt>static ConstantInt* get(const APInt&amp; Val)</tt>: Returns the
2707 ConstantInt object that represents the value provided by <tt>Val</tt>.
2708 The type is implied as the IntegerType that corresponds to the bit width
2709 of <tt>Val</tt>.</li>
Chris Lattner2b78d962007-02-03 20:02:25 +00002710 <li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
2711 Returns the ConstantInt object that represents the value provided by
2712 <tt>Val</tt> for integer type <tt>Ty</tt>.</li>
2713 </ul>
2714 </li>
2715 <li>ConstantFP : This class represents a floating point constant.
2716 <ul>
2717 <li><tt>double getValue() const</tt>: Returns the underlying value of
2718 this constant. </li>
2719 </ul>
2720 </li>
2721 <li>ConstantArray : This represents a constant array.
2722 <ul>
2723 <li><tt>const std::vector&lt;Use&gt; &amp;getValues() const</tt>: Returns
2724 a vector of component constants that makeup this array. </li>
2725 </ul>
2726 </li>
2727 <li>ConstantStruct : This represents a constant struct.
2728 <ul>
2729 <li><tt>const std::vector&lt;Use&gt; &amp;getValues() const</tt>: Returns
2730 a vector of component constants that makeup this array. </li>
2731 </ul>
2732 </li>
2733 <li>GlobalValue : This represents either a global variable or a function. In
2734 either case, the value is a constant fixed address (after linking).
2735 </li>
Chris Lattner261efe92003-11-25 01:02:51 +00002736</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002737</div>
2738
Chris Lattner2b78d962007-02-03 20:02:25 +00002739
Misha Brukman13fd15c2004-01-15 00:14:41 +00002740<!-- ======================================================================= -->
2741<div class="doc_subsection">
2742 <a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
2743</div>
2744
2745<div class="doc_text">
2746
2747<p><tt>#include "<a
2748href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
Misha Brukman384047f2004-06-03 23:29:12 +00002749doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
2750Class</a><br>
Reid Spencerbe5e85e2006-04-14 14:11:48 +00002751Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
2752<a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002753
2754<p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
2755href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
2756visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
2757Because they are visible at global scope, they are also subject to linking with
2758other globals defined in different translation units. To control the linking
2759process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
2760<tt>GlobalValue</tt>s know whether they have internal or external linkage, as
Reid Spencer8b2da7a2004-07-18 13:10:31 +00002761defined by the <tt>LinkageTypes</tt> enumeration.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002762
2763<p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
2764<tt>static</tt> in C), it is not visible to code outside the current translation
2765unit, and does not participate in linking. If it has external linkage, it is
2766visible to external code, and does participate in linking. In addition to
2767linkage information, <tt>GlobalValue</tt>s keep track of which <a
2768href="#Module"><tt>Module</tt></a> they are currently part of.</p>
2769
2770<p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
2771by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
2772global is always a pointer to its contents. It is important to remember this
2773when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
2774be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
2775subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
Reid Spencer06565dc2007-01-12 17:11:23 +00002776i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
Misha Brukman13fd15c2004-01-15 00:14:41 +00002777the address of the first element of this array and the value of the
2778<tt>GlobalVariable</tt> are the same, they have different types. The
Reid Spencer06565dc2007-01-12 17:11:23 +00002779<tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
2780is <tt>i32.</tt> Because of this, accessing a global value requires you to
Misha Brukman13fd15c2004-01-15 00:14:41 +00002781dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
2782can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
2783Language Reference Manual</a>.</p>
2784
2785</div>
2786
2787<!-- _______________________________________________________________________ -->
2788<div class="doc_subsubsection">
2789 <a name="m_GlobalValue">Important Public Members of the <tt>GlobalValue</tt>
2790 class</a>
2791</div>
2792
2793<div class="doc_text">
2794
Chris Lattner261efe92003-11-25 01:02:51 +00002795<ul>
2796 <li><tt>bool hasInternalLinkage() const</tt><br>
Chris Lattner9355b472002-09-06 02:50:58 +00002797 <tt>bool hasExternalLinkage() const</tt><br>
Chris Lattner261efe92003-11-25 01:02:51 +00002798 <tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
2799 <p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
2800 <p> </p>
2801 </li>
2802 <li><tt><a href="#Module">Module</a> *getParent()</tt>
2803 <p> This returns the <a href="#Module"><tt>Module</tt></a> that the
Misha Brukman13fd15c2004-01-15 00:14:41 +00002804GlobalValue is currently embedded into.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +00002805</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002806
2807</div>
2808
2809<!-- ======================================================================= -->
2810<div class="doc_subsection">
2811 <a name="Function">The <tt>Function</tt> class</a>
2812</div>
2813
2814<div class="doc_text">
2815
2816<p><tt>#include "<a
2817href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
Misha Brukman31ca1de2004-06-03 23:35:54 +00002818info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
Reid Spencerbe5e85e2006-04-14 14:11:48 +00002819Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
2820<a href="#Constant"><tt>Constant</tt></a>,
2821<a href="#User"><tt>User</tt></a>,
2822<a href="#Value"><tt>Value</tt></a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002823
2824<p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
2825actually one of the more complex classes in the LLVM heirarchy because it must
2826keep track of a large amount of data. The <tt>Function</tt> class keeps track
Reid Spencerbe5e85e2006-04-14 14:11:48 +00002827of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
2828<a href="#Argument"><tt>Argument</tt></a>s, and a
2829<a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002830
2831<p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
2832commonly used part of <tt>Function</tt> objects. The list imposes an implicit
2833ordering of the blocks in the function, which indicate how the code will be
2834layed out by the backend. Additionally, the first <a
2835href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
2836<tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
2837block. There are no implicit exit nodes, and in fact there may be multiple exit
2838nodes from a single <tt>Function</tt>. If the <a
2839href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
2840the <tt>Function</tt> is actually a function declaration: the actual body of the
2841function hasn't been linked in yet.</p>
2842
2843<p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
2844<tt>Function</tt> class also keeps track of the list of formal <a
2845href="#Argument"><tt>Argument</tt></a>s that the function receives. This
2846container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
2847nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
2848the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
2849
2850<p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
2851LLVM feature that is only used when you have to look up a value by name. Aside
2852from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
2853internally to make sure that there are not conflicts between the names of <a
2854href="#Instruction"><tt>Instruction</tt></a>s, <a
2855href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
2856href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
2857
Reid Spencer8b2da7a2004-07-18 13:10:31 +00002858<p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
2859and therefore also a <a href="#Constant">Constant</a>. The value of the function
2860is its address (after linking) which is guaranteed to be constant.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002861</div>
2862
2863<!-- _______________________________________________________________________ -->
2864<div class="doc_subsubsection">
2865 <a name="m_Function">Important Public Members of the <tt>Function</tt>
2866 class</a>
2867</div>
2868
2869<div class="doc_text">
2870
Chris Lattner261efe92003-11-25 01:02:51 +00002871<ul>
2872 <li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
Chris Lattnerac479e52004-08-04 05:10:48 +00002873 *Ty, LinkageTypes Linkage, const std::string &amp;N = "", Module* Parent = 0)</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002874
2875 <p>Constructor used when you need to create new <tt>Function</tt>s to add
2876 the the program. The constructor must specify the type of the function to
Chris Lattnerac479e52004-08-04 05:10:48 +00002877 create and what type of linkage the function should have. The <a
2878 href="#FunctionType"><tt>FunctionType</tt></a> argument
Misha Brukman13fd15c2004-01-15 00:14:41 +00002879 specifies the formal arguments and return value for the function. The same
Duncan Sands8036ca42007-03-30 12:22:09 +00002880 <a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
Misha Brukman13fd15c2004-01-15 00:14:41 +00002881 create multiple functions. The <tt>Parent</tt> argument specifies the Module
2882 in which the function is defined. If this argument is provided, the function
2883 will automatically be inserted into that module's list of
2884 functions.</p></li>
2885
Chris Lattner261efe92003-11-25 01:02:51 +00002886 <li><tt>bool isExternal()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002887
2888 <p>Return whether or not the <tt>Function</tt> has a body defined. If the
2889 function is "external", it does not have a body, and thus must be resolved
2890 by linking with a function defined in a different translation unit.</p></li>
2891
Chris Lattner261efe92003-11-25 01:02:51 +00002892 <li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
Chris Lattner9355b472002-09-06 02:50:58 +00002893 <tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002894
Chris Lattner77d69242005-03-15 05:19:20 +00002895 <tt>begin()</tt>, <tt>end()</tt>
2896 <tt>size()</tt>, <tt>empty()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002897
2898 <p>These are forwarding methods that make it easy to access the contents of
2899 a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
2900 list.</p></li>
2901
Chris Lattner261efe92003-11-25 01:02:51 +00002902 <li><tt>Function::BasicBlockListType &amp;getBasicBlockList()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002903
2904 <p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
2905 is necessary to use when you need to update the list or perform a complex
2906 action that doesn't have a forwarding method.</p></li>
2907
Chris Lattner89cc2652005-03-15 04:48:32 +00002908 <li><tt>Function::arg_iterator</tt> - Typedef for the argument list
Chris Lattner261efe92003-11-25 01:02:51 +00002909iterator<br>
Chris Lattner89cc2652005-03-15 04:48:32 +00002910 <tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002911
Chris Lattner77d69242005-03-15 05:19:20 +00002912 <tt>arg_begin()</tt>, <tt>arg_end()</tt>
Chris Lattner89cc2652005-03-15 04:48:32 +00002913 <tt>arg_size()</tt>, <tt>arg_empty()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002914
2915 <p>These are forwarding methods that make it easy to access the contents of
2916 a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
2917 list.</p></li>
2918
Chris Lattner261efe92003-11-25 01:02:51 +00002919 <li><tt>Function::ArgumentListType &amp;getArgumentList()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002920
2921 <p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
2922 necessary to use when you need to update the list or perform a complex
2923 action that doesn't have a forwarding method.</p></li>
2924
Chris Lattner261efe92003-11-25 01:02:51 +00002925 <li><tt><a href="#BasicBlock">BasicBlock</a> &amp;getEntryBlock()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002926
2927 <p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
2928 function. Because the entry block for the function is always the first
2929 block, this returns the first block of the <tt>Function</tt>.</p></li>
2930
Chris Lattner261efe92003-11-25 01:02:51 +00002931 <li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
2932 <tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002933
2934 <p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
2935 <tt>Function</tt> and returns the return type of the function, or the <a
2936 href="#FunctionType"><tt>FunctionType</tt></a> of the actual
2937 function.</p></li>
2938
Chris Lattner261efe92003-11-25 01:02:51 +00002939 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002940
Chris Lattner261efe92003-11-25 01:02:51 +00002941 <p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002942 for this <tt>Function</tt>.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +00002943</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002944
2945</div>
2946
2947<!-- ======================================================================= -->
2948<div class="doc_subsection">
2949 <a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
2950</div>
2951
2952<div class="doc_text">
2953
2954<p><tt>#include "<a
2955href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
2956<br>
Tanya Lattnera3da7772004-06-22 08:02:25 +00002957doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
Reid Spencerbe5e85e2006-04-14 14:11:48 +00002958 Class</a><br>
2959Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
2960<a href="#Constant"><tt>Constant</tt></a>,
2961<a href="#User"><tt>User</tt></a>,
2962<a href="#Value"><tt>Value</tt></a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002963
2964<p>Global variables are represented with the (suprise suprise)
2965<tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
2966subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
2967always referenced by their address (global values must live in memory, so their
Reid Spencerbe5e85e2006-04-14 14:11:48 +00002968"name" refers to their constant address). See
2969<a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
2970variables may have an initial value (which must be a
2971<a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
2972they may be marked as "constant" themselves (indicating that their contents
2973never change at runtime).</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002974</div>
2975
2976<!-- _______________________________________________________________________ -->
2977<div class="doc_subsubsection">
2978 <a name="m_GlobalVariable">Important Public Members of the
2979 <tt>GlobalVariable</tt> class</a>
2980</div>
2981
2982<div class="doc_text">
2983
Chris Lattner261efe92003-11-25 01:02:51 +00002984<ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00002985 <li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
2986 isConstant, LinkageTypes&amp; Linkage, <a href="#Constant">Constant</a>
2987 *Initializer = 0, const std::string &amp;Name = "", Module* Parent = 0)</tt>
2988
2989 <p>Create a new global variable of the specified type. If
2990 <tt>isConstant</tt> is true then the global variable will be marked as
2991 unchanging for the program. The Linkage parameter specifies the type of
2992 linkage (internal, external, weak, linkonce, appending) for the variable. If
2993 the linkage is InternalLinkage, WeakLinkage, or LinkOnceLinkage,&nbsp; then
2994 the resultant global variable will have internal linkage. AppendingLinkage
2995 concatenates together all instances (in different translation units) of the
2996 variable into a single variable but is only applicable to arrays. &nbsp;See
2997 the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
2998 further details on linkage types. Optionally an initializer, a name, and the
2999 module to put the variable into may be specified for the global variable as
3000 well.</p></li>
3001
Chris Lattner261efe92003-11-25 01:02:51 +00003002 <li><tt>bool isConstant() const</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003003
3004 <p>Returns true if this is a global variable that is known not to
3005 be modified at runtime.</p></li>
3006
Chris Lattner261efe92003-11-25 01:02:51 +00003007 <li><tt>bool hasInitializer()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003008
3009 <p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
3010
Chris Lattner261efe92003-11-25 01:02:51 +00003011 <li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003012
3013 <p>Returns the intial value for a <tt>GlobalVariable</tt>. It is not legal
3014 to call this method if there is no initializer.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +00003015</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003016
3017</div>
3018
Chris Lattner2b78d962007-02-03 20:02:25 +00003019
Misha Brukman13fd15c2004-01-15 00:14:41 +00003020<!-- ======================================================================= -->
3021<div class="doc_subsection">
Chris Lattner2b78d962007-02-03 20:02:25 +00003022 <a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003023</div>
3024
3025<div class="doc_text">
3026
3027<p><tt>#include "<a
Chris Lattner2b78d962007-02-03 20:02:25 +00003028href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
3029doxygen info: <a href="/doxygen/structllvm_1_1BasicBlock.html">BasicBlock
3030Class</a><br>
3031Superclass: <a href="#Value"><tt>Value</tt></a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003032
Chris Lattner2b78d962007-02-03 20:02:25 +00003033<p>This class represents a single entry multiple exit section of the code,
3034commonly known as a basic block by the compiler community. The
3035<tt>BasicBlock</tt> class maintains a list of <a
3036href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
3037Matching the language definition, the last element of this list of instructions
3038is always a terminator instruction (a subclass of the <a
3039href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
3040
3041<p>In addition to tracking the list of instructions that make up the block, the
3042<tt>BasicBlock</tt> class also keeps track of the <a
3043href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
3044
3045<p>Note that <tt>BasicBlock</tt>s themselves are <a
3046href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
3047like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
3048<tt>label</tt>.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003049
3050</div>
3051
3052<!-- _______________________________________________________________________ -->
3053<div class="doc_subsubsection">
Chris Lattner2b78d962007-02-03 20:02:25 +00003054 <a name="m_BasicBlock">Important Public Members of the <tt>BasicBlock</tt>
3055 class</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003056</div>
3057
3058<div class="doc_text">
Chris Lattner261efe92003-11-25 01:02:51 +00003059<ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003060
Chris Lattner2b78d962007-02-03 20:02:25 +00003061<li><tt>BasicBlock(const std::string &amp;Name = "", </tt><tt><a
3062 href="#Function">Function</a> *Parent = 0)</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003063
Chris Lattner2b78d962007-02-03 20:02:25 +00003064<p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
3065insertion into a function. The constructor optionally takes a name for the new
3066block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
3067the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
3068automatically inserted at the end of the specified <a
3069href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
3070manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003071
Chris Lattner2b78d962007-02-03 20:02:25 +00003072<li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
3073<tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
3074<tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
3075<tt>size()</tt>, <tt>empty()</tt>
3076STL-style functions for accessing the instruction list.
Misha Brukman13fd15c2004-01-15 00:14:41 +00003077
Chris Lattner2b78d962007-02-03 20:02:25 +00003078<p>These methods and typedefs are forwarding functions that have the same
3079semantics as the standard library methods of the same names. These methods
3080expose the underlying instruction list of a basic block in a way that is easy to
3081manipulate. To get the full complement of container operations (including
3082operations to update the list), you must use the <tt>getInstList()</tt>
3083method.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003084
Chris Lattner2b78d962007-02-03 20:02:25 +00003085<li><tt>BasicBlock::InstListType &amp;getInstList()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003086
Chris Lattner2b78d962007-02-03 20:02:25 +00003087<p>This method is used to get access to the underlying container that actually
3088holds the Instructions. This method must be used when there isn't a forwarding
3089function in the <tt>BasicBlock</tt> class for the operation that you would like
3090to perform. Because there are no forwarding functions for "updating"
3091operations, you need to use this if you want to update the contents of a
3092<tt>BasicBlock</tt>.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003093
Chris Lattner2b78d962007-02-03 20:02:25 +00003094<li><tt><a href="#Function">Function</a> *getParent()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003095
Chris Lattner2b78d962007-02-03 20:02:25 +00003096<p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
3097embedded into, or a null pointer if it is homeless.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003098
Chris Lattner2b78d962007-02-03 20:02:25 +00003099<li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003100
Chris Lattner2b78d962007-02-03 20:02:25 +00003101<p> Returns a pointer to the terminator instruction that appears at the end of
3102the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
3103instruction in the block is not a terminator, then a null pointer is
3104returned.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003105
Misha Brukman13fd15c2004-01-15 00:14:41 +00003106</ul>
3107
3108</div>
3109
Misha Brukman13fd15c2004-01-15 00:14:41 +00003110
Misha Brukman13fd15c2004-01-15 00:14:41 +00003111<!-- ======================================================================= -->
3112<div class="doc_subsection">
3113 <a name="Argument">The <tt>Argument</tt> class</a>
3114</div>
3115
3116<div class="doc_text">
3117
3118<p>This subclass of Value defines the interface for incoming formal
Chris Lattner58360822005-01-17 00:12:04 +00003119arguments to a function. A Function maintains a list of its formal
Misha Brukman13fd15c2004-01-15 00:14:41 +00003120arguments. An argument has a pointer to the parent Function.</p>
3121
3122</div>
3123
Chris Lattner9355b472002-09-06 02:50:58 +00003124<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +00003125<hr>
3126<address>
3127 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
3128 src="http://jigsaw.w3.org/css-validator/images/vcss" alt="Valid CSS!"></a>
3129 <a href="http://validator.w3.org/check/referer"><img
3130 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01!" /></a>
3131
3132 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
3133 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
Reid Spencer05fe4b02006-03-14 05:39:39 +00003134 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
Misha Brukman13fd15c2004-01-15 00:14:41 +00003135 Last modified: $Date$
3136</address>
3137
Chris Lattner261efe92003-11-25 01:02:51 +00003138</body>
3139</html>