Gitiles
Code Review Sign In
gerrit-public.fairphone.software / fp2-dev / platform / external / llvm / 81187ae82a322caba25211d5384df2e723d62533 / . / docs / ProgrammersManual.html
blob: e3d4e95a262dabb38ee4b33b8aa5af430d3d687e [file] [log] [blame]
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1<!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 +0000[diff] [blame]3<html>
4<head>
Owen Anderson5e8c50e2009-06-16 17:40:28 +0000[diff] [blame]5 <meta http-equiv="Content-type" content="text/html;charset=UTF-8">
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]6 <title>LLVM Programmer's Manual</title>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]7 <link rel="stylesheet" href="llvm.css" type="text/css">
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]8</head>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]9<body>
10
11<div class="doc_title">
12 LLVM Programmer's Manual
13</div>
14
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]15<ol>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]16 <li><a href="#introduction">Introduction</a></li>
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]17 <li><a href="#general">General Information</a>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]18 <ul>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000[diff] [blame]19 <li><a href="#stl">The C++ Standard Template Library</a></li>
20<!--
21 <li>The <tt>-time-passes</tt> option</li>
22 <li>How to use the LLVM Makefile system</li>
23 <li>How to write a regression test</li>
Chris Lattner61db4652004-12-08 19:05:44 +0000[diff] [blame]24
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000[diff] [blame]25-->
Chris Lattner84b7f8d2003-08-01 22:20:59 +0000[diff] [blame]26 </ul>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]27 </li>
28 <li><a href="#apis">Important and useful LLVM APIs</a>
29 <ul>
30 <li><a href="#isa">The <tt>isa&lt;&gt;</tt>, <tt>cast&lt;&gt;</tt>
31and <tt>dyn_cast&lt;&gt;</tt> templates</a> </li>
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]32 <li><a href="#string_apis">Passing strings (the <tt>StringRef</tt>
33and <tt>Twine</tt> classes)</li>
34 <ul>
35 <li><a href="#StringRef">The <tt>StringRef</tt> class</a> </li>
36 <li><a href="#Twine">The <tt>Twine</tt> class</a> </li>
37 </ul>
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]38 <li><a href="#DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]39option</a>
40 <ul>
41 <li><a href="#DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt>
42and the <tt>-debug-only</tt> option</a> </li>
43 </ul>
44 </li>
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000[diff] [blame]45 <li><a href="#Statistic">The <tt>Statistic</tt> class &amp; <tt>-stats</tt>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000[diff] [blame]46option</a></li>
47<!--
48 <li>The <tt>InstVisitor</tt> template
49 <li>The general graph API
50-->
Chris Lattnerf623a082005-10-17 01:36:23 +0000[diff] [blame]51 <li><a href="#ViewGraph">Viewing graphs while debugging code</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]52 </ul>
53 </li>
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]54 <li><a href="#datastructure">Picking the Right Data Structure for a Task</a>
55 <ul>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]56 <li><a href="#ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
57 <ul>
58 <li><a href="#dss_fixedarrays">Fixed Size Arrays</a></li>
59 <li><a href="#dss_heaparrays">Heap Allocated Arrays</a></li>
60 <li><a href="#dss_smallvector">"llvm/ADT/SmallVector.h"</a></li>
61 <li><a href="#dss_vector">&lt;vector&gt;</a></li>
62 <li><a href="#dss_deque">&lt;deque&gt;</a></li>
63 <li><a href="#dss_list">&lt;list&gt;</a></li>
Gabor Greif3899e492009-02-27 11:37:41 +0000[diff] [blame]64 <li><a href="#dss_ilist">llvm/ADT/ilist.h</a></li>
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]65 <li><a href="#dss_other">Other Sequential Container Options</a></li>
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]66 </ul></li>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]67 <li><a href="#ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
68 <ul>
69 <li><a href="#dss_sortedvectorset">A sorted 'vector'</a></li>
70 <li><a href="#dss_smallset">"llvm/ADT/SmallSet.h"</a></li>
71 <li><a href="#dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a></li>
Chris Lattnerc28476f2007-09-30 00:58:59 +0000[diff] [blame]72 <li><a href="#dss_denseset">"llvm/ADT/DenseSet.h"</a></li>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]73 <li><a href="#dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a></li>
74 <li><a href="#dss_set">&lt;set&gt;</a></li>
75 <li><a href="#dss_setvector">"llvm/ADT/SetVector.h"</a></li>
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]76 <li><a href="#dss_uniquevector">"llvm/ADT/UniqueVector.h"</a></li>
77 <li><a href="#dss_otherset">Other Set-Like ContainerOptions</a></li>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]78 </ul></li>
Chris Lattnerf3692522007-02-03 19:51:56 +0000[diff] [blame]79 <li><a href="#ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
80 <ul>
81 <li><a href="#dss_sortedvectormap">A sorted 'vector'</a></li>
Chris Lattner796f9fa2007-02-08 19:14:21 +0000[diff] [blame]82 <li><a href="#dss_stringmap">"llvm/ADT/StringMap.h"</a></li>
Chris Lattnerf3692522007-02-03 19:51:56 +0000[diff] [blame]83 <li><a href="#dss_indexedmap">"llvm/ADT/IndexedMap.h"</a></li>
84 <li><a href="#dss_densemap">"llvm/ADT/DenseMap.h"</a></li>
85 <li><a href="#dss_map">&lt;map&gt;</a></li>
86 <li><a href="#dss_othermap">Other Map-Like Container Options</a></li>
87 </ul></li>
Daniel Berlin1939ace2007-09-24 17:52:25 +0000[diff] [blame]88 <li><a href="#ds_bit">BitVector-like containers</a>
89 <ul>
90 <li><a href="#dss_bitvector">A dense bitvector</a></li>
91 <li><a href="#dss_sparsebitvector">A sparse bitvector</a></li>
92 </ul></li>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]93 </ul>
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]94 </li>
Chris Lattnerae7f7592002-09-06 18:31:18 +0000[diff] [blame]95 <li><a href="#common">Helpful Hints for Common Operations</a>
Chris Lattnerae7f7592002-09-06 18:31:18 +0000[diff] [blame]96 <ul>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]97 <li><a href="#inspection">Basic Inspection and Traversal Routines</a>
98 <ul>
99 <li><a href="#iterate_function">Iterating over the <tt>BasicBlock</tt>s
100in a <tt>Function</tt></a> </li>
101 <li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
102in a <tt>BasicBlock</tt></a> </li>
103 <li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
104in a <tt>Function</tt></a> </li>
105 <li><a href="#iterate_convert">Turning an iterator into a
106class pointer</a> </li>
107 <li><a href="#iterate_complex">Finding call sites: a more
108complex example</a> </li>
109 <li><a href="#calls_and_invokes">Treating calls and invokes
110the same way</a> </li>
111 <li><a href="#iterate_chains">Iterating over def-use &amp;
112use-def chains</a> </li>
Chris Lattner2e438ca2008-01-03 16:56:04 +0000[diff] [blame]113 <li><a href="#iterate_preds">Iterating over predecessors &amp;
114successors of blocks</a></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]115 </ul>
116 </li>
117 <li><a href="#simplechanges">Making simple changes</a>
118 <ul>
119 <li><a href="#schanges_creating">Creating and inserting new
120 <tt>Instruction</tt>s</a> </li>
121 <li><a href="#schanges_deleting">Deleting <tt>Instruction</tt>s</a> </li>
122 <li><a href="#schanges_replacing">Replacing an <tt>Instruction</tt>
123with another <tt>Value</tt></a> </li>
Tanya Lattnerb011c662007-06-20 18:33:15 +0000[diff] [blame]124 <li><a href="#schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a> </li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]125 </ul>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000[diff] [blame]126 </li>
Jeffrey Yasskin714257f2009-04-30 22:33:41 +0000[diff] [blame]127 <li><a href="#create_types">How to Create Types</a></li>
Chris Lattnerae7f7592002-09-06 18:31:18 +0000[diff] [blame]128<!--
129 <li>Working with the Control Flow Graph
130 <ul>
131 <li>Accessing predecessors and successors of a <tt>BasicBlock</tt>
132 <li>
133 <li>
134 </ul>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000[diff] [blame]135-->
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]136 </ul>
137 </li>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]138
Owen Anderson8bc1b3b2009-06-16 01:17:16 +0000[diff] [blame]139 <li><a href="#threading">Threads and LLVM</a>
140 <ul>
Owen Anderson1ad70e32009-06-16 18:04:19 +0000[diff] [blame]141 <li><a href="#startmultithreaded">Entering and Exiting Multithreaded Mode
142 </a></li>
Owen Anderson8bc1b3b2009-06-16 01:17:16 +0000[diff] [blame]143 <li><a href="#shutdown">Ending execution with <tt>llvm_shutdown()</tt></a></li>
144 <li><a href="#managedstatic">Lazy initialization with <tt>ManagedStatic</tt></a></li>
145 </ul>
146 </li>
147
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]148 <li><a href="#advanced">Advanced Topics</a>
149 <ul>
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]150 <li><a href="#TypeResolve">LLVM Type Resolution</a>
151 <ul>
152 <li><a href="#BuildRecType">Basic Recursive Type Construction</a></li>
153 <li><a href="#refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a></li>
154 <li><a href="#PATypeHolder">The PATypeHolder Class</a></li>
155 <li><a href="#AbstractTypeUser">The AbstractTypeUser Class</a></li>
156 </ul></li>
157
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]158 <li><a href="#SymbolTable">The <tt>ValueSymbolTable</tt> and <tt>TypeSymbolTable</tt> classes</a></li>
159 <li><a href="#UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a></li>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]160 </ul></li>
161
Joel Stanley9b96c442002-09-06 21:55:13 +0000[diff] [blame]162 <li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]163 <ul>
Reid Spencer303c4b42007-01-12 17:26:25 +0000[diff] [blame]164 <li><a href="#Type">The <tt>Type</tt> class</a> </li>
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]165 <li><a href="#Module">The <tt>Module</tt> class</a></li>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000[diff] [blame]166 <li><a href="#Value">The <tt>Value</tt> class</a>
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]167 <ul>
168 <li><a href="#User">The <tt>User</tt> class</a>
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]169 <ul>
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]170 <li><a href="#Instruction">The <tt>Instruction</tt> class</a></li>
171 <li><a href="#Constant">The <tt>Constant</tt> class</a>
172 <ul>
173 <li><a href="#GlobalValue">The <tt>GlobalValue</tt> class</a>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]174 <ul>
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]175 <li><a href="#Function">The <tt>Function</tt> class</a></li>
176 <li><a href="#GlobalVariable">The <tt>GlobalVariable</tt> class</a></li>
177 </ul>
178 </li>
179 </ul>
180 </li>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000[diff] [blame]181 </ul>
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]182 </li>
183 <li><a href="#BasicBlock">The <tt>BasicBlock</tt> class</a></li>
184 <li><a href="#Argument">The <tt>Argument</tt> class</a></li>
185 </ul>
Reid Spencerfe8f4ff2004-11-01 09:02:53 +0000[diff] [blame]186 </li>
187 </ul>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]188 </li>
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]189</ol>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]190
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]191<div class="doc_author">
192 <p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
Chris Lattner94c43592004-05-26 16:52:55 +0000[diff] [blame]193 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>,
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]194 <a href="mailto:ggreif@gmail.com">Gabor Greif</a>,
Owen Anderson8bc1b3b2009-06-16 01:17:16 +0000[diff] [blame]195 <a href="mailto:jstanley@cs.uiuc.edu">Joel Stanley</a>,
196 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a> and
197 <a href="mailto:owen@apple.com">Owen Anderson</a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]198</div>
199
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]200<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]201<div class="doc_section">
202 <a name="introduction">Introduction </a>
203</div>
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]204<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]205
206<div class="doc_text">
207
208<p>This document is meant to highlight some of the important classes and
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]209interfaces available in the LLVM source-base. This manual is not
210intended to explain what LLVM is, how it works, and what LLVM code looks
211like. It assumes that you know the basics of LLVM and are interested
212in writing transformations or otherwise analyzing or manipulating the
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]213code.</p>
214
215<p>This document should get you oriented so that you can find your
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]216way in the continuously growing source code that makes up the LLVM
217infrastructure. Note that this manual is not intended to serve as a
218replacement for reading the source code, so if you think there should be
219a method in one of these classes to do something, but it's not listed,
220check the source. Links to the <a href="/doxygen/">doxygen</a> sources
221are provided to make this as easy as possible.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]222
223<p>The first section of this document describes general information that is
224useful to know when working in the LLVM infrastructure, and the second describes
225the Core LLVM classes. In the future this manual will be extended with
226information describing how to use extension libraries, such as dominator
227information, CFG traversal routines, and useful utilities like the <tt><a
228href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.</p>
229
230</div>
231
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]232<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]233<div class="doc_section">
234 <a name="general">General Information</a>
235</div>
236<!-- *********************************************************************** -->
237
238<div class="doc_text">
239
240<p>This section contains general information that is useful if you are working
241in the LLVM source-base, but that isn't specific to any particular API.</p>
242
243</div>
244
245<!-- ======================================================================= -->
246<div class="doc_subsection">
247 <a name="stl">The C++ Standard Template Library</a>
248</div>
249
250<div class="doc_text">
251
252<p>LLVM makes heavy use of the C++ Standard Template Library (STL),
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]253perhaps much more than you are used to, or have seen before. Because of
254this, you might want to do a little background reading in the
255techniques used and capabilities of the library. There are many good
256pages that discuss the STL, and several books on the subject that you
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]257can get, so it will not be discussed in this document.</p>
258
259<p>Here are some useful links:</p>
260
261<ol>
262
263<li><a href="http://www.dinkumware.com/refxcpp.html">Dinkumware C++ Library
264reference</a> - an excellent reference for the STL and other parts of the
265standard C++ library.</li>
266
267<li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
Gabor Greif0cbcabe2009-03-12 09:47:03 +0000[diff] [blame]268O'Reilly book in the making. It has a decent Standard Library
269Reference that rivals Dinkumware's, and is unfortunately no longer free since the
270book has been published.</li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]271
272<li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
273Questions</a></li>
274
275<li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
276Contains a useful <a
277href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
278STL</a>.</li>
279
280<li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
281Page</a></li>
282
Tanya Lattner79445ba2004-12-08 18:34:56 +0000[diff] [blame]283<li><a href="http://64.78.49.204/">
Reid Spencer096603a2004-05-26 08:41:35 +0000[diff] [blame]284Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
285the book).</a></li>
286
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]287</ol>
288
289<p>You are also encouraged to take a look at the <a
290href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
291to write maintainable code more than where to put your curly braces.</p>
292
293</div>
294
295<!-- ======================================================================= -->
296<div class="doc_subsection">
297 <a name="stl">Other useful references</a>
298</div>
299
300<div class="doc_text">
301
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]302<ol>
303<li><a href="http://www.psc.edu/%7Esemke/cvs_branches.html">CVS
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]304Branch and Tag Primer</a></li>
Misha Brukmana0f71e42004-06-18 18:39:00 +0000[diff] [blame]305<li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
306static and shared libraries across platforms</a></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]307</ol>
308
309</div>
310
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]311<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]312<div class="doc_section">
313 <a name="apis">Important and useful LLVM APIs</a>
314</div>
315<!-- *********************************************************************** -->
316
317<div class="doc_text">
318
319<p>Here we highlight some LLVM APIs that are generally useful and good to
320know about when writing transformations.</p>
321
322</div>
323
324<!-- ======================================================================= -->
325<div class="doc_subsection">
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]326 <a name="isa">The <tt>isa&lt;&gt;</tt>, <tt>cast&lt;&gt;</tt> and
327 <tt>dyn_cast&lt;&gt;</tt> templates</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]328</div>
329
330<div class="doc_text">
331
332<p>The LLVM source-base makes extensive use of a custom form of RTTI.
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]333These templates have many similarities to the C++ <tt>dynamic_cast&lt;&gt;</tt>
334operator, but they don't have some drawbacks (primarily stemming from
335the fact that <tt>dynamic_cast&lt;&gt;</tt> only works on classes that
336have a v-table). Because they are used so often, you must know what they
337do and how they work. All of these templates are defined in the <a
Chris Lattner695b78b2005-04-26 22:56:16 +0000[diff] [blame]338 href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]339file (note that you very rarely have to include this file directly).</p>
340
341<dl>
342 <dt><tt>isa&lt;&gt;</tt>: </dt>
343
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]344 <dd><p>The <tt>isa&lt;&gt;</tt> operator works exactly like the Java
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]345 "<tt>instanceof</tt>" operator. It returns true or false depending on whether
346 a reference or pointer points to an instance of the specified class. This can
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]347 be very useful for constraint checking of various sorts (example below).</p>
348 </dd>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]349
350 <dt><tt>cast&lt;&gt;</tt>: </dt>
351
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]352 <dd><p>The <tt>cast&lt;&gt;</tt> operator is a "checked cast" operation. It
Chris Lattner28e6ff52008-06-20 05:03:17 +0000[diff] [blame]353 converts a pointer or reference from a base class to a derived class, causing
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]354 an assertion failure if it is not really an instance of the right type. This
355 should be used in cases where you have some information that makes you believe
356 that something is of the right type. An example of the <tt>isa&lt;&gt;</tt>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]357 and <tt>cast&lt;&gt;</tt> template is:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]358
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]359<div class="doc_code">
360<pre>
361static bool isLoopInvariant(const <a href="#Value">Value</a> *V, const Loop *L) {
362 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))
363 return true;
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]364
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]365 // <i>Otherwise, it must be an instruction...</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]366 return !L-&gt;contains(cast&lt;<a href="#Instruction">Instruction</a>&gt;(V)-&gt;getParent());
367}
368</pre>
369</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]370
371 <p>Note that you should <b>not</b> use an <tt>isa&lt;&gt;</tt> test followed
372 by a <tt>cast&lt;&gt;</tt>, for that use the <tt>dyn_cast&lt;&gt;</tt>
373 operator.</p>
374
375 </dd>
376
377 <dt><tt>dyn_cast&lt;&gt;</tt>:</dt>
378
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]379 <dd><p>The <tt>dyn_cast&lt;&gt;</tt> operator is a "checking cast" operation.
380 It checks to see if the operand is of the specified type, and if so, returns a
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]381 pointer to it (this operator does not work with references). If the operand is
382 not of the correct type, a null pointer is returned. Thus, this works very
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]383 much like the <tt>dynamic_cast&lt;&gt;</tt> operator in C++, and should be
384 used in the same circumstances. Typically, the <tt>dyn_cast&lt;&gt;</tt>
385 operator is used in an <tt>if</tt> statement or some other flow control
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]386 statement like this:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]387
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]388<div class="doc_code">
389<pre>
390if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast&lt;<a href="#AllocationInst">AllocationInst</a>&gt;(Val)) {
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]391 // <i>...</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]392}
393</pre>
394</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]395
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]396 <p>This form of the <tt>if</tt> statement effectively combines together a call
397 to <tt>isa&lt;&gt;</tt> and a call to <tt>cast&lt;&gt;</tt> into one
398 statement, which is very convenient.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]399
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]400 <p>Note that the <tt>dyn_cast&lt;&gt;</tt> operator, like C++'s
401 <tt>dynamic_cast&lt;&gt;</tt> or Java's <tt>instanceof</tt> operator, can be
402 abused. In particular, you should not use big chained <tt>if/then/else</tt>
403 blocks to check for lots of different variants of classes. If you find
404 yourself wanting to do this, it is much cleaner and more efficient to use the
405 <tt>InstVisitor</tt> class to dispatch over the instruction type directly.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]406
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]407 </dd>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]408
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]409 <dt><tt>cast_or_null&lt;&gt;</tt>: </dt>
410
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]411 <dd><p>The <tt>cast_or_null&lt;&gt;</tt> operator works just like the
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]412 <tt>cast&lt;&gt;</tt> operator, except that it allows for a null pointer as an
413 argument (which it then propagates). This can sometimes be useful, allowing
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]414 you to combine several null checks into one.</p></dd>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]415
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]416 <dt><tt>dyn_cast_or_null&lt;&gt;</tt>: </dt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]417
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]418 <dd><p>The <tt>dyn_cast_or_null&lt;&gt;</tt> operator works just like the
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]419 <tt>dyn_cast&lt;&gt;</tt> operator, except that it allows for a null pointer
420 as an argument (which it then propagates). This can sometimes be useful,
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]421 allowing you to combine several null checks into one.</p></dd>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]422
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]423</dl>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]424
425<p>These five templates can be used with any classes, whether they have a
426v-table or not. To add support for these templates, you simply need to add
427<tt>classof</tt> static methods to the class you are interested casting
428to. Describing this is currently outside the scope of this document, but there
429are lots of examples in the LLVM source base.</p>
430
431</div>
432
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]433
434<!-- ======================================================================= -->
435<div class="doc_subsection">
436 <a name="string_apis">Passing strings (the <tt>StringRef</tt>
437and <tt>Twine</tt> classes)</a>
438</div>
439
440<div class="doc_text">
441
442<p>Although LLVM generally does not do much string manipulation, we do have
Chris Lattner81187ae2009-07-25 07:16:59 +0000[diff] [blame^]443several important APIs which take strings. Two important examples are the
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]444Value class -- which has names for instructions, functions, etc. -- and the
445StringMap class which is used extensively in LLVM and Clang.</p>
446
447<p>These are generic classes, and they need to be able to accept strings which
448may have embedded null characters. Therefore, they cannot simply take
Chris Lattner81187ae2009-07-25 07:16:59 +0000[diff] [blame^]449a <tt>const char *</tt>, and taking a <tt>const std::string&amp;</tt> requires
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]450clients to perform a heap allocation which is usually unnecessary. Instead,
Chris Lattner81187ae2009-07-25 07:16:59 +0000[diff] [blame^]451many LLVM APIs use a <tt>const StringRef&amp;</tt> or a <tt>const
452Twine&amp;</tt> for passing strings efficiently.</p>
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]453
454</div>
455
456<!-- _______________________________________________________________________ -->
457<div class="doc_subsubsection">
458 <a name="StringRef">The <tt>StringRef</tt> class</a>
459</div>
460
461<div class="doc_text">
462
463<p>The <tt>StringRef</tt> data type represents a reference to a constant string
464(a character array and a length) and supports the common operations available
465on <tt>std:string</tt>, but does not require heap allocation.</p>
466
Chris Lattner81187ae2009-07-25 07:16:59 +0000[diff] [blame^]467<p>It can be implicitly constructed using a C style null-terminated string,
468an <tt>std::string</tt>, or explicitly with a character pointer and length.
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]469For example, the <tt>StringRef</tt> find function is declared as:</p>
Chris Lattner81187ae2009-07-25 07:16:59 +0000[diff] [blame^]470
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]471<div class="doc_code">
Chris Lattner81187ae2009-07-25 07:16:59 +0000[diff] [blame^]472 iterator find(const StringRef &amp;Key);
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]473</div>
474
475<p>and clients can call it using any one of:</p>
476
477<div class="doc_code">
478<pre>
479 Map.find("foo"); <i>// Lookup "foo"</i>
480 Map.find(std::string("bar")); <i>// Lookup "bar"</i>
481 Map.find(StringRef("\0baz", 4)); <i>// Lookup "\0baz"</i>
482</pre>
483</div>
484
485<p>Similarly, APIs which need to return a string may return a <tt>StringRef</tt>
486instance, which can be used directly or converted to an <tt>std::string</tt>
487using the <tt>str</tt> member function. See
488"<tt><a href="/doxygen/classllvm_1_1StringRef_8h-source.html">llvm/ADT/StringRef.h</a></tt>"
489for more information.</p>
490
491<p>You should rarely use the <tt>StringRef</tt> class directly, because it contains
492pointers to external memory it is not generally safe to store an instance of the
Chris Lattner81187ae2009-07-25 07:16:59 +0000[diff] [blame^]493class (unless you know that the external storage will not be freed).</p>
Daniel Dunbar6e0d1cb2009-07-25 04:41:11 +0000[diff] [blame]494
495</div>
496
497<!-- _______________________________________________________________________ -->
498<div class="doc_subsubsection">
499 <a name="Twine">The <tt>Twine</tt> class</a>
500</div>
501
502<div class="doc_text">
503
504<p>The <tt>Twine</tt> class is an efficient way for APIs to accept concatenated
505strings. For example, a common LLVM paradigm is to name one instruction based on
506the name of another instruction with a suffix, for example:</p>
507
508<div class="doc_code">
509<pre>
510 New = CmpInst::Create(<i>...</i>, SO->getName() + ".cmp");
511</pre>
512</div>
513
514<p>The <tt>Twine</tt> class is effectively a
515lightweight <a href="http://en.wikipedia.org/wiki/Rope_(computer_science)">rope</a>
516which points to temporary (stack allocated) objects. Twines can be implicitly
517constructed as the result of the plus operator applied to strings (i.e., a C
518strings, an <tt>std::string</tt>, or a <tt>StringRef</tt>). The twine delays the
519actual concatentation of strings until it is actually required, at which point
520it can be efficiently rendered directly into a character array. This avoids
521unnecessary heap allocation involved in constructing the temporary results of
522string concatenation. See
523"<tt><a href="/doxygen/classllvm_1_1Twine_8h-source.html">llvm/ADT/Twine.h</a></tt>"
524for more information.</p></tt>
525
526<p>As with a <tt>StringRef</tt>, <tt>Twine</tt> objects point to external memory
527and should almost never be stored or mentioned directly. They are intended
528solely for use when defining a function which should be able to efficiently
529accept concatenated strings.</p>
530
531</div>
532
533
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]534<!-- ======================================================================= -->
535<div class="doc_subsection">
Misha Brukman2c122ce2005-11-01 21:12:49 +0000[diff] [blame]536 <a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]537</div>
538
539<div class="doc_text">
540
541<p>Often when working on your pass you will put a bunch of debugging printouts
542and other code into your pass. After you get it working, you want to remove
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]543it, but you may need it again in the future (to work out new bugs that you run
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]544across).</p>
545
546<p> Naturally, because of this, you don't want to delete the debug printouts,
547but you don't want them to always be noisy. A standard compromise is to comment
548them out, allowing you to enable them if you need them in the future.</p>
549
Chris Lattner695b78b2005-04-26 22:56:16 +0000[diff] [blame]550<p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]551file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
552this problem. Basically, you can put arbitrary code into the argument of the
553<tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
554tool) is run with the '<tt>-debug</tt>' command line argument:</p>
555
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]556<div class="doc_code">
557<pre>
Daniel Dunbar06388ae2009-07-25 01:55:32 +0000[diff] [blame]558DEBUG(errs() &lt;&lt; "I am here!\n");
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]559</pre>
560</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]561
562<p>Then you can run your pass like this:</p>
563
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]564<div class="doc_code">
565<pre>
566$ opt &lt; a.bc &gt; /dev/null -mypass
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]567<i>&lt;no output&gt;</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]568$ opt &lt; a.bc &gt; /dev/null -mypass -debug
569I am here!
570</pre>
571</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]572
573<p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
574to not have to create "yet another" command line option for the debug output for
575your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
576so they do not cause a performance impact at all (for the same reason, they
577should also not contain side-effects!).</p>
578
579<p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
580enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
581"<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
582program hasn't been started yet, you can always just run it with
583<tt>-debug</tt>.</p>
584
585</div>
586
587<!-- _______________________________________________________________________ -->
588<div class="doc_subsubsection">
Chris Lattnerc9151082005-04-26 22:57:07 +0000[diff] [blame]589 <a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]590 the <tt>-debug-only</tt> option</a>
591</div>
592
593<div class="doc_text">
594
595<p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
596just turns on <b>too much</b> information (such as when working on the code
597generator). If you want to enable debug information with more fine-grained
598control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
599option as follows:</p>
600
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]601<div class="doc_code">
602<pre>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]603#undef DEBUG_TYPE
Daniel Dunbar06388ae2009-07-25 01:55:32 +0000[diff] [blame]604DEBUG(errs() &lt;&lt; "No debug type\n");
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]605#define DEBUG_TYPE "foo"
Daniel Dunbar06388ae2009-07-25 01:55:32 +0000[diff] [blame]606DEBUG(errs() &lt;&lt; "'foo' debug type\n");
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]607#undef DEBUG_TYPE
608#define DEBUG_TYPE "bar"
Daniel Dunbar06388ae2009-07-25 01:55:32 +0000[diff] [blame]609DEBUG(errs() &lt;&lt; "'bar' debug type\n"));
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]610#undef DEBUG_TYPE
611#define DEBUG_TYPE ""
Daniel Dunbar06388ae2009-07-25 01:55:32 +0000[diff] [blame]612DEBUG(errs() &lt;&lt; "No debug type (2)\n");
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]613</pre>
614</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]615
616<p>Then you can run your pass like this:</p>
617
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]618<div class="doc_code">
619<pre>
620$ opt &lt; a.bc &gt; /dev/null -mypass
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]621<i>&lt;no output&gt;</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]622$ opt &lt; a.bc &gt; /dev/null -mypass -debug
623No debug type
624'foo' debug type
625'bar' debug type
626No debug type (2)
627$ opt &lt; a.bc &gt; /dev/null -mypass -debug-only=foo
628'foo' debug type
629$ opt &lt; a.bc &gt; /dev/null -mypass -debug-only=bar
630'bar' debug type
631</pre>
632</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]633
634<p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
635a file, to specify the debug type for the entire module (if you do this before
Chris Lattner695b78b2005-04-26 22:56:16 +0000[diff] [blame]636you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]637<tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
638"bar", because there is no system in place to ensure that names do not
639conflict. If two different modules use the same string, they will all be turned
640on when the name is specified. This allows, for example, all debug information
641for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]642even if the source lives in multiple files.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]643
644</div>
645
646<!-- ======================================================================= -->
647<div class="doc_subsection">
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000[diff] [blame]648 <a name="Statistic">The <tt>Statistic</tt> class &amp; <tt>-stats</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]649 option</a>
650</div>
651
652<div class="doc_text">
653
654<p>The "<tt><a
Chris Lattner695b78b2005-04-26 22:56:16 +0000[diff] [blame]655href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000[diff] [blame]656provides a class named <tt>Statistic</tt> that is used as a unified way to
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]657keep track of what the LLVM compiler is doing and how effective various
658optimizations are. It is useful to see what optimizations are contributing to
659making a particular program run faster.</p>
660
661<p>Often you may run your pass on some big program, and you're interested to see
662how many times it makes a certain transformation. Although you can do this with
663hand inspection, or some ad-hoc method, this is a real pain and not very useful
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000[diff] [blame]664for big programs. Using the <tt>Statistic</tt> class makes it very easy to
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]665keep track of this information, and the calculated information is presented in a
666uniform manner with the rest of the passes being executed.</p>
667
668<p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
669it are as follows:</p>
670
671<ol>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]672 <li><p>Define your statistic like this:</p>
673
674<div class="doc_code">
675<pre>
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000[diff] [blame]676#define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
677STATISTIC(NumXForms, "The # of times I did stuff");
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]678</pre>
679</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]680
Chris Lattner0be6fdf2006-12-19 21:46:21 +0000[diff] [blame]681 <p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
682 specified by the first argument. The pass name is taken from the DEBUG_TYPE
683 macro, and the description is taken from the second argument. The variable
Reid Spencer06565dc2007-01-12 17:11:23 +0000[diff] [blame]684 defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]685
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]686 <li><p>Whenever you make a transformation, bump the counter:</p>
687
688<div class="doc_code">
689<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]690++NumXForms; // <i>I did stuff!</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]691</pre>
692</div>
693
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]694 </li>
695 </ol>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]696
697 <p>That's all you have to do. To get '<tt>opt</tt>' to print out the
698 statistics gathered, use the '<tt>-stats</tt>' option:</p>
699
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]700<div class="doc_code">
701<pre>
702$ opt -stats -mypassname &lt; program.bc &gt; /dev/null
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]703<i>... statistics output ...</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]704</pre>
705</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]706
Reid Spencer6b6c73e2007-02-09 16:00:28 +0000[diff] [blame]707 <p> When running <tt>opt</tt> on a C file from the SPEC benchmark
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]708suite, it gives a report that looks like this:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]709
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]710<div class="doc_code">
711<pre>
Gabor Greif04367bf2007-07-06 22:07:22 +0000[diff] [blame]712 7646 bitcodewriter - Number of normal instructions
713 725 bitcodewriter - Number of oversized instructions
714 129996 bitcodewriter - Number of bitcode bytes written
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]715 2817 raise - Number of insts DCEd or constprop'd
716 3213 raise - Number of cast-of-self removed
717 5046 raise - Number of expression trees converted
718 75 raise - Number of other getelementptr's formed
719 138 raise - Number of load/store peepholes
720 42 deadtypeelim - Number of unused typenames removed from symtab
721 392 funcresolve - Number of varargs functions resolved
722 27 globaldce - Number of global variables removed
723 2 adce - Number of basic blocks removed
724 134 cee - Number of branches revectored
725 49 cee - Number of setcc instruction eliminated
726 532 gcse - Number of loads removed
727 2919 gcse - Number of instructions removed
728 86 indvars - Number of canonical indvars added
729 87 indvars - Number of aux indvars removed
730 25 instcombine - Number of dead inst eliminate
731 434 instcombine - Number of insts combined
732 248 licm - Number of load insts hoisted
733 1298 licm - Number of insts hoisted to a loop pre-header
734 3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
735 75 mem2reg - Number of alloca's promoted
736 1444 cfgsimplify - Number of blocks simplified
737</pre>
738</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]739
740<p>Obviously, with so many optimizations, having a unified framework for this
741stuff is very nice. Making your pass fit well into the framework makes it more
742maintainable and useful.</p>
743
744</div>
745
Chris Lattnerf623a082005-10-17 01:36:23 +0000[diff] [blame]746<!-- ======================================================================= -->
747<div class="doc_subsection">
748 <a name="ViewGraph">Viewing graphs while debugging code</a>
749</div>
750
751<div class="doc_text">
752
753<p>Several of the important data structures in LLVM are graphs: for example
754CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
755LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
756<a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
757DAGs</a>. In many cases, while debugging various parts of the compiler, it is
758nice to instantly visualize these graphs.</p>
759
760<p>LLVM provides several callbacks that are available in a debug build to do
761exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
762the current LLVM tool will pop up a window containing the CFG for the function
763where each basic block is a node in the graph, and each node contains the
764instructions in the block. Similarly, there also exists
765<tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
766<tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
767and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
Jim Laskey543a0ee2006-10-02 12:28:07 +0000[diff] [blame]768you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
Chris Lattnerf623a082005-10-17 01:36:23 +0000[diff] [blame]769up a window. Alternatively, you can sprinkle calls to these functions in your
770code in places you want to debug.</p>
771
772<p>Getting this to work requires a small amount of configuration. On Unix
773systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
774toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
775Mac OS/X, download and install the Mac OS/X <a
776href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]777<tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
Chris Lattnerf623a082005-10-17 01:36:23 +0000[diff] [blame]778it) to your path. Once in your system and path are set up, rerun the LLVM
779configure script and rebuild LLVM to enable this functionality.</p>
780
Jim Laskey543a0ee2006-10-02 12:28:07 +0000[diff] [blame]781<p><tt>SelectionDAG</tt> has been extended to make it easier to locate
782<i>interesting</i> nodes in large complex graphs. From gdb, if you
783<tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]784next <tt>call DAG.viewGraph()</tt> would highlight the node in the
Jim Laskey543a0ee2006-10-02 12:28:07 +0000[diff] [blame]785specified color (choices of colors can be found at <a
Chris Lattner302da1e2007-02-03 03:05:57 +0000[diff] [blame]786href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
Jim Laskey543a0ee2006-10-02 12:28:07 +0000[diff] [blame]787complex node attributes can be provided with <tt>call
788DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
789found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
790Attributes</a>.) If you want to restart and clear all the current graph
791attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
792
Chris Lattnerf623a082005-10-17 01:36:23 +0000[diff] [blame]793</div>
794
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]795<!-- *********************************************************************** -->
796<div class="doc_section">
797 <a name="datastructure">Picking the Right Data Structure for a Task</a>
798</div>
799<!-- *********************************************************************** -->
800
801<div class="doc_text">
802
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]803<p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
804 and we commonly use STL data structures. This section describes the trade-offs
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]805 you should consider when you pick one.</p>
806
807<p>
808The first step is a choose your own adventure: do you want a sequential
809container, a set-like container, or a map-like container? The most important
810thing when choosing a container is the algorithmic properties of how you plan to
811access the container. Based on that, you should use:</p>
812
813<ul>
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]814<li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]815 of an value based on another value. Map-like containers also support
816 efficient queries for containment (whether a key is in the map). Map-like
817 containers generally do not support efficient reverse mapping (values to
818 keys). If you need that, use two maps. Some map-like containers also
819 support efficient iteration through the keys in sorted order. Map-like
820 containers are the most expensive sort, only use them if you need one of
821 these capabilities.</li>
822
823<li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
824 stuff into a container that automatically eliminates duplicates. Some
825 set-like containers support efficient iteration through the elements in
826 sorted order. Set-like containers are more expensive than sequential
827 containers.
828</li>
829
830<li>a <a href="#ds_sequential">sequential</a> container provides
831 the most efficient way to add elements and keeps track of the order they are
832 added to the collection. They permit duplicates and support efficient
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]833 iteration, but do not support efficient look-up based on a key.
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]834</li>
835
Daniel Berlin1939ace2007-09-24 17:52:25 +0000[diff] [blame]836<li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
837 perform set operations on sets of numeric id's, while automatically
838 eliminating duplicates. Bit containers require a maximum of 1 bit for each
839 identifier you want to store.
840</li>
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]841</ul>
842
843<p>
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]844Once the proper category of container is determined, you can fine tune the
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]845memory use, constant factors, and cache behaviors of access by intelligently
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]846picking a member of the category. Note that constant factors and cache behavior
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]847can be a big deal. If you have a vector that usually only contains a few
848elements (but could contain many), for example, it's much better to use
849<a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
850. Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
851cost of adding the elements to the container. </p>
852
853</div>
854
855<!-- ======================================================================= -->
856<div class="doc_subsection">
857 <a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
858</div>
859
860<div class="doc_text">
861There are a variety of sequential containers available for you, based on your
862needs. Pick the first in this section that will do what you want.
863</div>
864
865<!-- _______________________________________________________________________ -->
866<div class="doc_subsubsection">
867 <a name="dss_fixedarrays">Fixed Size Arrays</a>
868</div>
869
870<div class="doc_text">
871<p>Fixed size arrays are very simple and very fast. They are good if you know
872exactly how many elements you have, or you have a (low) upper bound on how many
873you have.</p>
874</div>
875
876<!-- _______________________________________________________________________ -->
877<div class="doc_subsubsection">
878 <a name="dss_heaparrays">Heap Allocated Arrays</a>
879</div>
880
881<div class="doc_text">
882<p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
883the number of elements is variable, if you know how many elements you will need
884before the array is allocated, and if the array is usually large (if not,
885consider a <a href="#dss_smallvector">SmallVector</a>). The cost of a heap
886allocated array is the cost of the new/delete (aka malloc/free). Also note that
887if you are allocating an array of a type with a constructor, the constructor and
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]888destructors will be run for every element in the array (re-sizable vectors only
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]889construct those elements actually used).</p>
890</div>
891
892<!-- _______________________________________________________________________ -->
893<div class="doc_subsubsection">
894 <a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
895</div>
896
897<div class="doc_text">
898<p><tt>SmallVector&lt;Type, N&gt;</tt> is a simple class that looks and smells
899just like <tt>vector&lt;Type&gt;</tt>:
900it supports efficient iteration, lays out elements in memory order (so you can
901do pointer arithmetic between elements), supports efficient push_back/pop_back
902operations, supports efficient random access to its elements, etc.</p>
903
904<p>The advantage of SmallVector is that it allocates space for
905some number of elements (N) <b>in the object itself</b>. Because of this, if
906the SmallVector is dynamically smaller than N, no malloc is performed. This can
907be a big win in cases where the malloc/free call is far more expensive than the
908code that fiddles around with the elements.</p>
909
910<p>This is good for vectors that are "usually small" (e.g. the number of
911predecessors/successors of a block is usually less than 8). On the other hand,
912this makes the size of the SmallVector itself large, so you don't want to
913allocate lots of them (doing so will waste a lot of space). As such,
914SmallVectors are most useful when on the stack.</p>
915
916<p>SmallVector also provides a nice portable and efficient replacement for
917<tt>alloca</tt>.</p>
918
919</div>
920
921<!-- _______________________________________________________________________ -->
922<div class="doc_subsubsection">
923 <a name="dss_vector">&lt;vector&gt;</a>
924</div>
925
926<div class="doc_text">
927<p>
928std::vector is well loved and respected. It is useful when SmallVector isn't:
929when the size of the vector is often large (thus the small optimization will
930rarely be a benefit) or if you will be allocating many instances of the vector
931itself (which would waste space for elements that aren't in the container).
932vector is also useful when interfacing with code that expects vectors :).
933</p>
Chris Lattner32d84762007-02-05 06:30:51 +0000[diff] [blame]934
935<p>One worthwhile note about std::vector: avoid code like this:</p>
936
937<div class="doc_code">
938<pre>
939for ( ... ) {
Chris Lattner9bb3dbb2007-03-28 18:27:57 +0000[diff] [blame]940 std::vector&lt;foo&gt; V;
Chris Lattner32d84762007-02-05 06:30:51 +0000[diff] [blame]941 use V;
942}
943</pre>
944</div>
945
946<p>Instead, write this as:</p>
947
948<div class="doc_code">
949<pre>
Chris Lattner9bb3dbb2007-03-28 18:27:57 +0000[diff] [blame]950std::vector&lt;foo&gt; V;
Chris Lattner32d84762007-02-05 06:30:51 +0000[diff] [blame]951for ( ... ) {
952 use V;
953 V.clear();
954}
955</pre>
956</div>
957
958<p>Doing so will save (at least) one heap allocation and free per iteration of
959the loop.</p>
960
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]961</div>
962
963<!-- _______________________________________________________________________ -->
964<div class="doc_subsubsection">
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]965 <a name="dss_deque">&lt;deque&gt;</a>
966</div>
967
968<div class="doc_text">
969<p>std::deque is, in some senses, a generalized version of std::vector. Like
970std::vector, it provides constant time random access and other similar
971properties, but it also provides efficient access to the front of the list. It
972does not guarantee continuity of elements within memory.</p>
973
974<p>In exchange for this extra flexibility, std::deque has significantly higher
975constant factor costs than std::vector. If possible, use std::vector or
976something cheaper.</p>
977</div>
978
979<!-- _______________________________________________________________________ -->
980<div class="doc_subsubsection">
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]981 <a name="dss_list">&lt;list&gt;</a>
982</div>
983
984<div class="doc_text">
985<p>std::list is an extremely inefficient class that is rarely useful.
986It performs a heap allocation for every element inserted into it, thus having an
987extremely high constant factor, particularly for small data types. std::list
988also only supports bidirectional iteration, not random access iteration.</p>
989
990<p>In exchange for this high cost, std::list supports efficient access to both
991ends of the list (like std::deque, but unlike std::vector or SmallVector). In
992addition, the iterator invalidation characteristics of std::list are stronger
993than that of a vector class: inserting or removing an element into the list does
994not invalidate iterator or pointers to other elements in the list.</p>
995</div>
996
997<!-- _______________________________________________________________________ -->
998<div class="doc_subsubsection">
Gabor Greif3899e492009-02-27 11:37:41 +0000[diff] [blame]999 <a name="dss_ilist">llvm/ADT/ilist.h</a>
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]1000</div>
1001
1002<div class="doc_text">
1003<p><tt>ilist&lt;T&gt;</tt> implements an 'intrusive' doubly-linked list. It is
1004intrusive, because it requires the element to store and provide access to the
1005prev/next pointers for the list.</p>
1006
Gabor Greif2946d1c2009-02-27 12:02:19 +0000[diff] [blame]1007<p><tt>ilist</tt> has the same drawbacks as <tt>std::list</tt>, and additionally
1008requires an <tt>ilist_traits</tt> implementation for the element type, but it
1009provides some novel characteristics. In particular, it can efficiently store
1010polymorphic objects, the traits class is informed when an element is inserted or
Gabor Greif0cbcabe2009-03-12 09:47:03 +0000[diff] [blame]1011removed from the list, and <tt>ilist</tt>s are guaranteed to support a
1012constant-time splice operation.</p>
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]1013
Gabor Greif0cbcabe2009-03-12 09:47:03 +0000[diff] [blame]1014<p>These properties are exactly what we want for things like
1015<tt>Instruction</tt>s and basic blocks, which is why these are implemented with
1016<tt>ilist</tt>s.</p>
Gabor Greif3899e492009-02-27 11:37:41 +0000[diff] [blame]1017
1018Related classes of interest are explained in the following subsections:
1019 <ul>
Gabor Greif01862502009-02-27 13:28:07 +0000[diff] [blame]1020 <li><a href="#dss_ilist_traits">ilist_traits</a></li>
Gabor Greif2946d1c2009-02-27 12:02:19 +0000[diff] [blame]1021 <li><a href="#dss_iplist">iplist</a></li>
Gabor Greif3899e492009-02-27 11:37:41 +0000[diff] [blame]1022 <li><a href="#dss_ilist_node">llvm/ADT/ilist_node.h</a></li>
Gabor Greif6a65f422009-03-12 10:30:31 +0000[diff] [blame]1023 <li><a href="#dss_ilist_sentinel">Sentinels</a></li>
Gabor Greif3899e492009-02-27 11:37:41 +0000[diff] [blame]1024 </ul>
1025</div>
1026
1027<!-- _______________________________________________________________________ -->
1028<div class="doc_subsubsection">
Gabor Greif01862502009-02-27 13:28:07 +0000[diff] [blame]1029 <a name="dss_ilist_traits">ilist_traits</a>
1030</div>
1031
1032<div class="doc_text">
1033<p><tt>ilist_traits&lt;T&gt;</tt> is <tt>ilist&lt;T&gt;</tt>'s customization
1034mechanism. <tt>iplist&lt;T&gt;</tt> (and consequently <tt>ilist&lt;T&gt;</tt>)
1035publicly derive from this traits class.</p>
1036</div>
1037
1038<!-- _______________________________________________________________________ -->
1039<div class="doc_subsubsection">
Gabor Greif2946d1c2009-02-27 12:02:19 +0000[diff] [blame]1040 <a name="dss_iplist">iplist</a>
1041</div>
1042
1043<div class="doc_text">
1044<p><tt>iplist&lt;T&gt;</tt> is <tt>ilist&lt;T&gt;</tt>'s base and as such
Gabor Greif0cbcabe2009-03-12 09:47:03 +0000[diff] [blame]1045supports a slightly narrower interface. Notably, inserters from
1046<tt>T&amp;</tt> are absent.</p>
Gabor Greif01862502009-02-27 13:28:07 +0000[diff] [blame]1047
1048<p><tt>ilist_traits&lt;T&gt;</tt> is a public base of this class and can be
1049used for a wide variety of customizations.</p>
Gabor Greif2946d1c2009-02-27 12:02:19 +0000[diff] [blame]1050</div>
1051
1052<!-- _______________________________________________________________________ -->
1053<div class="doc_subsubsection">
Gabor Greif3899e492009-02-27 11:37:41 +0000[diff] [blame]1054 <a name="dss_ilist_node">llvm/ADT/ilist_node.h</a>
1055</div>
1056
1057<div class="doc_text">
1058<p><tt>ilist_node&lt;T&gt;</tt> implements a the forward and backward links
1059that are expected by the <tt>ilist&lt;T&gt;</tt> (and analogous containers)
1060in the default manner.</p>
1061
1062<p><tt>ilist_node&lt;T&gt;</tt>s are meant to be embedded in the node type
Gabor Greif0cbcabe2009-03-12 09:47:03 +0000[diff] [blame]1063<tt>T</tt>, usually <tt>T</tt> publicly derives from
1064<tt>ilist_node&lt;T&gt;</tt>.</p>
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]1065</div>
1066
1067<!-- _______________________________________________________________________ -->
1068<div class="doc_subsubsection">
Gabor Greif6a65f422009-03-12 10:30:31 +0000[diff] [blame]1069 <a name="dss_ilist_sentinel">Sentinels</a>
1070</div>
1071
1072<div class="doc_text">
1073<p><tt>ilist</tt>s have another speciality that must be considered. To be a good
1074citizen in the C++ ecosystem, it needs to support the standard container
1075operations, such as <tt>begin</tt> and <tt>end</tt> iterators, etc. Also, the
1076<tt>operator--</tt> must work correctly on the <tt>end</tt> iterator in the
1077case of non-empty <tt>ilist</tt>s.</p>
1078
1079<p>The only sensible solution to this problem is to allocate a so-called
1080<i>sentinel</i> along with the intrusive list, which serves as the <tt>end</tt>
1081iterator, providing the back-link to the last element. However conforming to the
1082C++ convention it is illegal to <tt>operator++</tt> beyond the sentinel and it
1083also must not be dereferenced.</p>
1084
1085<p>These constraints allow for some implementation freedom to the <tt>ilist</tt>
1086how to allocate and store the sentinel. The corresponding policy is dictated
1087by <tt>ilist_traits&lt;T&gt;</tt>. By default a <tt>T</tt> gets heap-allocated
1088whenever the need for a sentinel arises.</p>
1089
1090<p>While the default policy is sufficient in most cases, it may break down when
1091<tt>T</tt> does not provide a default constructor. Also, in the case of many
1092instances of <tt>ilist</tt>s, the memory overhead of the associated sentinels
1093is wasted. To alleviate the situation with numerous and voluminous
1094<tt>T</tt>-sentinels, sometimes a trick is employed, leading to <i>ghostly
1095sentinels</i>.</p>
1096
1097<p>Ghostly sentinels are obtained by specially-crafted <tt>ilist_traits&lt;T&gt;</tt>
1098which superpose the sentinel with the <tt>ilist</tt> instance in memory. Pointer
1099arithmetic is used to obtain the sentinel, which is relative to the
1100<tt>ilist</tt>'s <tt>this</tt> pointer. The <tt>ilist</tt> is augmented by an
1101extra pointer, which serves as the back-link of the sentinel. This is the only
1102field in the ghostly sentinel which can be legally accessed.</p>
1103</div>
1104
1105<!-- _______________________________________________________________________ -->
1106<div class="doc_subsubsection">
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1107 <a name="dss_other">Other Sequential Container options</a>
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]1108</div>
1109
1110<div class="doc_text">
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1111<p>Other STL containers are available, such as std::string.</p>
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]1112
1113<p>There are also various STL adapter classes such as std::queue,
1114std::priority_queue, std::stack, etc. These provide simplified access to an
1115underlying container but don't affect the cost of the container itself.</p>
1116
1117</div>
1118
1119
1120<!-- ======================================================================= -->
1121<div class="doc_subsection">
1122 <a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
1123</div>
1124
1125<div class="doc_text">
1126
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1127<p>Set-like containers are useful when you need to canonicalize multiple values
1128into a single representation. There are several different choices for how to do
1129this, providing various trade-offs.</p>
1130
1131</div>
1132
1133
1134<!-- _______________________________________________________________________ -->
1135<div class="doc_subsubsection">
1136 <a name="dss_sortedvectorset">A sorted 'vector'</a>
1137</div>
1138
1139<div class="doc_text">
1140
Chris Lattner3b23a8c2007-02-03 08:10:45 +0000[diff] [blame]1141<p>If you intend to insert a lot of elements, then do a lot of queries, a
1142great approach is to use a vector (or other sequential container) with
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1143std::sort+std::unique to remove duplicates. This approach works really well if
Chris Lattner3b23a8c2007-02-03 08:10:45 +0000[diff] [blame]1144your usage pattern has these two distinct phases (insert then query), and can be
1145coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
1146</p>
1147
1148<p>
1149This combination provides the several nice properties: the result data is
1150contiguous in memory (good for cache locality), has few allocations, is easy to
1151address (iterators in the final vector are just indices or pointers), and can be
1152efficiently queried with a standard binary or radix search.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1153
1154</div>
1155
1156<!-- _______________________________________________________________________ -->
1157<div class="doc_subsubsection">
1158 <a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
1159</div>
1160
1161<div class="doc_text">
1162
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]1163<p>If you have a set-like data structure that is usually small and whose elements
Chris Lattner4ddfac12007-02-03 07:59:51 +0000[diff] [blame]1164are reasonably small, a <tt>SmallSet&lt;Type, N&gt;</tt> is a good choice. This set
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1165has space for N elements in place (thus, if the set is dynamically smaller than
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1166N, no malloc traffic is required) and accesses them with a simple linear search.
1167When the set grows beyond 'N' elements, it allocates a more expensive representation that
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1168guarantees efficient access (for most types, it falls back to std::set, but for
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1169pointers it uses something far better, <a
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1170href="#dss_smallptrset">SmallPtrSet</a>).</p>
1171
1172<p>The magic of this class is that it handles small sets extremely efficiently,
1173but gracefully handles extremely large sets without loss of efficiency. The
1174drawback is that the interface is quite small: it supports insertion, queries
1175and erasing, but does not support iteration.</p>
1176
1177</div>
1178
1179<!-- _______________________________________________________________________ -->
1180<div class="doc_subsubsection">
1181 <a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
1182</div>
1183
1184<div class="doc_text">
1185
1186<p>SmallPtrSet has all the advantages of SmallSet (and a SmallSet of pointers is
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]1187transparently implemented with a SmallPtrSet), but also supports iterators. If
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1188more than 'N' insertions are performed, a single quadratically
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1189probed hash table is allocated and grows as needed, providing extremely
1190efficient access (constant time insertion/deleting/queries with low constant
1191factors) and is very stingy with malloc traffic.</p>
1192
1193<p>Note that, unlike std::set, the iterators of SmallPtrSet are invalidated
1194whenever an insertion occurs. Also, the values visited by the iterators are not
1195visited in sorted order.</p>
1196
1197</div>
1198
1199<!-- _______________________________________________________________________ -->
1200<div class="doc_subsubsection">
Chris Lattnerc28476f2007-09-30 00:58:59 +0000[diff] [blame]1201 <a name="dss_denseset">"llvm/ADT/DenseSet.h"</a>
1202</div>
1203
1204<div class="doc_text">
1205
1206<p>
1207DenseSet is a simple quadratically probed hash table. It excels at supporting
1208small values: it uses a single allocation to hold all of the pairs that
1209are currently inserted in the set. DenseSet is a great way to unique small
1210values that are not simple pointers (use <a
1211href="#dss_smallptrset">SmallPtrSet</a> for pointers). Note that DenseSet has
1212the same requirements for the value type that <a
1213href="#dss_densemap">DenseMap</a> has.
1214</p>
1215
1216</div>
1217
1218<!-- _______________________________________________________________________ -->
1219<div class="doc_subsubsection">
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1220 <a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
1221</div>
1222
1223<div class="doc_text">
1224
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]1225<p>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1226FoldingSet is an aggregate class that is really good at uniquing
1227expensive-to-create or polymorphic objects. It is a combination of a chained
1228hash table with intrusive links (uniqued objects are required to inherit from
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1229FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
1230its ID process.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1231
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1232<p>Consider a case where you want to implement a "getOrCreateFoo" method for
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1233a complex object (for example, a node in the code generator). The client has a
1234description of *what* it wants to generate (it knows the opcode and all the
1235operands), but we don't want to 'new' a node, then try inserting it into a set
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1236only to find out it already exists, at which point we would have to delete it
1237and return the node that already exists.
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]1238</p>
1239
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1240<p>To support this style of client, FoldingSet perform a query with a
1241FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
1242element that we want to query for. The query either returns the element
1243matching the ID or it returns an opaque ID that indicates where insertion should
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1244take place. Construction of the ID usually does not require heap traffic.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1245
1246<p>Because FoldingSet uses intrusive links, it can support polymorphic objects
1247in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
1248Because the elements are individually allocated, pointers to the elements are
1249stable: inserting or removing elements does not invalidate any pointers to other
1250elements.
1251</p>
1252
1253</div>
1254
1255<!-- _______________________________________________________________________ -->
1256<div class="doc_subsubsection">
1257 <a name="dss_set">&lt;set&gt;</a>
1258</div>
1259
1260<div class="doc_text">
1261
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1262<p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
1263many things but great at nothing. std::set allocates memory for each element
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1264inserted (thus it is very malloc intensive) and typically stores three pointers
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1265per element in the set (thus adding a large amount of per-element space
1266overhead). It offers guaranteed log(n) performance, which is not particularly
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1267fast from a complexity standpoint (particularly if the elements of the set are
1268expensive to compare, like strings), and has extremely high constant factors for
1269lookup, insertion and removal.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1270
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1271<p>The advantages of std::set are that its iterators are stable (deleting or
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1272inserting an element from the set does not affect iterators or pointers to other
1273elements) and that iteration over the set is guaranteed to be in sorted order.
1274If the elements in the set are large, then the relative overhead of the pointers
1275and malloc traffic is not a big deal, but if the elements of the set are small,
1276std::set is almost never a good choice.</p>
1277
1278</div>
1279
1280<!-- _______________________________________________________________________ -->
1281<div class="doc_subsubsection">
1282 <a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
1283</div>
1284
1285<div class="doc_text">
Chris Lattneredca3c52007-02-04 00:00:26 +0000[diff] [blame]1286<p>LLVM's SetVector&lt;Type&gt; is an adapter class that combines your choice of
1287a set-like container along with a <a href="#ds_sequential">Sequential
1288Container</a>. The important property
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1289that this provides is efficient insertion with uniquing (duplicate elements are
1290ignored) with iteration support. It implements this by inserting elements into
1291both a set-like container and the sequential container, using the set-like
1292container for uniquing and the sequential container for iteration.
1293</p>
1294
1295<p>The difference between SetVector and other sets is that the order of
1296iteration is guaranteed to match the order of insertion into the SetVector.
1297This property is really important for things like sets of pointers. Because
1298pointer values are non-deterministic (e.g. vary across runs of the program on
Chris Lattneredca3c52007-02-04 00:00:26 +0000[diff] [blame]1299different machines), iterating over the pointers in the set will
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1300not be in a well-defined order.</p>
1301
1302<p>
1303The drawback of SetVector is that it requires twice as much space as a normal
1304set and has the sum of constant factors from the set-like container and the
1305sequential container that it uses. Use it *only* if you need to iterate over
1306the elements in a deterministic order. SetVector is also expensive to delete
Chris Lattneredca3c52007-02-04 00:00:26 +0000[diff] [blame]1307elements out of (linear time), unless you use it's "pop_back" method, which is
1308faster.
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1309</p>
1310
Chris Lattneredca3c52007-02-04 00:00:26 +0000[diff] [blame]1311<p>SetVector is an adapter class that defaults to using std::vector and std::set
1312for the underlying containers, so it is quite expensive. However,
1313<tt>"llvm/ADT/SetVector.h"</tt> also provides a SmallSetVector class, which
1314defaults to using a SmallVector and SmallSet of a specified size. If you use
1315this, and if your sets are dynamically smaller than N, you will save a lot of
1316heap traffic.</p>
1317
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1318</div>
1319
1320<!-- _______________________________________________________________________ -->
1321<div class="doc_subsubsection">
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1322 <a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
1323</div>
1324
1325<div class="doc_text">
1326
1327<p>
1328UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
1329retains a unique ID for each element inserted into the set. It internally
1330contains a map and a vector, and it assigns a unique ID for each value inserted
1331into the set.</p>
1332
1333<p>UniqueVector is very expensive: its cost is the sum of the cost of
1334maintaining both the map and vector, it has high complexity, high constant
1335factors, and produces a lot of malloc traffic. It should be avoided.</p>
1336
1337</div>
1338
1339
1340<!-- _______________________________________________________________________ -->
1341<div class="doc_subsubsection">
1342 <a name="dss_otherset">Other Set-Like Container Options</a>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1343</div>
1344
1345<div class="doc_text">
1346
1347<p>
1348The STL provides several other options, such as std::multiset and the various
Chris Lattnerf1a30822009-03-09 05:20:45 +0000[diff] [blame]1349"hash_set" like containers (whether from C++ TR1 or from the SGI library). We
1350never use hash_set and unordered_set because they are generally very expensive
1351(each insertion requires a malloc) and very non-portable.
1352</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1353
1354<p>std::multiset is useful if you're not interested in elimination of
Chris Lattner14868db2007-02-03 08:20:15 +0000[diff] [blame]1355duplicates, but has all the drawbacks of std::set. A sorted vector (where you
1356don't delete duplicate entries) or some other approach is almost always
1357better.</p>
Chris Lattner74c4ca12007-02-03 07:59:07 +0000[diff] [blame]1358
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]1359</div>
1360
1361<!-- ======================================================================= -->
1362<div class="doc_subsection">
1363 <a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
1364</div>
1365
1366<div class="doc_text">
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1367Map-like containers are useful when you want to associate data to a key. As
1368usual, there are a lot of different ways to do this. :)
1369</div>
1370
1371<!-- _______________________________________________________________________ -->
1372<div class="doc_subsubsection">
1373 <a name="dss_sortedvectormap">A sorted 'vector'</a>
1374</div>
1375
1376<div class="doc_text">
1377
1378<p>
1379If your usage pattern follows a strict insert-then-query approach, you can
1380trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
1381for set-like containers</a>. The only difference is that your query function
1382(which uses std::lower_bound to get efficient log(n) lookup) should only compare
1383the key, not both the key and value. This yields the same advantages as sorted
1384vectors for sets.
1385</p>
1386</div>
1387
1388<!-- _______________________________________________________________________ -->
1389<div class="doc_subsubsection">
Chris Lattner796f9fa2007-02-08 19:14:21 +0000[diff] [blame]1390 <a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1391</div>
1392
1393<div class="doc_text">
1394
1395<p>
1396Strings are commonly used as keys in maps, and they are difficult to support
1397efficiently: they are variable length, inefficient to hash and compare when
Chris Lattner796f9fa2007-02-08 19:14:21 +0000[diff] [blame]1398long, expensive to copy, etc. StringMap is a specialized container designed to
1399cope with these issues. It supports mapping an arbitrary range of bytes to an
1400arbitrary other object.</p>
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1401
Chris Lattner796f9fa2007-02-08 19:14:21 +0000[diff] [blame]1402<p>The StringMap implementation uses a quadratically-probed hash table, where
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1403the buckets store a pointer to the heap allocated entries (and some other
1404stuff). The entries in the map must be heap allocated because the strings are
1405variable length. The string data (key) and the element object (value) are
1406stored in the same allocation with the string data immediately after the element
1407object. This container guarantees the "<tt>(char*)(&amp;Value+1)</tt>" points
1408to the key string for a value.</p>
1409
Chris Lattner796f9fa2007-02-08 19:14:21 +0000[diff] [blame]1410<p>The StringMap is very fast for several reasons: quadratic probing is very
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1411cache efficient for lookups, the hash value of strings in buckets is not
Chris Lattner796f9fa2007-02-08 19:14:21 +0000[diff] [blame]1412recomputed when lookup up an element, StringMap rarely has to touch the
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1413memory for unrelated objects when looking up a value (even when hash collisions
1414happen), hash table growth does not recompute the hash values for strings
1415already in the table, and each pair in the map is store in a single allocation
1416(the string data is stored in the same allocation as the Value of a pair).</p>
1417
Chris Lattner796f9fa2007-02-08 19:14:21 +0000[diff] [blame]1418<p>StringMap also provides query methods that take byte ranges, so it only ever
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1419copies a string if a value is inserted into the table.</p>
1420</div>
1421
1422<!-- _______________________________________________________________________ -->
1423<div class="doc_subsubsection">
1424 <a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
1425</div>
1426
1427<div class="doc_text">
1428<p>
1429IndexedMap is a specialized container for mapping small dense integers (or
1430values that can be mapped to small dense integers) to some other type. It is
1431internally implemented as a vector with a mapping function that maps the keys to
1432the dense integer range.
1433</p>
1434
1435<p>
1436This is useful for cases like virtual registers in the LLVM code generator: they
1437have a dense mapping that is offset by a compile-time constant (the first
1438virtual register ID).</p>
1439
1440</div>
1441
1442<!-- _______________________________________________________________________ -->
1443<div class="doc_subsubsection">
1444 <a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
1445</div>
1446
1447<div class="doc_text">
1448
1449<p>
1450DenseMap is a simple quadratically probed hash table. It excels at supporting
1451small keys and values: it uses a single allocation to hold all of the pairs that
1452are currently inserted in the map. DenseMap is a great way to map pointers to
1453pointers, or map other small types to each other.
1454</p>
1455
1456<p>
1457There are several aspects of DenseMap that you should be aware of, however. The
1458iterators in a densemap are invalidated whenever an insertion occurs, unlike
1459map. Also, because DenseMap allocates space for a large number of key/value
Chris Lattnera4a264d2007-02-03 20:17:53 +0000[diff] [blame]1460pairs (it starts with 64 by default), it will waste a lot of space if your keys
1461or values are large. Finally, you must implement a partial specialization of
Chris Lattner76c1b972007-09-17 18:34:04 +0000[diff] [blame]1462DenseMapInfo for the key that you want, if it isn't already supported. This
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1463is required to tell DenseMap about two special marker values (which can never be
Chris Lattnera4a264d2007-02-03 20:17:53 +0000[diff] [blame]1464inserted into the map) that it needs internally.</p>
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1465
1466</div>
1467
1468<!-- _______________________________________________________________________ -->
1469<div class="doc_subsubsection">
1470 <a name="dss_map">&lt;map&gt;</a>
1471</div>
1472
1473<div class="doc_text">
1474
1475<p>
1476std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
1477a single allocation per pair inserted into the map, it offers log(n) lookup with
1478an extremely large constant factor, imposes a space penalty of 3 pointers per
1479pair in the map, etc.</p>
1480
1481<p>std::map is most useful when your keys or values are very large, if you need
1482to iterate over the collection in sorted order, or if you need stable iterators
1483into the map (i.e. they don't get invalidated if an insertion or deletion of
1484another element takes place).</p>
1485
1486</div>
1487
1488<!-- _______________________________________________________________________ -->
1489<div class="doc_subsubsection">
1490 <a name="dss_othermap">Other Map-Like Container Options</a>
1491</div>
1492
1493<div class="doc_text">
1494
1495<p>
1496The STL provides several other options, such as std::multimap and the various
Chris Lattnerf1a30822009-03-09 05:20:45 +0000[diff] [blame]1497"hash_map" like containers (whether from C++ TR1 or from the SGI library). We
1498never use hash_set and unordered_set because they are generally very expensive
1499(each insertion requires a malloc) and very non-portable.</p>
Chris Lattnerc5722432007-02-03 19:49:31 +0000[diff] [blame]1500
1501<p>std::multimap is useful if you want to map a key to multiple values, but has
1502all the drawbacks of std::map. A sorted vector or some other approach is almost
1503always better.</p>
1504
Chris Lattner098129a2007-02-03 03:04:03 +0000[diff] [blame]1505</div>
1506
Daniel Berlin1939ace2007-09-24 17:52:25 +0000[diff] [blame]1507<!-- ======================================================================= -->
1508<div class="doc_subsection">
1509 <a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
1510</div>
1511
1512<div class="doc_text">
Chris Lattner7086ce72007-09-25 22:37:50 +0000[diff] [blame]1513<p>Unlike the other containers, there are only two bit storage containers, and
1514choosing when to use each is relatively straightforward.</p>
1515
1516<p>One additional option is
1517<tt>std::vector&lt;bool&gt;</tt>: we discourage its use for two reasons 1) the
1518implementation in many common compilers (e.g. commonly available versions of
1519GCC) is extremely inefficient and 2) the C++ standards committee is likely to
1520deprecate this container and/or change it significantly somehow. In any case,
1521please don't use it.</p>
Daniel Berlin1939ace2007-09-24 17:52:25 +0000[diff] [blame]1522</div>
1523
1524<!-- _______________________________________________________________________ -->
1525<div class="doc_subsubsection">
1526 <a name="dss_bitvector">BitVector</a>
1527</div>
1528
1529<div class="doc_text">
1530<p> The BitVector container provides a fixed size set of bits for manipulation.
1531It supports individual bit setting/testing, as well as set operations. The set
1532operations take time O(size of bitvector), but operations are performed one word
1533at a time, instead of one bit at a time. This makes the BitVector very fast for
1534set operations compared to other containers. Use the BitVector when you expect
1535the number of set bits to be high (IE a dense set).
1536</p>
1537</div>
1538
1539<!-- _______________________________________________________________________ -->
1540<div class="doc_subsubsection">
1541 <a name="dss_sparsebitvector">SparseBitVector</a>
1542</div>
1543
1544<div class="doc_text">
1545<p> The SparseBitVector container is much like BitVector, with one major
1546difference: Only the bits that are set, are stored. This makes the
1547SparseBitVector much more space efficient than BitVector when the set is sparse,
1548as well as making set operations O(number of set bits) instead of O(size of
1549universe). 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
1550(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).
1551</p>
1552</div>
Chris Lattnerf623a082005-10-17 01:36:23 +0000[diff] [blame]1553
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1554<!-- *********************************************************************** -->
1555<div class="doc_section">
1556 <a name="common">Helpful Hints for Common Operations</a>
1557</div>
1558<!-- *********************************************************************** -->
1559
1560<div class="doc_text">
1561
1562<p>This section describes how to perform some very simple transformations of
1563LLVM code. This is meant to give examples of common idioms used, showing the
1564practical side of LLVM transformations. <p> Because this is a "how-to" section,
1565you should also read about the main classes that you will be working with. The
1566<a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
1567and descriptions of the main classes that you should know about.</p>
1568
1569</div>
1570
1571<!-- NOTE: this section should be heavy on example code -->
1572<!-- ======================================================================= -->
1573<div class="doc_subsection">
1574 <a name="inspection">Basic Inspection and Traversal Routines</a>
1575</div>
1576
1577<div class="doc_text">
1578
1579<p>The LLVM compiler infrastructure have many different data structures that may
1580be traversed. Following the example of the C++ standard template library, the
1581techniques used to traverse these various data structures are all basically the
1582same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
1583method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
1584function returns an iterator pointing to one past the last valid element of the
1585sequence, and there is some <tt>XXXiterator</tt> data type that is common
1586between the two operations.</p>
1587
1588<p>Because the pattern for iteration is common across many different aspects of
1589the program representation, the standard template library algorithms may be used
1590on them, and it is easier to remember how to iterate. First we show a few common
1591examples of the data structures that need to be traversed. Other data
1592structures are traversed in very similar ways.</p>
1593
1594</div>
1595
1596<!-- _______________________________________________________________________ -->
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1597<div class="doc_subsubsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1598 <a name="iterate_function">Iterating over the </a><a
1599 href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
1600 href="#Function"><tt>Function</tt></a>
1601</div>
1602
1603<div class="doc_text">
1604
1605<p>It's quite common to have a <tt>Function</tt> instance that you'd like to
1606transform in some way; in particular, you'd like to manipulate its
1607<tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
1608the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
1609an example that prints the name of a <tt>BasicBlock</tt> and the number of
1610<tt>Instruction</tt>s it contains:</p>
1611
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1612<div class="doc_code">
1613<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]1614// <i>func is a pointer to a Function instance</i>
1615for (Function::iterator i = func-&gt;begin(), e = func-&gt;end(); i != e; ++i)
1616 // <i>Print out the name of the basic block if it has one, and then the</i>
1617 // <i>number of instructions that it contains</i>
Bill Wendling832171c2006-12-07 20:04:42 +0000[diff] [blame]1618 llvm::cerr &lt;&lt; "Basic block (name=" &lt;&lt; i-&gt;getName() &lt;&lt; ") has "
1619 &lt;&lt; i-&gt;size() &lt;&lt; " instructions.\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1620</pre>
1621</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1622
1623<p>Note that i can be used as if it were a pointer for the purposes of
Joel Stanley9b96c442002-09-06 21:55:13 +0000[diff] [blame]1624invoking member functions of the <tt>Instruction</tt> class. This is
1625because the indirection operator is overloaded for the iterator
Chris Lattner7496ec52003-08-05 22:54:23 +0000[diff] [blame]1626classes. In the above code, the expression <tt>i-&gt;size()</tt> is
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1627exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
1628
1629</div>
1630
1631<!-- _______________________________________________________________________ -->
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1632<div class="doc_subsubsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1633 <a name="iterate_basicblock">Iterating over the </a><a
1634 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1635 href="#BasicBlock"><tt>BasicBlock</tt></a>
1636</div>
1637
1638<div class="doc_text">
1639
1640<p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
1641easy to iterate over the individual instructions that make up
1642<tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
1643a <tt>BasicBlock</tt>:</p>
1644
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1645<div class="doc_code">
Chris Lattner55c04612005-03-06 06:00:13 +0000[diff] [blame]1646<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]1647// <i>blk is a pointer to a BasicBlock instance</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1648for (BasicBlock::iterator i = blk-&gt;begin(), e = blk-&gt;end(); i != e; ++i)
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]1649 // <i>The next statement works since operator&lt;&lt;(ostream&amp;,...)</i>
1650 // <i>is overloaded for Instruction&amp;</i>
Bill Wendling832171c2006-12-07 20:04:42 +0000[diff] [blame]1651 llvm::cerr &lt;&lt; *i &lt;&lt; "\n";
Chris Lattner55c04612005-03-06 06:00:13 +0000[diff] [blame]1652</pre>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1653</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1654
1655<p>However, this isn't really the best way to print out the contents of a
1656<tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
1657anything you'll care about, you could have just invoked the print routine on the
Bill Wendling832171c2006-12-07 20:04:42 +0000[diff] [blame]1658basic block itself: <tt>llvm::cerr &lt;&lt; *blk &lt;&lt; "\n";</tt>.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1659
1660</div>
1661
1662<!-- _______________________________________________________________________ -->
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1663<div class="doc_subsubsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1664 <a name="iterate_institer">Iterating over the </a><a
1665 href="#Instruction"><tt>Instruction</tt></a>s in a <a
1666 href="#Function"><tt>Function</tt></a>
1667</div>
1668
1669<div class="doc_text">
1670
1671<p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
1672<tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
1673<tt>InstIterator</tt> should be used instead. You'll need to include <a
1674href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
1675and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1676small example that shows how to dump all instructions in a function to the standard error stream:<p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1677
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1678<div class="doc_code">
1679<pre>
1680#include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
1681
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]1682// <i>F is a pointer to a Function instance</i>
Chris Lattnerda021aa2008-06-04 18:20:42 +0000[diff] [blame]1683for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
1684 llvm::cerr &lt;&lt; *I &lt;&lt; "\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1685</pre>
1686</div>
1687
1688<p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]1689work list with its initial contents. For example, if you wanted to
1690initialize a work list to contain all instructions in a <tt>Function</tt>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1691F, all you would need to do is something like:</p>
1692
1693<div class="doc_code">
1694<pre>
1695std::set&lt;Instruction*&gt; worklist;
Chris Lattnerda021aa2008-06-04 18:20:42 +0000[diff] [blame]1696// or better yet, SmallPtrSet&lt;Instruction*, 64&gt; worklist;
1697
1698for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
1699 worklist.insert(&amp;*I);
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1700</pre>
1701</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1702
1703<p>The STL set <tt>worklist</tt> would now contain all instructions in the
1704<tt>Function</tt> pointed to by F.</p>
1705
1706</div>
1707
1708<!-- _______________________________________________________________________ -->
1709<div class="doc_subsubsection">
1710 <a name="iterate_convert">Turning an iterator into a class pointer (and
1711 vice-versa)</a>
1712</div>
1713
1714<div class="doc_text">
1715
1716<p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
Joel Stanley9b96c442002-09-06 21:55:13 +0000[diff] [blame]1717instance when all you've got at hand is an iterator. Well, extracting
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1718a reference or a pointer from an iterator is very straight-forward.
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]1719Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1720is a <tt>BasicBlock::const_iterator</tt>:</p>
1721
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1722<div class="doc_code">
1723<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]1724Instruction&amp; inst = *i; // <i>Grab reference to instruction reference</i>
1725Instruction* pinst = &amp;*i; // <i>Grab pointer to instruction reference</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1726const Instruction&amp; inst = *j;
1727</pre>
1728</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1729
1730<p>However, the iterators you'll be working with in the LLVM framework are
1731special: they will automatically convert to a ptr-to-instance type whenever they
1732need to. Instead of dereferencing the iterator and then taking the address of
1733the result, you can simply assign the iterator to the proper pointer type and
1734you get the dereference and address-of operation as a result of the assignment
1735(behind the scenes, this is a result of overloading casting mechanisms). Thus
1736the last line of the last example,</p>
1737
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1738<div class="doc_code">
1739<pre>
Chris Lattner2e438ca2008-01-03 16:56:04 +0000[diff] [blame]1740Instruction *pinst = &amp;*i;
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1741</pre>
1742</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1743
1744<p>is semantically equivalent to</p>
1745
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1746<div class="doc_code">
1747<pre>
Chris Lattner2e438ca2008-01-03 16:56:04 +0000[diff] [blame]1748Instruction *pinst = i;
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1749</pre>
1750</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1751
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1752<p>It's also possible to turn a class pointer into the corresponding iterator,
1753and this is a constant time operation (very efficient). The following code
1754snippet illustrates use of the conversion constructors provided by LLVM
1755iterators. By using these, you can explicitly grab the iterator of something
1756without actually obtaining it via iteration over some structure:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1757
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1758<div class="doc_code">
1759<pre>
1760void printNextInstruction(Instruction* inst) {
1761 BasicBlock::iterator it(inst);
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]1762 ++it; // <i>After this line, it refers to the instruction after *inst</i>
Bill Wendling832171c2006-12-07 20:04:42 +0000[diff] [blame]1763 if (it != inst-&gt;getParent()-&gt;end()) llvm::cerr &lt;&lt; *it &lt;&lt; "\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1764}
1765</pre>
1766</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1767
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1768</div>
1769
1770<!--_______________________________________________________________________-->
1771<div class="doc_subsubsection">
1772 <a name="iterate_complex">Finding call sites: a slightly more complex
1773 example</a>
1774</div>
1775
1776<div class="doc_text">
1777
1778<p>Say that you're writing a FunctionPass and would like to count all the
1779locations in the entire module (that is, across every <tt>Function</tt>) where a
1780certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
1781learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1782much more straight-forward manner, but this example will allow us to explore how
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]1783you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1784is what we want to do:</p>
1785
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1786<div class="doc_code">
1787<pre>
1788initialize callCounter to zero
1789for each Function f in the Module
1790 for each BasicBlock b in f
1791 for each Instruction i in b
1792 if (i is a CallInst and calls the given function)
1793 increment callCounter
1794</pre>
1795</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1796
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1797<p>And the actual code is (remember, because we're writing a
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1798<tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1799override the <tt>runOnFunction</tt> method):</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1800
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1801<div class="doc_code">
1802<pre>
1803Function* targetFunc = ...;
1804
1805class OurFunctionPass : public FunctionPass {
1806 public:
1807 OurFunctionPass(): callCounter(0) { }
1808
1809 virtual runOnFunction(Function&amp; F) {
1810 for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
Eric Christopher203e71d2008-11-08 08:20:49 +0000[diff] [blame]1811 for (BasicBlock::iterator i = b-&gt;begin(), ie = b-&gt;end(); i != ie; ++i) {
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1812 if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a>&lt;<a
1813 href="#CallInst">CallInst</a>&gt;(&amp;*i)) {
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]1814 // <i>We know we've encountered a call instruction, so we</i>
1815 // <i>need to determine if it's a call to the</i>
Chris Lattner2e438ca2008-01-03 16:56:04 +0000[diff] [blame]1816 // <i>function pointed to by m_func or not.</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1817 if (callInst-&gt;getCalledFunction() == targetFunc)
1818 ++callCounter;
1819 }
1820 }
1821 }
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]1822 }
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1823
1824 private:
Chris Lattner2e438ca2008-01-03 16:56:04 +0000[diff] [blame]1825 unsigned callCounter;
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1826};
1827</pre>
1828</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1829
1830</div>
1831
Brian Gaekef1972c62003-11-07 19:25:45 +0000[diff] [blame]1832<!--_______________________________________________________________________-->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1833<div class="doc_subsubsection">
1834 <a name="calls_and_invokes">Treating calls and invokes the same way</a>
1835</div>
1836
1837<div class="doc_text">
1838
1839<p>You may have noticed that the previous example was a bit oversimplified in
1840that it did not deal with call sites generated by 'invoke' instructions. In
1841this, and in other situations, you may find that you want to treat
1842<tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
1843most-specific common base class is <tt>Instruction</tt>, which includes lots of
1844less closely-related things. For these cases, LLVM provides a handy wrapper
1845class called <a
Reid Spencer05fe4b02006-03-14 05:39:39 +0000[diff] [blame]1846href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1847It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
1848methods that provide functionality common to <tt>CallInst</tt>s and
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1849<tt>InvokeInst</tt>s.</p>
1850
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1851<p>This class has "value semantics": it should be passed by value, not by
1852reference and it should not be dynamically allocated or deallocated using
1853<tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
1854assignable and constructable, with costs equivalents to that of a bare pointer.
1855If you look at its definition, it has only a single pointer member.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1856
1857</div>
1858
Chris Lattner1a3105b2002-09-09 05:49:39 +0000[diff] [blame]1859<!--_______________________________________________________________________-->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1860<div class="doc_subsubsection">
1861 <a name="iterate_chains">Iterating over def-use &amp; use-def chains</a>
1862</div>
1863
1864<div class="doc_text">
1865
1866<p>Frequently, we might have an instance of the <a
Chris Lattner00815172007-01-04 22:01:45 +0000[diff] [blame]1867href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
Misha Brukman384047f2004-06-03 23:29:12 +0000[diff] [blame]1868determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
1869<tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
1870For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
1871particular function <tt>foo</tt>. Finding all of the instructions that
1872<i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
1873of <tt>F</tt>:</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1874
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1875<div class="doc_code">
1876<pre>
Chris Lattner2e438ca2008-01-03 16:56:04 +0000[diff] [blame]1877Function *F = ...;
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1878
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]1879for (Value::use_iterator i = F-&gt;use_begin(), e = F-&gt;use_end(); i != e; ++i)
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1880 if (Instruction *Inst = dyn_cast&lt;Instruction&gt;(*i)) {
Bill Wendling832171c2006-12-07 20:04:42 +0000[diff] [blame]1881 llvm::cerr &lt;&lt; "F is used in instruction:\n";
1882 llvm::cerr &lt;&lt; *Inst &lt;&lt; "\n";
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1883 }
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1884</pre>
1885</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1886
1887<p>Alternately, it's common to have an instance of the <a
Misha Brukman384047f2004-06-03 23:29:12 +0000[diff] [blame]1888href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1889<tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
1890<tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
1891<tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
1892all of the values that a particular instruction uses (that is, the operands of
1893the particular <tt>Instruction</tt>):</p>
1894
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1895<div class="doc_code">
1896<pre>
Chris Lattner2e438ca2008-01-03 16:56:04 +0000[diff] [blame]1897Instruction *pi = ...;
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1898
1899for (User::op_iterator i = pi-&gt;op_begin(), e = pi-&gt;op_end(); i != e; ++i) {
Chris Lattner2e438ca2008-01-03 16:56:04 +0000[diff] [blame]1900 Value *v = *i;
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]1901 // <i>...</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1902}
1903</pre>
1904</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1905
Chris Lattner1a3105b2002-09-09 05:49:39 +0000[diff] [blame]1906<!--
1907 def-use chains ("finding all users of"): Value::use_begin/use_end
1908 use-def chains ("finding all values used"): User::op_begin/op_end [op=operand]
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1909-->
1910
1911</div>
1912
Chris Lattner2e438ca2008-01-03 16:56:04 +0000[diff] [blame]1913<!--_______________________________________________________________________-->
1914<div class="doc_subsubsection">
1915 <a name="iterate_preds">Iterating over predecessors &amp;
1916successors of blocks</a>
1917</div>
1918
1919<div class="doc_text">
1920
1921<p>Iterating over the predecessors and successors of a block is quite easy
1922with the routines defined in <tt>"llvm/Support/CFG.h"</tt>. Just use code like
1923this to iterate over all predecessors of BB:</p>
1924
1925<div class="doc_code">
1926<pre>
1927#include "llvm/Support/CFG.h"
1928BasicBlock *BB = ...;
1929
1930for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
1931 BasicBlock *Pred = *PI;
1932 // <i>...</i>
1933}
1934</pre>
1935</div>
1936
1937<p>Similarly, to iterate over successors use
1938succ_iterator/succ_begin/succ_end.</p>
1939
1940</div>
1941
1942
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1943<!-- ======================================================================= -->
1944<div class="doc_subsection">
1945 <a name="simplechanges">Making simple changes</a>
1946</div>
1947
1948<div class="doc_text">
1949
1950<p>There are some primitive transformation operations present in the LLVM
Joel Stanley753eb712002-09-11 22:32:24 +0000[diff] [blame]1951infrastructure that are worth knowing about. When performing
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]1952transformations, it's fairly common to manipulate the contents of basic
1953blocks. This section describes some of the common methods for doing so
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1954and gives example code.</p>
1955
1956</div>
1957
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]1958<!--_______________________________________________________________________-->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1959<div class="doc_subsubsection">
1960 <a name="schanges_creating">Creating and inserting new
1961 <tt>Instruction</tt>s</a>
1962</div>
1963
1964<div class="doc_text">
1965
1966<p><i>Instantiating Instructions</i></p>
1967
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]1968<p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1969constructor for the kind of instruction to instantiate and provide the necessary
1970parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
1971(const-ptr-to) <tt>Type</tt>. Thus:</p>
1972
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1973<div class="doc_code">
1974<pre>
Nick Lewycky10d64b92007-12-03 01:52:52 +0000[diff] [blame]1975AllocaInst* ai = new AllocaInst(Type::Int32Ty);
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]1976</pre>
1977</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1978
1979<p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]1980one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1981subclass is likely to have varying default parameters which change the semantics
1982of the instruction, so refer to the <a
Misha Brukman31ca1de2004-06-03 23:35:54 +0000[diff] [blame]1983href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1984Instruction</a> that you're interested in instantiating.</p>
1985
1986<p><i>Naming values</i></p>
1987
1988<p>It is very useful to name the values of instructions when you're able to, as
1989this facilitates the debugging of your transformations. If you end up looking
1990at generated LLVM machine code, you definitely want to have logical names
1991associated with the results of instructions! By supplying a value for the
1992<tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
1993associate a logical name with the result of the instruction's execution at
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]1994run time. For example, say that I'm writing a transformation that dynamically
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]1995allocates space for an integer on the stack, and that integer is going to be
1996used as some kind of index by some other code. To accomplish this, I place an
1997<tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
1998<tt>Function</tt>, and I'm intending to use it within the same
1999<tt>Function</tt>. I might do:</p>
2000
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2001<div class="doc_code">
2002<pre>
Nick Lewycky10d64b92007-12-03 01:52:52 +0000[diff] [blame]2003AllocaInst* pa = new AllocaInst(Type::Int32Ty, 0, "indexLoc");
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2004</pre>
2005</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2006
2007<p>where <tt>indexLoc</tt> is now the logical name of the instruction's
Reid Spencer128a7a72007-02-03 21:06:43 +0000[diff] [blame]2008execution value, which is a pointer to an integer on the run time stack.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2009
2010<p><i>Inserting instructions</i></p>
2011
2012<p>There are essentially two ways to insert an <tt>Instruction</tt>
2013into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
2014
Joel Stanley9dd1ad62002-09-18 03:17:23 +0000[diff] [blame]2015<ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2016 <li>Insertion into an explicit instruction list
2017
2018 <p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
2019 <tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
2020 before <tt>*pi</tt>, we do the following: </p>
2021
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2022<div class="doc_code">
2023<pre>
2024BasicBlock *pb = ...;
2025Instruction *pi = ...;
2026Instruction *newInst = new Instruction(...);
2027
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]2028pb-&gt;getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2029</pre>
2030</div>
Alkis Evlogimenos9a5dc4f2004-05-27 00:57:51 +0000[diff] [blame]2031
2032 <p>Appending to the end of a <tt>BasicBlock</tt> is so common that
2033 the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
2034 classes provide constructors which take a pointer to a
2035 <tt>BasicBlock</tt> to be appended to. For example code that
2036 looked like: </p>
2037
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2038<div class="doc_code">
2039<pre>
2040BasicBlock *pb = ...;
2041Instruction *newInst = new Instruction(...);
2042
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]2043pb-&gt;getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2044</pre>
2045</div>
Alkis Evlogimenos9a5dc4f2004-05-27 00:57:51 +0000[diff] [blame]2046
2047 <p>becomes: </p>
2048
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2049<div class="doc_code">
2050<pre>
2051BasicBlock *pb = ...;
2052Instruction *newInst = new Instruction(..., pb);
2053</pre>
2054</div>
Alkis Evlogimenos9a5dc4f2004-05-27 00:57:51 +0000[diff] [blame]2055
2056 <p>which is much cleaner, especially if you are creating
2057 long instruction streams.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2058
2059 <li>Insertion into an implicit instruction list
2060
2061 <p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
2062 are implicitly associated with an existing instruction list: the instruction
2063 list of the enclosing basic block. Thus, we could have accomplished the same
2064 thing as the above code without being given a <tt>BasicBlock</tt> by doing:
2065 </p>
2066
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2067<div class="doc_code">
2068<pre>
2069Instruction *pi = ...;
2070Instruction *newInst = new Instruction(...);
2071
2072pi-&gt;getParent()-&gt;getInstList().insert(pi, newInst);
2073</pre>
2074</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2075
2076 <p>In fact, this sequence of steps occurs so frequently that the
2077 <tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
2078 constructors which take (as a default parameter) a pointer to an
2079 <tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
2080 precede. That is, <tt>Instruction</tt> constructors are capable of
2081 inserting the newly-created instance into the <tt>BasicBlock</tt> of a
2082 provided instruction, immediately before that instruction. Using an
2083 <tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
2084 parameter, the above code becomes:</p>
2085
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2086<div class="doc_code">
2087<pre>
2088Instruction* pi = ...;
2089Instruction* newInst = new Instruction(..., pi);
2090</pre>
2091</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2092
2093 <p>which is much cleaner, especially if you're creating a lot of
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2094 instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2095</ul>
2096
2097</div>
2098
2099<!--_______________________________________________________________________-->
2100<div class="doc_subsubsection">
2101 <a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
2102</div>
2103
2104<div class="doc_text">
2105
2106<p>Deleting an instruction from an existing sequence of instructions that form a
Chris Lattner69bf8a92004-05-23 21:06:58 +0000[diff] [blame]2107<a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward. First,
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2108you must have a pointer to the instruction that you wish to delete. Second, you
2109need to obtain the pointer to that instruction's basic block. You use the
2110pointer to the basic block to get its list of instructions and then use the
2111erase function to remove your instruction. For example:</p>
2112
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2113<div class="doc_code">
2114<pre>
2115<a href="#Instruction">Instruction</a> *I = .. ;
Chris Lattner9f8ec252008-02-15 22:57:17 +0000[diff] [blame]2116I-&gt;eraseFromParent();
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2117</pre>
2118</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2119
2120</div>
2121
2122<!--_______________________________________________________________________-->
2123<div class="doc_subsubsection">
2124 <a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
2125 <tt>Value</tt></a>
2126</div>
2127
2128<div class="doc_text">
2129
2130<p><i>Replacing individual instructions</i></p>
2131
2132<p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]2133permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2134and <tt>ReplaceInstWithInst</tt>.</p>
2135
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]2136<h4><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h4>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2137
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]2138<ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2139 <li><tt>ReplaceInstWithValue</tt>
2140
Nick Lewyckyb6d1f392008-09-15 06:31:52 +0000[diff] [blame]2141 <p>This function replaces all uses of a given instruction with a value,
2142 and then removes the original instruction. The following example
2143 illustrates the replacement of the result of a particular
Chris Lattner58360822005-01-17 00:12:04 +0000[diff] [blame]2144 <tt>AllocaInst</tt> that allocates memory for a single integer with a null
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2145 pointer to an integer.</p>
2146
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2147<div class="doc_code">
2148<pre>
2149AllocaInst* instToReplace = ...;
2150BasicBlock::iterator ii(instToReplace);
2151
2152ReplaceInstWithValue(instToReplace-&gt;getParent()-&gt;getInstList(), ii,
Daniel Dunbar58c2ac02008-10-03 22:17:25 +0000[diff] [blame]2153 Constant::getNullValue(PointerType::getUnqual(Type::Int32Ty)));
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2154</pre></div></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2155
2156 <li><tt>ReplaceInstWithInst</tt>
2157
2158 <p>This function replaces a particular instruction with another
Nick Lewyckyb6d1f392008-09-15 06:31:52 +0000[diff] [blame]2159 instruction, inserting the new instruction into the basic block at the
2160 location where the old instruction was, and replacing any uses of the old
2161 instruction with the new instruction. The following example illustrates
2162 the replacement of one <tt>AllocaInst</tt> with another.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2163
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2164<div class="doc_code">
2165<pre>
2166AllocaInst* instToReplace = ...;
2167BasicBlock::iterator ii(instToReplace);
2168
2169ReplaceInstWithInst(instToReplace-&gt;getParent()-&gt;getInstList(), ii,
Nick Lewycky10d64b92007-12-03 01:52:52 +0000[diff] [blame]2170 new AllocaInst(Type::Int32Ty, 0, "ptrToReplacedInt"));
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2171</pre></div></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]2172</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2173
2174<p><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></p>
2175
2176<p>You can use <tt>Value::replaceAllUsesWith</tt> and
2177<tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
Chris Lattner00815172007-01-04 22:01:45 +0000[diff] [blame]2178doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
Misha Brukman384047f2004-06-03 23:29:12 +0000[diff] [blame]2179and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2180information.</p>
2181
2182<!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
2183include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
2184ReplaceInstWithValue, ReplaceInstWithInst -->
2185
2186</div>
2187
Tanya Lattnerb011c662007-06-20 18:33:15 +0000[diff] [blame]2188<!--_______________________________________________________________________-->
2189<div class="doc_subsubsection">
2190 <a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
2191</div>
2192
2193<div class="doc_text">
2194
Tanya Lattnerc5dfcdb2007-06-20 20:46:37 +0000[diff] [blame]2195<p>Deleting a global variable from a module is just as easy as deleting an
2196Instruction. First, you must have a pointer to the global variable that you wish
2197 to delete. You use this pointer to erase it from its parent, the module.
Tanya Lattnerb011c662007-06-20 18:33:15 +0000[diff] [blame]2198 For example:</p>
2199
2200<div class="doc_code">
2201<pre>
2202<a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
Tanya Lattnerb011c662007-06-20 18:33:15 +0000[diff] [blame]2203
Tanya Lattnerc5dfcdb2007-06-20 20:46:37 +0000[diff] [blame]2204GV-&gt;eraseFromParent();
Tanya Lattnerb011c662007-06-20 18:33:15 +0000[diff] [blame]2205</pre>
2206</div>
2207
2208</div>
2209
Jeffrey Yasskin714257f2009-04-30 22:33:41 +0000[diff] [blame]2210<!-- ======================================================================= -->
2211<div class="doc_subsection">
2212 <a name="create_types">How to Create Types</a>
2213</div>
2214
2215<div class="doc_text">
2216
2217<p>In generating IR, you may need some complex types. If you know these types
Misha Brukman1af789f2009-05-01 20:40:51 +0000[diff] [blame]2218statically, you can use <tt>TypeBuilder&lt;...&gt;::get()</tt>, defined
Jeffrey Yasskin714257f2009-04-30 22:33:41 +0000[diff] [blame]2219in <tt>llvm/Support/TypeBuilder.h</tt>, to retrieve them. <tt>TypeBuilder</tt>
2220has two forms depending on whether you're building types for cross-compilation
Misha Brukman1af789f2009-05-01 20:40:51 +0000[diff] [blame]2221or native library use. <tt>TypeBuilder&lt;T, true&gt;</tt> requires
Jeffrey Yasskin714257f2009-04-30 22:33:41 +0000[diff] [blame]2222that <tt>T</tt> be independent of the host environment, meaning that it's built
2223out of types from
2224the <a href="/doxygen/namespacellvm_1_1types.html"><tt>llvm::types</tt></a>
2225namespace and pointers, functions, arrays, etc. built of
Misha Brukman1af789f2009-05-01 20:40:51 +0000[diff] [blame]2226those. <tt>TypeBuilder&lt;T, false&gt;</tt> additionally allows native C types
Jeffrey Yasskin714257f2009-04-30 22:33:41 +0000[diff] [blame]2227whose size may depend on the host compiler. For example,</p>
2228
2229<div class="doc_code">
2230<pre>
Misha Brukman1af789f2009-05-01 20:40:51 +0000[diff] [blame]2231FunctionType *ft = TypeBuilder&lt;types::i&lt;8&gt;(types::i&lt;32&gt;*), true&gt;::get();
Jeffrey Yasskin714257f2009-04-30 22:33:41 +0000[diff] [blame]2232</pre>
2233</div>
2234
2235<p>is easier to read and write than the equivalent</p>
2236
2237<div class="doc_code">
2238<pre>
Owen Anderson5e8c50e2009-06-16 17:40:28 +0000[diff] [blame]2239std::vector&lt;const Type*&gt; params;
Jeffrey Yasskin714257f2009-04-30 22:33:41 +0000[diff] [blame]2240params.push_back(PointerType::getUnqual(Type::Int32Ty));
2241FunctionType *ft = FunctionType::get(Type::Int8Ty, params, false);
2242</pre>
2243</div>
2244
2245<p>See the <a href="/doxygen/TypeBuilder_8h-source.html#l00001">class
2246comment</a> for more details.</p>
2247
2248</div>
2249
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]2250<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2251<div class="doc_section">
Owen Anderson8bc1b3b2009-06-16 01:17:16 +0000[diff] [blame]2252 <a name="threading">Threads and LLVM</a>
2253</div>
2254<!-- *********************************************************************** -->
2255
2256<div class="doc_text">
2257<p>
2258This section describes the interaction of the LLVM APIs with multithreading,
2259both on the part of client applications, and in the JIT, in the hosted
2260application.
2261</p>
2262
2263<p>
2264Note that LLVM's support for multithreading is still relatively young. Up
2265through version 2.5, the execution of threaded hosted applications was
2266supported, but not threaded client access to the APIs. While this use case is
2267now supported, clients <em>must</em> adhere to the guidelines specified below to
2268ensure proper operation in multithreaded mode.
2269</p>
2270
2271<p>
2272Note that, on Unix-like platforms, LLVM requires the presence of GCC's atomic
2273intrinsics in order to support threaded operation. If you need a
2274multhreading-capable LLVM on a platform without a suitably modern system
2275compiler, consider compiling LLVM and LLVM-GCC in single-threaded mode, and
2276using the resultant compiler to build a copy of LLVM with multithreading
2277support.
2278</p>
2279</div>
2280
2281<!-- ======================================================================= -->
2282<div class="doc_subsection">
Owen Anderson1ad70e32009-06-16 18:04:19 +0000[diff] [blame]2283 <a name="startmultithreaded">Entering and Exiting Multithreaded Mode</a>
Owen Anderson8bc1b3b2009-06-16 01:17:16 +0000[diff] [blame]2284</div>
2285
2286<div class="doc_text">
2287
2288<p>
2289In order to properly protect its internal data structures while avoiding
Owen Anderson1ad70e32009-06-16 18:04:19 +0000[diff] [blame]2290excessive locking overhead in the single-threaded case, the LLVM must intialize
2291certain data structures necessary to provide guards around its internals. To do
2292so, the client program must invoke <tt>llvm_start_multithreaded()</tt> before
2293making any concurrent LLVM API calls. To subsequently tear down these
2294structures, use the <tt>llvm_stop_multithreaded()</tt> call. You can also use
2295the <tt>llvm_is_multithreaded()</tt> call to check the status of multithreaded
2296mode.
Owen Anderson8bc1b3b2009-06-16 01:17:16 +0000[diff] [blame]2297</p>
2298
2299<p>
Owen Anderson1ad70e32009-06-16 18:04:19 +0000[diff] [blame]2300Note that both of these calls must be made <em>in isolation</em>. That is to
2301say that no other LLVM API calls may be executing at any time during the
2302execution of <tt>llvm_start_multithreaded()</tt> or <tt>llvm_stop_multithreaded
2303</tt>. It's is the client's responsibility to enforce this isolation.
2304</p>
2305
2306<p>
2307The return value of <tt>llvm_start_multithreaded()</tt> indicates the success or
2308failure of the initialization. Failure typically indicates that your copy of
2309LLVM was built without multithreading support, typically because GCC atomic
2310intrinsics were not found in your system compiler. In this case, the LLVM API
2311will not be safe for concurrent calls. However, it <em>will</em> be safe for
2312hosting threaded applications in the JIT, though care must be taken to ensure
2313that side exits and the like do not accidentally result in concurrent LLVM API
2314calls.
Owen Anderson8bc1b3b2009-06-16 01:17:16 +0000[diff] [blame]2315</p>
2316</div>
2317
2318<!-- ======================================================================= -->
2319<div class="doc_subsection">
2320 <a name="shutdown">Ending Execution with <tt>llvm_shutdown()</tt></a>
2321</div>
2322
2323<div class="doc_text">
2324<p>
2325When you are done using the LLVM APIs, you should call <tt>llvm_shutdown()</tt>
Owen Anderson1ad70e32009-06-16 18:04:19 +0000[diff] [blame]2326to deallocate memory used for internal structures. This will also invoke
2327<tt>llvm_stop_multithreaded()</tt> if LLVM is operating in multithreaded mode.
2328As such, <tt>llvm_shutdown()</tt> requires the same isolation guarantees as
2329<tt>llvm_stop_multithreaded()</tt>.
Owen Anderson8bc1b3b2009-06-16 01:17:16 +0000[diff] [blame]2330</p>
2331
2332<p>
2333Note that, if you use scope-based shutdown, you can use the
2334<tt>llvm_shutdown_obj</tt> class, which calls <tt>llvm_shutdown()</tt> in its
2335destructor.
2336</div>
2337
2338<!-- ======================================================================= -->
2339<div class="doc_subsection">
2340 <a name="managedstatic">Lazy Initialization with <tt>ManagedStatic</tt></a>
2341</div>
2342
2343<div class="doc_text">
2344<p>
2345<tt>ManagedStatic</tt> is a utility class in LLVM used to implement static
2346initialization of static resources, such as the global type tables. Before the
2347invocation of <tt>llvm_shutdown()</tt>, it implements a simple lazy
2348initialization scheme. Once <tt>llvm_start_multithreaded()</tt> returns,
2349however, it uses double-checked locking to implement thread-safe lazy
2350initialization.
2351</p>
2352
2353<p>
2354Note that, because no other threads are allowed to issue LLVM API calls before
2355<tt>llvm_start_multithreaded()</tt> returns, it is possible to have
2356<tt>ManagedStatic</tt>s of <tt>llvm::sys::Mutex</tt>s.
2357</p>
Owen Anderson1ad70e32009-06-16 18:04:19 +0000[diff] [blame]2358
2359<p>
2360The <tt>llvm_acquire_global_lock()</tt> and <tt>llvm_release_global_lock</tt>
2361APIs provide access to the global lock used to implement the double-checked
2362locking for lazy initialization. These should only be used internally to LLVM,
2363and only if you know what you're doing!
2364</p>
Owen Anderson8bc1b3b2009-06-16 01:17:16 +0000[diff] [blame]2365</div>
2366
2367<!-- *********************************************************************** -->
2368<div class="doc_section">
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2369 <a name="advanced">Advanced Topics</a>
2370</div>
2371<!-- *********************************************************************** -->
2372
2373<div class="doc_text">
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2374<p>
2375This section describes some of the advanced or obscure API's that most clients
2376do not need to be aware of. These API's tend manage the inner workings of the
2377LLVM system, and only need to be accessed in unusual circumstances.
2378</p>
2379</div>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2380
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2381<!-- ======================================================================= -->
2382<div class="doc_subsection">
2383 <a name="TypeResolve">LLVM Type Resolution</a>
2384</div>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2385
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2386<div class="doc_text">
2387
2388<p>
2389The LLVM type system has a very simple goal: allow clients to compare types for
2390structural equality with a simple pointer comparison (aka a shallow compare).
2391This goal makes clients much simpler and faster, and is used throughout the LLVM
2392system.
2393</p>
2394
2395<p>
2396Unfortunately achieving this goal is not a simple matter. In particular,
2397recursive types and late resolution of opaque types makes the situation very
2398difficult to handle. Fortunately, for the most part, our implementation makes
2399most clients able to be completely unaware of the nasty internal details. The
2400primary case where clients are exposed to the inner workings of it are when
Gabor Greif04367bf2007-07-06 22:07:22 +0000[diff] [blame]2401building a recursive type. In addition to this case, the LLVM bitcode reader,
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2402assembly parser, and linker also have to be aware of the inner workings of this
2403system.
2404</p>
2405
Chris Lattner0f876db2005-04-25 15:47:57 +0000[diff] [blame]2406<p>
2407For our purposes below, we need three concepts. First, an "Opaque Type" is
2408exactly as defined in the <a href="LangRef.html#t_opaque">language
2409reference</a>. Second an "Abstract Type" is any type which includes an
Reid Spencer06565dc2007-01-12 17:11:23 +0000[diff] [blame]2410opaque type as part of its type graph (for example "<tt>{ opaque, i32 }</tt>").
2411Third, a concrete type is a type that is not an abstract type (e.g. "<tt>{ i32,
Chris Lattner0f876db2005-04-25 15:47:57 +0000[diff] [blame]2412float }</tt>").
2413</p>
2414
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2415</div>
2416
2417<!-- ______________________________________________________________________ -->
2418<div class="doc_subsubsection">
2419 <a name="BuildRecType">Basic Recursive Type Construction</a>
2420</div>
2421
2422<div class="doc_text">
2423
2424<p>
2425Because the most common question is "how do I build a recursive type with LLVM",
2426we answer it now and explain it as we go. Here we include enough to cause this
2427to be emitted to an output .ll file:
2428</p>
2429
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2430<div class="doc_code">
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2431<pre>
Reid Spencer06565dc2007-01-12 17:11:23 +0000[diff] [blame]2432%mylist = type { %mylist*, i32 }
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2433</pre>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2434</div>
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2435
2436<p>
2437To build this, use the following LLVM APIs:
2438</p>
2439
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2440<div class="doc_code">
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2441<pre>
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]2442// <i>Create the initial outer struct</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2443<a href="#PATypeHolder">PATypeHolder</a> StructTy = OpaqueType::get();
2444std::vector&lt;const Type*&gt; Elts;
Daniel Dunbar58c2ac02008-10-03 22:17:25 +0000[diff] [blame]2445Elts.push_back(PointerType::getUnqual(StructTy));
Nick Lewycky10d64b92007-12-03 01:52:52 +0000[diff] [blame]2446Elts.push_back(Type::Int32Ty);
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2447StructType *NewSTy = StructType::get(Elts);
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2448
Reid Spencer06565dc2007-01-12 17:11:23 +0000[diff] [blame]2449// <i>At this point, NewSTy = "{ opaque*, i32 }". Tell VMCore that</i>
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]2450// <i>the struct and the opaque type are actually the same.</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2451cast&lt;OpaqueType&gt;(StructTy.get())-&gt;<a href="#refineAbstractTypeTo">refineAbstractTypeTo</a>(NewSTy);
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2452
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2453// <i>NewSTy is potentially invalidated, but StructTy (a <a href="#PATypeHolder">PATypeHolder</a>) is</i>
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]2454// <i>kept up-to-date</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2455NewSTy = cast&lt;StructType&gt;(StructTy.get());
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2456
Bill Wendling82e2eea2006-10-11 18:00:22 +0000[diff] [blame]2457// <i>Add a name for the type to the module symbol table (optional)</i>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2458MyModule-&gt;addTypeName("mylist", NewSTy);
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2459</pre>
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]2460</div>
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2461
2462<p>
2463This code shows the basic approach used to build recursive types: build a
2464non-recursive type using 'opaque', then use type unification to close the cycle.
2465The type unification step is performed by the <tt><a
Chris Lattneraff26d12007-02-03 03:06:52 +0000[diff] [blame]2466href="#refineAbstractTypeTo">refineAbstractTypeTo</a></tt> method, which is
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2467described next. After that, we describe the <a
2468href="#PATypeHolder">PATypeHolder class</a>.
2469</p>
2470
2471</div>
2472
2473<!-- ______________________________________________________________________ -->
2474<div class="doc_subsubsection">
2475 <a name="refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a>
2476</div>
2477
2478<div class="doc_text">
2479<p>
2480The <tt>refineAbstractTypeTo</tt> method starts the type unification process.
2481While this method is actually a member of the DerivedType class, it is most
2482often used on OpaqueType instances. Type unification is actually a recursive
2483process. After unification, types can become structurally isomorphic to
2484existing types, and all duplicates are deleted (to preserve pointer equality).
2485</p>
2486
2487<p>
2488In the example above, the OpaqueType object is definitely deleted.
Reid Spencer06565dc2007-01-12 17:11:23 +0000[diff] [blame]2489Additionally, if there is an "{ \2*, i32}" type already created in the system,
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2490the pointer and struct type created are <b>also</b> deleted. Obviously whenever
2491a type is deleted, any "Type*" pointers in the program are invalidated. As
2492such, it is safest to avoid having <i>any</i> "Type*" pointers to abstract types
2493live across a call to <tt>refineAbstractTypeTo</tt> (note that non-abstract
2494types can never move or be deleted). To deal with this, the <a
2495href="#PATypeHolder">PATypeHolder</a> class is used to maintain a stable
2496reference to a possibly refined type, and the <a
2497href="#AbstractTypeUser">AbstractTypeUser</a> class is used to update more
2498complex datastructures.
2499</p>
2500
2501</div>
2502
2503<!-- ______________________________________________________________________ -->
2504<div class="doc_subsubsection">
2505 <a name="PATypeHolder">The PATypeHolder Class</a>
2506</div>
2507
2508<div class="doc_text">
2509<p>
2510PATypeHolder is a form of a "smart pointer" for Type objects. When VMCore
2511happily goes about nuking types that become isomorphic to existing types, it
2512automatically updates all PATypeHolder objects to point to the new type. In the
2513example above, this allows the code to maintain a pointer to the resultant
2514resolved recursive type, even though the Type*'s are potentially invalidated.
2515</p>
2516
2517<p>
2518PATypeHolder is an extremely light-weight object that uses a lazy union-find
2519implementation to update pointers. For example the pointer from a Value to its
2520Type is maintained by PATypeHolder objects.
2521</p>
2522
2523</div>
2524
2525<!-- ______________________________________________________________________ -->
2526<div class="doc_subsubsection">
2527 <a name="AbstractTypeUser">The AbstractTypeUser Class</a>
2528</div>
2529
2530<div class="doc_text">
2531
2532<p>
2533Some data structures need more to perform more complex updates when types get
Chris Lattner263a98e2007-02-16 04:37:31 +0000[diff] [blame]2534resolved. To support this, a class can derive from the AbstractTypeUser class.
2535This class
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2536allows it to get callbacks when certain types are resolved. To register to get
2537callbacks for a particular type, the DerivedType::{add/remove}AbstractTypeUser
Chris Lattner0f876db2005-04-25 15:47:57 +0000[diff] [blame]2538methods can be called on a type. Note that these methods only work for <i>
Reid Spencer06565dc2007-01-12 17:11:23 +0000[diff] [blame]2539 abstract</i> types. Concrete types (those that do not include any opaque
2540objects) can never be refined.
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2541</p>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2542</div>
2543
2544
2545<!-- ======================================================================= -->
2546<div class="doc_subsection">
Chris Lattner263a98e2007-02-16 04:37:31 +0000[diff] [blame]2547 <a name="SymbolTable">The <tt>ValueSymbolTable</tt> and
2548 <tt>TypeSymbolTable</tt> classes</a>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2549</div>
Chris Lattnerf1b200b2005-04-23 17:27:36 +0000[diff] [blame]2550
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2551<div class="doc_text">
Chris Lattner263a98e2007-02-16 04:37:31 +0000[diff] [blame]2552<p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
2553ValueSymbolTable</a></tt> class provides a symbol table that the <a
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2554href="#Function"><tt>Function</tt></a> and <a href="#Module">
Chris Lattner263a98e2007-02-16 04:37:31 +0000[diff] [blame]2555<tt>Module</tt></a> classes use for naming value definitions. The symbol table
2556can provide a name for any <a href="#Value"><tt>Value</tt></a>.
2557The <tt><a href="http://llvm.org/doxygen/classllvm_1_1TypeSymbolTable.html">
2558TypeSymbolTable</a></tt> class is used by the <tt>Module</tt> class to store
2559names for types.</p>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2560
Reid Spencera6362242007-01-07 00:41:39 +0000[diff] [blame]2561<p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
2562by most clients. It should only be used when iteration over the symbol table
2563names themselves are required, which is very special purpose. Note that not
2564all LLVM
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2565<tt><a href="#Value">Value</a></tt>s have names, and those without names (i.e. they have
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2566an empty name) do not exist in the symbol table.
2567</p>
2568
Chris Lattner263a98e2007-02-16 04:37:31 +0000[diff] [blame]2569<p>These symbol tables support iteration over the values/types in the symbol
2570table with <tt>begin/end/iterator</tt> and supports querying to see if a
2571specific name is in the symbol table (with <tt>lookup</tt>). The
2572<tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
2573simply call <tt>setName</tt> on a value, which will autoinsert it into the
2574appropriate symbol table. For types, use the Module::addTypeName method to
2575insert entries into the symbol table.</p>
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2576
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2577</div>
2578
2579
2580
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2581<!-- ======================================================================= -->
2582<div class="doc_subsection">
2583 <a name="UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a>
2584</div>
2585
2586<div class="doc_text">
2587<p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1User.html">
Gabor Greiffd095b62009-01-05 16:05:32 +0000[diff] [blame]2588User</a></tt> class provides a basis for expressing the ownership of <tt>User</tt>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2589towards other <tt><a href="http://llvm.org/doxygen/classllvm_1_1Value.html">
2590Value</a></tt>s. The <tt><a href="http://llvm.org/doxygen/classllvm_1_1Use.html">
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2591Use</a></tt> helper class is employed to do the bookkeeping and to facilitate <i>O(1)</i>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2592addition and removal.</p>
2593
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2594<!-- ______________________________________________________________________ -->
2595<div class="doc_subsubsection">
Gabor Greiffd095b62009-01-05 16:05:32 +0000[diff] [blame]2596 <a name="Use2User">Interaction and relationship between <tt>User</tt> and <tt>Use</tt> objects</a>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2597</div>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2598
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2599<div class="doc_text">
2600<p>
2601A subclass of <tt>User</tt> can choose between incorporating its <tt>Use</tt> objects
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2602or refer to them out-of-line by means of a pointer. A mixed variant
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2603(some <tt>Use</tt>s inline others hung off) is impractical and breaks the invariant
2604that the <tt>Use</tt> objects belonging to the same <tt>User</tt> form a contiguous array.
2605</p>
2606</div>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2607
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2608<p>
2609We have 2 different layouts in the <tt>User</tt> (sub)classes:
2610<ul>
2611<li><p>Layout a)
2612The <tt>Use</tt> object(s) are inside (resp. at fixed offset) of the <tt>User</tt>
2613object and there are a fixed number of them.</p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2614
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2615<li><p>Layout b)
2616The <tt>Use</tt> object(s) are referenced by a pointer to an
2617array from the <tt>User</tt> object and there may be a variable
2618number of them.</p>
2619</ul>
2620<p>
Gabor Greifd41720a2008-06-25 00:10:22 +0000[diff] [blame]2621As of v2.4 each layout still possesses a direct pointer to the
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2622start of the array of <tt>Use</tt>s. Though not mandatory for layout a),
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2623we stick to this redundancy for the sake of simplicity.
Gabor Greifd41720a2008-06-25 00:10:22 +0000[diff] [blame]2624The <tt>User</tt> object also stores the number of <tt>Use</tt> objects it
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2625has. (Theoretically this information can also be calculated
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2626given the scheme presented below.)</p>
2627<p>
2628Special forms of allocation operators (<tt>operator new</tt>)
Gabor Greifd41720a2008-06-25 00:10:22 +0000[diff] [blame]2629enforce the following memory layouts:</p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2630
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2631<ul>
Gabor Greifd41720a2008-06-25 00:10:22 +0000[diff] [blame]2632<li><p>Layout a) is modelled by prepending the <tt>User</tt> object by the <tt>Use[]</tt> array.</p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2633
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2634<pre>
2635...---.---.---.---.-------...
2636 | P | P | P | P | User
2637'''---'---'---'---'-------'''
2638</pre>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2639
Gabor Greifd41720a2008-06-25 00:10:22 +0000[diff] [blame]2640<li><p>Layout b) is modelled by pointing at the <tt>Use[]</tt> array.</p>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2641<pre>
2642.-------...
2643| User
2644'-------'''
2645 |
2646 v
2647 .---.---.---.---...
2648 | P | P | P | P |
2649 '---'---'---'---'''
2650</pre>
2651</ul>
2652<i>(In the above figures '<tt>P</tt>' stands for the <tt>Use**</tt> that
2653 is stored in each <tt>Use</tt> object in the member <tt>Use::Prev</tt>)</i>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2654
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2655<!-- ______________________________________________________________________ -->
2656<div class="doc_subsubsection">
Gabor Greiffd095b62009-01-05 16:05:32 +0000[diff] [blame]2657 <a name="Waymarking">The waymarking algorithm</a>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2658</div>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2659
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2660<div class="doc_text">
2661<p>
Gabor Greifd41720a2008-06-25 00:10:22 +0000[diff] [blame]2662Since the <tt>Use</tt> objects are deprived of the direct (back)pointer to
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2663their <tt>User</tt> objects, there must be a fast and exact method to
2664recover it. This is accomplished by the following scheme:</p>
2665</div>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2666
Gabor Greifd41720a2008-06-25 00:10:22 +0000[diff] [blame]2667A bit-encoding in the 2 LSBits (least significant bits) of the <tt>Use::Prev</tt> allows to find the
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2668start of the <tt>User</tt> object:
2669<ul>
2670<li><tt>00</tt> &mdash;&gt; binary digit 0</li>
2671<li><tt>01</tt> &mdash;&gt; binary digit 1</li>
2672<li><tt>10</tt> &mdash;&gt; stop and calculate (<tt>s</tt>)</li>
2673<li><tt>11</tt> &mdash;&gt; full stop (<tt>S</tt>)</li>
2674</ul>
2675<p>
2676Given a <tt>Use*</tt>, all we have to do is to walk till we get
2677a stop and we either have a <tt>User</tt> immediately behind or
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2678we have to walk to the next stop picking up digits
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2679and calculating the offset:</p>
2680<pre>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2681.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.----------------
2682| 1 | s | 1 | 0 | 1 | 0 | s | 1 | 1 | 0 | s | 1 | 1 | s | 1 | S | User (or User*)
2683'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'----------------
2684 |+15 |+10 |+6 |+3 |+1
2685 | | | | |__>
2686 | | | |__________>
2687 | | |______________________>
2688 | |______________________________________>
2689 |__________________________________________________________>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2690</pre>
2691<p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2692Only the significant number of bits need to be stored between the
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2693stops, so that the <i>worst case is 20 memory accesses</i> when there are
26941000 <tt>Use</tt> objects associated with a <tt>User</tt>.</p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2695
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2696<!-- ______________________________________________________________________ -->
2697<div class="doc_subsubsection">
Gabor Greiffd095b62009-01-05 16:05:32 +0000[diff] [blame]2698 <a name="ReferenceImpl">Reference implementation</a>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2699</div>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2700
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2701<div class="doc_text">
2702<p>
2703The following literate Haskell fragment demonstrates the concept:</p>
2704</div>
2705
2706<div class="doc_code">
2707<pre>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2708> import Test.QuickCheck
2709>
2710> digits :: Int -> [Char] -> [Char]
2711> digits 0 acc = '0' : acc
2712> digits 1 acc = '1' : acc
2713> digits n acc = digits (n `div` 2) $ digits (n `mod` 2) acc
2714>
2715> dist :: Int -> [Char] -> [Char]
2716> dist 0 [] = ['S']
2717> dist 0 acc = acc
2718> dist 1 acc = let r = dist 0 acc in 's' : digits (length r) r
2719> dist n acc = dist (n - 1) $ dist 1 acc
2720>
2721> takeLast n ss = reverse $ take n $ reverse ss
2722>
2723> test = takeLast 40 $ dist 20 []
2724>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2725</pre>
2726</div>
2727<p>
2728Printing &lt;test&gt; gives: <tt>"1s100000s11010s10100s1111s1010s110s11s1S"</tt></p>
2729<p>
2730The reverse algorithm computes the length of the string just by examining
2731a certain prefix:</p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2732
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2733<div class="doc_code">
2734<pre>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2735> pref :: [Char] -> Int
2736> pref "S" = 1
2737> pref ('s':'1':rest) = decode 2 1 rest
2738> pref (_:rest) = 1 + pref rest
2739>
2740> decode walk acc ('0':rest) = decode (walk + 1) (acc * 2) rest
2741> decode walk acc ('1':rest) = decode (walk + 1) (acc * 2 + 1) rest
2742> decode walk acc _ = walk + acc
2743>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2744</pre>
2745</div>
2746<p>
2747Now, as expected, printing &lt;pref test&gt; gives <tt>40</tt>.</p>
2748<p>
2749We can <i>quickCheck</i> this with following property:</p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2750
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2751<div class="doc_code">
2752<pre>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2753> testcase = dist 2000 []
2754> testcaseLength = length testcase
2755>
2756> identityProp n = n > 0 && n <= testcaseLength ==> length arr == pref arr
2757> where arr = takeLast n testcase
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2758>
2759</pre>
2760</div>
2761<p>
2762As expected &lt;quickCheck identityProp&gt; gives:</p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2763
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2764<pre>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2765*Main> quickCheck identityProp
2766OK, passed 100 tests.
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2767</pre>
2768<p>
2769Let's be a bit more exhaustive:</p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2770
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2771<div class="doc_code">
2772<pre>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2773>
2774> deepCheck p = check (defaultConfig { configMaxTest = 500 }) p
2775>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2776</pre>
2777</div>
2778<p>
2779And here is the result of &lt;deepCheck identityProp&gt;:</p>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2780
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2781<pre>
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2782*Main> deepCheck identityProp
2783OK, passed 500 tests.
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2784</pre>
2785
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2786<!-- ______________________________________________________________________ -->
2787<div class="doc_subsubsection">
Gabor Greiffd095b62009-01-05 16:05:32 +0000[diff] [blame]2788 <a name="Tagging">Tagging considerations</a>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2789</div>
2790
2791<p>
2792To maintain the invariant that the 2 LSBits of each <tt>Use**</tt> in <tt>Use</tt>
2793never change after being set up, setters of <tt>Use::Prev</tt> must re-tag the
2794new <tt>Use**</tt> on every modification. Accordingly getters must strip the
2795tag bits.</p>
2796<p>
Gabor Greifd41720a2008-06-25 00:10:22 +0000[diff] [blame]2797For layout b) instead of the <tt>User</tt> we find a pointer (<tt>User*</tt> with LSBit set).
2798Following this pointer brings us to the <tt>User</tt>. A portable trick ensures
2799that the first bytes of <tt>User</tt> (if interpreted as a pointer) never has
Gabor Greiffd095b62009-01-05 16:05:32 +0000[diff] [blame]2800the LSBit set. (Portability is relying on the fact that all known compilers place the
2801<tt>vptr</tt> in the first word of the instances.)</p>
Gabor Greifdfed1182008-06-18 13:44:57 +0000[diff] [blame]2802
Gabor Greife98fc272008-06-16 21:06:12 +0000[diff] [blame]2803</div>
2804
2805 <!-- *********************************************************************** -->
Chris Lattnerd9d6e102005-04-23 16:10:52 +0000[diff] [blame]2806<div class="doc_section">
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2807 <a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
2808</div>
2809<!-- *********************************************************************** -->
2810
2811<div class="doc_text">
Reid Spencer303c4b42007-01-12 17:26:25 +0000[diff] [blame]2812<p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
2813<br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2814
2815<p>The Core LLVM classes are the primary means of representing the program
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]2816being inspected or transformed. The core LLVM classes are defined in
2817header files in the <tt>include/llvm/</tt> directory, and implemented in
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]2818the <tt>lib/VMCore</tt> directory.</p>
2819
2820</div>
2821
2822<!-- ======================================================================= -->
2823<div class="doc_subsection">
Reid Spencer303c4b42007-01-12 17:26:25 +0000[diff] [blame]2824 <a name="Type">The <tt>Type</tt> class and Derived Types</a>
2825</div>
2826
2827<div class="doc_text">
2828
2829 <p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
2830 a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
2831 through its subclasses. Certain primitive types (<tt>VoidType</tt>,
2832 <tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
2833 subclasses. They are hidden because they offer no useful functionality beyond
2834 what the <tt>Type</tt> class offers except to distinguish themselves from
2835 other subclasses of <tt>Type</tt>.</p>
2836 <p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
2837 named, but this is not a requirement. There exists exactly
2838 one instance of a given shape at any one time. This allows type equality to
2839 be performed with address equality of the Type Instance. That is, given two
2840 <tt>Type*</tt> values, the types are identical if the pointers are identical.
2841 </p>
2842</div>
2843
2844<!-- _______________________________________________________________________ -->
2845<div class="doc_subsubsection">
Gabor Greiffd095b62009-01-05 16:05:32 +0000[diff] [blame]2846 <a name="m_Type">Important Public Methods</a>
Reid Spencer303c4b42007-01-12 17:26:25 +0000[diff] [blame]2847</div>
2848
2849<div class="doc_text">
2850
2851<ul>
Chris Lattner8f79df32007-01-15 01:55:32 +0000[diff] [blame]2852 <li><tt>bool isInteger() const</tt>: Returns true for any integer type.</li>
Reid Spencer303c4b42007-01-12 17:26:25 +0000[diff] [blame]2853
2854 <li><tt>bool isFloatingPoint()</tt>: Return true if this is one of the two
2855 floating point types.</li>
2856
2857 <li><tt>bool isAbstract()</tt>: Return true if the type is abstract (contains
2858 an OpaqueType anywhere in its definition).</li>
2859
2860 <li><tt>bool isSized()</tt>: Return true if the type has known size. Things
2861 that don't have a size are abstract types, labels and void.</li>
2862
2863</ul>
2864</div>
2865
2866<!-- _______________________________________________________________________ -->
2867<div class="doc_subsubsection">
Gabor Greiffd095b62009-01-05 16:05:32 +0000[diff] [blame]2868 <a name="derivedtypes">Important Derived Types</a>
Reid Spencer303c4b42007-01-12 17:26:25 +0000[diff] [blame]2869</div>
2870<div class="doc_text">
2871<dl>
2872 <dt><tt>IntegerType</tt></dt>
2873 <dd>Subclass of DerivedType that represents integer types of any bit width.
2874 Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
2875 <tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
2876 <ul>
2877 <li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
2878 type of a specific bit width.</li>
2879 <li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
2880 type.</li>
2881 </ul>
2882 </dd>
2883 <dt><tt>SequentialType</tt></dt>
2884 <dd>This is subclassed by ArrayType and PointerType
2885 <ul>
2886 <li><tt>const Type * getElementType() const</tt>: Returns the type of each
2887 of the elements in the sequential type. </li>
2888 </ul>
2889 </dd>
2890 <dt><tt>ArrayType</tt></dt>
2891 <dd>This is a subclass of SequentialType and defines the interface for array
2892 types.
2893 <ul>
2894 <li><tt>unsigned getNumElements() const</tt>: Returns the number of
2895 elements in the array. </li>
2896 </ul>
2897 </dd>
2898 <dt><tt>PointerType</tt></dt>
Chris Lattner302da1e2007-02-03 03:05:57 +0000[diff] [blame]2899 <dd>Subclass of SequentialType for pointer types.</dd>
Reid Spencer9d6565a2007-02-15 02:26:10 +0000[diff] [blame]2900 <dt><tt>VectorType</tt></dt>
Reid Spencer485bad12007-02-15 03:07:05 +0000[diff] [blame]2901 <dd>Subclass of SequentialType for vector types. A
2902 vector type is similar to an ArrayType but is distinguished because it is
2903 a first class type wherease ArrayType is not. Vector types are used for
Reid Spencer303c4b42007-01-12 17:26:25 +0000[diff] [blame]2904 vector operations and are usually small vectors of of an integer or floating
2905 point type.</dd>
2906 <dt><tt>StructType</tt></dt>
2907 <dd>Subclass of DerivedTypes for struct types.</dd>
Duncan Sands8036ca42007-03-30 12:22:09 +0000[diff] [blame]2908 <dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
Reid Spencer303c4b42007-01-12 17:26:25 +0000[diff] [blame]2909 <dd>Subclass of DerivedTypes for function types.
2910 <ul>
2911 <li><tt>bool isVarArg() const</tt>: Returns true if its a vararg
2912 function</li>
2913 <li><tt> const Type * getReturnType() const</tt>: Returns the
2914 return type of the function.</li>
2915 <li><tt>const Type * getParamType (unsigned i)</tt>: Returns
2916 the type of the ith parameter.</li>
2917 <li><tt> const unsigned getNumParams() const</tt>: Returns the
2918 number of formal parameters.</li>
2919 </ul>
2920 </dd>
2921 <dt><tt>OpaqueType</tt></dt>
2922 <dd>Sublcass of DerivedType for abstract types. This class
2923 defines no content and is used as a placeholder for some other type. Note
2924 that OpaqueType is used (temporarily) during type resolution for forward
2925 references of types. Once the referenced type is resolved, the OpaqueType
2926 is replaced with the actual type. OpaqueType can also be used for data
2927 abstraction. At link time opaque types can be resolved to actual types
2928 of the same name.</dd>
2929</dl>
2930</div>
2931
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]2932
2933
2934<!-- ======================================================================= -->
2935<div class="doc_subsection">
2936 <a name="Module">The <tt>Module</tt> class</a>
2937</div>
2938
2939<div class="doc_text">
2940
2941<p><tt>#include "<a
2942href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
2943<a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
2944
2945<p>The <tt>Module</tt> class represents the top level structure present in LLVM
2946programs. An LLVM module is effectively either a translation unit of the
2947original program or a combination of several translation units merged by the
2948linker. The <tt>Module</tt> class keeps track of a list of <a
2949href="#Function"><tt>Function</tt></a>s, a list of <a
2950href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
2951href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
2952helpful member functions that try to make common operations easy.</p>
2953
2954</div>
2955
2956<!-- _______________________________________________________________________ -->
2957<div class="doc_subsubsection">
2958 <a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
2959</div>
2960
2961<div class="doc_text">
2962
2963<ul>
2964 <li><tt>Module::Module(std::string name = "")</tt></li>
2965</ul>
2966
2967<p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
2968provide a name for it (probably based on the name of the translation unit).</p>
2969
2970<ul>
2971 <li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
2972 <tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
2973
2974 <tt>begin()</tt>, <tt>end()</tt>
2975 <tt>size()</tt>, <tt>empty()</tt>
2976
2977 <p>These are forwarding methods that make it easy to access the contents of
2978 a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
2979 list.</p></li>
2980
2981 <li><tt>Module::FunctionListType &amp;getFunctionList()</tt>
2982
2983 <p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
2984 necessary to use when you need to update the list or perform a complex
2985 action that doesn't have a forwarding method.</p>
2986
2987 <p><!-- Global Variable --></p></li>
2988</ul>
2989
2990<hr>
2991
2992<ul>
2993 <li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
2994
2995 <tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
2996
2997 <tt>global_begin()</tt>, <tt>global_end()</tt>
2998 <tt>global_size()</tt>, <tt>global_empty()</tt>
2999
3000 <p> These are forwarding methods that make it easy to access the contents of
3001 a <tt>Module</tt> object's <a
3002 href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
3003
3004 <li><tt>Module::GlobalListType &amp;getGlobalList()</tt>
3005
3006 <p>Returns the list of <a
3007 href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
3008 use when you need to update the list or perform a complex action that
3009 doesn't have a forwarding method.</p>
3010
3011 <p><!-- Symbol table stuff --> </p></li>
3012</ul>
3013
3014<hr>
3015
3016<ul>
3017 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
3018
3019 <p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3020 for this <tt>Module</tt>.</p>
3021
3022 <p><!-- Convenience methods --></p></li>
3023</ul>
3024
3025<hr>
3026
3027<ul>
3028 <li><tt><a href="#Function">Function</a> *getFunction(const std::string
3029 &amp;Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt>
3030
3031 <p>Look up the specified function in the <tt>Module</tt> <a
3032 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
3033 <tt>null</tt>.</p></li>
3034
3035 <li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
3036 std::string &amp;Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
3037
3038 <p>Look up the specified function in the <tt>Module</tt> <a
3039 href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
3040 external declaration for the function and return it.</p></li>
3041
3042 <li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
3043
3044 <p>If there is at least one entry in the <a
3045 href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
3046 href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
3047 string.</p></li>
3048
3049 <li><tt>bool addTypeName(const std::string &amp;Name, const <a
3050 href="#Type">Type</a> *Ty)</tt>
3051
3052 <p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
3053 mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
3054 name, true is returned and the <a
3055 href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
3056</ul>
3057
3058</div>
3059
3060
Reid Spencer303c4b42007-01-12 17:26:25 +0000[diff] [blame]3061<!-- ======================================================================= -->
3062<div class="doc_subsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3063 <a name="Value">The <tt>Value</tt> class</a>
3064</div>
3065
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3066<div class="doc_text">
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3067
3068<p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
3069<br>
Chris Lattner00815172007-01-04 22:01:45 +0000[diff] [blame]3070doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3071
3072<p>The <tt>Value</tt> class is the most important class in the LLVM Source
3073base. It represents a typed value that may be used (among other things) as an
3074operand to an instruction. There are many different types of <tt>Value</tt>s,
3075such as <a href="#Constant"><tt>Constant</tt></a>s,<a
3076href="#Argument"><tt>Argument</tt></a>s. Even <a
3077href="#Instruction"><tt>Instruction</tt></a>s and <a
3078href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
3079
3080<p>A particular <tt>Value</tt> may be used many times in the LLVM representation
3081for a program. For example, an incoming argument to a function (represented
3082with an instance of the <a href="#Argument">Argument</a> class) is "used" by
3083every instruction in the function that references the argument. To keep track
3084of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
3085href="#User"><tt>User</tt></a>s that is using it (the <a
3086href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
3087graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
3088def-use information in the program, and is accessible through the <tt>use_</tt>*
3089methods, shown below.</p>
3090
3091<p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
3092and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
3093method. In addition, all LLVM values can be named. The "name" of the
3094<tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
3095
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]3096<div class="doc_code">
3097<pre>
Reid Spencer06565dc2007-01-12 17:11:23 +0000[diff] [blame]3098%<b>foo</b> = add i32 1, 2
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]3099</pre>
3100</div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3101
Duncan Sands8036ca42007-03-30 12:22:09 +0000[diff] [blame]3102<p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3103that the name of any value may be missing (an empty string), so names should
3104<b>ONLY</b> be used for debugging (making the source code easier to read,
3105debugging printouts), they should not be used to keep track of values or map
3106between them. For this purpose, use a <tt>std::map</tt> of pointers to the
3107<tt>Value</tt> itself instead.</p>
3108
3109<p>One important aspect of LLVM is that there is no distinction between an SSA
3110variable and the operation that produces it. Because of this, any reference to
3111the value produced by an instruction (or the value available as an incoming
Chris Lattnerd5fc4fc2004-03-18 14:58:55 +0000[diff] [blame]3112argument, for example) is represented as a direct pointer to the instance of
3113the class that
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3114represents this value. Although this may take some getting used to, it
3115simplifies the representation and makes it easier to manipulate.</p>
3116
3117</div>
3118
3119<!-- _______________________________________________________________________ -->
3120<div class="doc_subsubsection">
3121 <a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
3122</div>
3123
3124<div class="doc_text">
3125
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3126<ul>
3127 <li><tt>Value::use_iterator</tt> - Typedef for iterator over the
3128use-list<br>
3129 <tt>Value::use_const_iterator</tt> - Typedef for const_iterator over
3130the use-list<br>
3131 <tt>unsigned use_size()</tt> - Returns the number of users of the
3132value.<br>
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]3133 <tt>bool use_empty()</tt> - Returns true if there are no users.<br>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3134 <tt>use_iterator use_begin()</tt> - Get an iterator to the start of
3135the use-list.<br>
3136 <tt>use_iterator use_end()</tt> - Get an iterator to the end of the
3137use-list.<br>
3138 <tt><a href="#User">User</a> *use_back()</tt> - Returns the last
3139element in the list.
3140 <p> These methods are the interface to access the def-use
3141information in LLVM. As with all other iterators in LLVM, the naming
3142conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3143 </li>
3144 <li><tt><a href="#Type">Type</a> *getType() const</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3145 <p>This method returns the Type of the Value.</p>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3146 </li>
3147 <li><tt>bool hasName() const</tt><br>
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]3148 <tt>std::string getName() const</tt><br>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3149 <tt>void setName(const std::string &amp;Name)</tt>
3150 <p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
3151be aware of the <a href="#nameWarning">precaution above</a>.</p>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3152 </li>
3153 <li><tt>void replaceAllUsesWith(Value *V)</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3154
3155 <p>This method traverses the use list of a <tt>Value</tt> changing all <a
3156 href="#User"><tt>User</tt>s</a> of the current value to refer to
3157 "<tt>V</tt>" instead. For example, if you detect that an instruction always
3158 produces a constant value (for example through constant folding), you can
3159 replace all uses of the instruction with the constant like this:</p>
3160
Bill Wendling3cd5ca62006-10-11 06:30:10 +0000[diff] [blame]3161<div class="doc_code">
3162<pre>
3163Inst-&gt;replaceAllUsesWith(ConstVal);
3164</pre>
3165</div>
3166
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3167</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3168
3169</div>
3170
3171<!-- ======================================================================= -->
3172<div class="doc_subsection">
3173 <a name="User">The <tt>User</tt> class</a>
3174</div>
3175
3176<div class="doc_text">
3177
3178<p>
3179<tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
Misha Brukman384047f2004-06-03 23:29:12 +0000[diff] [blame]3180doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3181Superclass: <a href="#Value"><tt>Value</tt></a></p>
3182
3183<p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
3184refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
3185that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
3186referring to. The <tt>User</tt> class itself is a subclass of
3187<tt>Value</tt>.</p>
3188
3189<p>The operands of a <tt>User</tt> point directly to the LLVM <a
3190href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
3191Single Assignment (SSA) form, there can only be one definition referred to,
3192allowing this direct connection. This connection provides the use-def
3193information in LLVM.</p>
3194
3195</div>
3196
3197<!-- _______________________________________________________________________ -->
3198<div class="doc_subsubsection">
3199 <a name="m_User">Important Public Members of the <tt>User</tt> class</a>
3200</div>
3201
3202<div class="doc_text">
3203
3204<p>The <tt>User</tt> class exposes the operand list in two ways: through
3205an index access interface and through an iterator based interface.</p>
3206
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3207<ul>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3208 <li><tt>Value *getOperand(unsigned i)</tt><br>
3209 <tt>unsigned getNumOperands()</tt>
3210 <p> These two methods expose the operands of the <tt>User</tt> in a
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3211convenient form for direct access.</p></li>
3212
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3213 <li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
3214list<br>
Chris Lattner58360822005-01-17 00:12:04 +0000[diff] [blame]3215 <tt>op_iterator op_begin()</tt> - Get an iterator to the start of
3216the operand list.<br>
3217 <tt>op_iterator op_end()</tt> - Get an iterator to the end of the
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3218operand list.
3219 <p> Together, these methods make up the iterator based interface to
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3220the operands of a <tt>User</tt>.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3221</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3222
3223</div>
3224
3225<!-- ======================================================================= -->
3226<div class="doc_subsection">
3227 <a name="Instruction">The <tt>Instruction</tt> class</a>
3228</div>
3229
3230<div class="doc_text">
3231
3232<p><tt>#include "</tt><tt><a
3233href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
Misha Brukman31ca1de2004-06-03 23:35:54 +0000[diff] [blame]3234doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3235Superclasses: <a href="#User"><tt>User</tt></a>, <a
3236href="#Value"><tt>Value</tt></a></p>
3237
3238<p>The <tt>Instruction</tt> class is the common base class for all LLVM
3239instructions. It provides only a few methods, but is a very commonly used
3240class. The primary data tracked by the <tt>Instruction</tt> class itself is the
3241opcode (instruction type) and the parent <a
3242href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
3243into. To represent a specific type of instruction, one of many subclasses of
3244<tt>Instruction</tt> are used.</p>
3245
3246<p> Because the <tt>Instruction</tt> class subclasses the <a
3247href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
3248way as for other <a href="#User"><tt>User</tt></a>s (with the
3249<tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
3250<tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
3251the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
3252file contains some meta-data about the various different types of instructions
3253in LLVM. It describes the enum values that are used as opcodes (for example
Reid Spencerc92d25d2006-12-19 19:47:19 +0000[diff] [blame]3254<tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3255concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
3256example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
Reid Spencerc92d25d2006-12-19 19:47:19 +0000[diff] [blame]3257href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3258this file confuses doxygen, so these enum values don't show up correctly in the
Misha Brukman31ca1de2004-06-03 23:35:54 +0000[diff] [blame]3259<a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3260
3261</div>
3262
3263<!-- _______________________________________________________________________ -->
3264<div class="doc_subsubsection">
Reid Spencerc92d25d2006-12-19 19:47:19 +0000[diff] [blame]3265 <a name="s_Instruction">Important Subclasses of the <tt>Instruction</tt>
3266 class</a>
3267</div>
3268<div class="doc_text">
3269 <ul>
3270 <li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
3271 <p>This subclasses represents all two operand instructions whose operands
3272 must be the same type, except for the comparison instructions.</p></li>
3273 <li><tt><a name="CastInst">CastInst</a></tt>
3274 <p>This subclass is the parent of the 12 casting instructions. It provides
3275 common operations on cast instructions.</p>
3276 <li><tt><a name="CmpInst">CmpInst</a></tt>
3277 <p>This subclass respresents the two comparison instructions,
3278 <a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
3279 <a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
3280 <li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
3281 <p>This subclass is the parent of all terminator instructions (those which
3282 can terminate a block).</p>
3283 </ul>
3284 </div>
3285
3286<!-- _______________________________________________________________________ -->
3287<div class="doc_subsubsection">
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3288 <a name="m_Instruction">Important Public Members of the <tt>Instruction</tt>
3289 class</a>
3290</div>
3291
3292<div class="doc_text">
3293
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3294<ul>
3295 <li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3296 <p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
3297this <tt>Instruction</tt> is embedded into.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3298 <li><tt>bool mayWriteToMemory()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3299 <p>Returns true if the instruction writes to memory, i.e. it is a
3300 <tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3301 <li><tt>unsigned getOpcode()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3302 <p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3303 <li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3304 <p>Returns another instance of the specified instruction, identical
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3305in all ways to the original except that the instruction has no parent
3306(ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3307and it has no name</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3308</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3309
3310</div>
3311
3312<!-- ======================================================================= -->
3313<div class="doc_subsection">
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3314 <a name="Constant">The <tt>Constant</tt> class and subclasses</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3315</div>
3316
3317<div class="doc_text">
3318
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3319<p>Constant represents a base class for different types of constants. It
3320is subclassed by ConstantInt, ConstantArray, etc. for representing
3321the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
3322a subclass, which represents the address of a global variable or function.
3323</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3324
3325</div>
3326
3327<!-- _______________________________________________________________________ -->
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3328<div class="doc_subsubsection">Important Subclasses of Constant </div>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3329<div class="doc_text">
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3330<ul>
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3331 <li>ConstantInt : This subclass of Constant represents an integer constant of
3332 any width.
3333 <ul>
Reid Spencer97b4ee32007-03-01 21:05:33 +0000[diff] [blame]3334 <li><tt>const APInt&amp; getValue() const</tt>: Returns the underlying
3335 value of this constant, an APInt value.</li>
3336 <li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
3337 value to an int64_t via sign extension. If the value (not the bit width)
3338 of the APInt is too large to fit in an int64_t, an assertion will result.
3339 For this reason, use of this method is discouraged.</li>
3340 <li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
3341 value to a uint64_t via zero extension. IF the value (not the bit width)
3342 of the APInt is too large to fit in a uint64_t, an assertion will result.
Reid Spencer4474d872007-03-02 01:31:31 +0000[diff] [blame]3343 For this reason, use of this method is discouraged.</li>
Reid Spencer97b4ee32007-03-01 21:05:33 +0000[diff] [blame]3344 <li><tt>static ConstantInt* get(const APInt&amp; Val)</tt>: Returns the
3345 ConstantInt object that represents the value provided by <tt>Val</tt>.
3346 The type is implied as the IntegerType that corresponds to the bit width
3347 of <tt>Val</tt>.</li>
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3348 <li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
3349 Returns the ConstantInt object that represents the value provided by
3350 <tt>Val</tt> for integer type <tt>Ty</tt>.</li>
3351 </ul>
3352 </li>
3353 <li>ConstantFP : This class represents a floating point constant.
3354 <ul>
3355 <li><tt>double getValue() const</tt>: Returns the underlying value of
3356 this constant. </li>
3357 </ul>
3358 </li>
3359 <li>ConstantArray : This represents a constant array.
3360 <ul>
3361 <li><tt>const std::vector&lt;Use&gt; &amp;getValues() const</tt>: Returns
3362 a vector of component constants that makeup this array. </li>
3363 </ul>
3364 </li>
3365 <li>ConstantStruct : This represents a constant struct.
3366 <ul>
3367 <li><tt>const std::vector&lt;Use&gt; &amp;getValues() const</tt>: Returns
3368 a vector of component constants that makeup this array. </li>
3369 </ul>
3370 </li>
3371 <li>GlobalValue : This represents either a global variable or a function. In
3372 either case, the value is a constant fixed address (after linking).
3373 </li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3374</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3375</div>
3376
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3377
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3378<!-- ======================================================================= -->
3379<div class="doc_subsection">
3380 <a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
3381</div>
3382
3383<div class="doc_text">
3384
3385<p><tt>#include "<a
3386href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
Misha Brukman384047f2004-06-03 23:29:12 +0000[diff] [blame]3387doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
3388Class</a><br>
Reid Spencerbe5e85e2006-04-14 14:11:48 +0000[diff] [blame]3389Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
3390<a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3391
3392<p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
3393href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
3394visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
3395Because they are visible at global scope, they are also subject to linking with
3396other globals defined in different translation units. To control the linking
3397process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
3398<tt>GlobalValue</tt>s know whether they have internal or external linkage, as
Reid Spencer8b2da7a2004-07-18 13:10:31 +0000[diff] [blame]3399defined by the <tt>LinkageTypes</tt> enumeration.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3400
3401<p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
3402<tt>static</tt> in C), it is not visible to code outside the current translation
3403unit, and does not participate in linking. If it has external linkage, it is
3404visible to external code, and does participate in linking. In addition to
3405linkage information, <tt>GlobalValue</tt>s keep track of which <a
3406href="#Module"><tt>Module</tt></a> they are currently part of.</p>
3407
3408<p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
3409by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
3410global is always a pointer to its contents. It is important to remember this
3411when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
3412be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
3413subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
Reid Spencer06565dc2007-01-12 17:11:23 +0000[diff] [blame]3414i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3415the address of the first element of this array and the value of the
3416<tt>GlobalVariable</tt> are the same, they have different types. The
Reid Spencer06565dc2007-01-12 17:11:23 +0000[diff] [blame]3417<tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
3418is <tt>i32.</tt> Because of this, accessing a global value requires you to
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3419dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
3420can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
3421Language Reference Manual</a>.</p>
3422
3423</div>
3424
3425<!-- _______________________________________________________________________ -->
3426<div class="doc_subsubsection">
3427 <a name="m_GlobalValue">Important Public Members of the <tt>GlobalValue</tt>
3428 class</a>
3429</div>
3430
3431<div class="doc_text">
3432
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3433<ul>
3434 <li><tt>bool hasInternalLinkage() const</tt><br>
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]3435 <tt>bool hasExternalLinkage() const</tt><br>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3436 <tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
3437 <p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
3438 <p> </p>
3439 </li>
3440 <li><tt><a href="#Module">Module</a> *getParent()</tt>
3441 <p> This returns the <a href="#Module"><tt>Module</tt></a> that the
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3442GlobalValue is currently embedded into.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3443</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3444
3445</div>
3446
3447<!-- ======================================================================= -->
3448<div class="doc_subsection">
3449 <a name="Function">The <tt>Function</tt> class</a>
3450</div>
3451
3452<div class="doc_text">
3453
3454<p><tt>#include "<a
3455href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
Misha Brukman31ca1de2004-06-03 23:35:54 +0000[diff] [blame]3456info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
Reid Spencerbe5e85e2006-04-14 14:11:48 +0000[diff] [blame]3457Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3458<a href="#Constant"><tt>Constant</tt></a>,
3459<a href="#User"><tt>User</tt></a>,
3460<a href="#Value"><tt>Value</tt></a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3461
3462<p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
3463actually one of the more complex classes in the LLVM heirarchy because it must
3464keep track of a large amount of data. The <tt>Function</tt> class keeps track
Reid Spencerbe5e85e2006-04-14 14:11:48 +0000[diff] [blame]3465of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
3466<a href="#Argument"><tt>Argument</tt></a>s, and a
3467<a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3468
3469<p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
3470commonly used part of <tt>Function</tt> objects. The list imposes an implicit
3471ordering of the blocks in the function, which indicate how the code will be
3472layed out by the backend. Additionally, the first <a
3473href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
3474<tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
3475block. There are no implicit exit nodes, and in fact there may be multiple exit
3476nodes from a single <tt>Function</tt>. If the <a
3477href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
3478the <tt>Function</tt> is actually a function declaration: the actual body of the
3479function hasn't been linked in yet.</p>
3480
3481<p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
3482<tt>Function</tt> class also keeps track of the list of formal <a
3483href="#Argument"><tt>Argument</tt></a>s that the function receives. This
3484container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
3485nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
3486the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
3487
3488<p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
3489LLVM feature that is only used when you have to look up a value by name. Aside
3490from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
3491internally to make sure that there are not conflicts between the names of <a
3492href="#Instruction"><tt>Instruction</tt></a>s, <a
3493href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
3494href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
3495
Reid Spencer8b2da7a2004-07-18 13:10:31 +0000[diff] [blame]3496<p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
3497and therefore also a <a href="#Constant">Constant</a>. The value of the function
3498is its address (after linking) which is guaranteed to be constant.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3499</div>
3500
3501<!-- _______________________________________________________________________ -->
3502<div class="doc_subsubsection">
3503 <a name="m_Function">Important Public Members of the <tt>Function</tt>
3504 class</a>
3505</div>
3506
3507<div class="doc_text">
3508
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3509<ul>
3510 <li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
Chris Lattnerac479e52004-08-04 05:10:48 +0000[diff] [blame]3511 *Ty, LinkageTypes Linkage, const std::string &amp;N = "", Module* Parent = 0)</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3512
3513 <p>Constructor used when you need to create new <tt>Function</tt>s to add
3514 the the program. The constructor must specify the type of the function to
Chris Lattnerac479e52004-08-04 05:10:48 +0000[diff] [blame]3515 create and what type of linkage the function should have. The <a
3516 href="#FunctionType"><tt>FunctionType</tt></a> argument
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3517 specifies the formal arguments and return value for the function. The same
Duncan Sands8036ca42007-03-30 12:22:09 +0000[diff] [blame]3518 <a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3519 create multiple functions. The <tt>Parent</tt> argument specifies the Module
3520 in which the function is defined. If this argument is provided, the function
3521 will automatically be inserted into that module's list of
3522 functions.</p></li>
3523
Chris Lattner62810e32008-11-25 18:34:50 +0000[diff] [blame]3524 <li><tt>bool isDeclaration()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3525
3526 <p>Return whether or not the <tt>Function</tt> has a body defined. If the
3527 function is "external", it does not have a body, and thus must be resolved
3528 by linking with a function defined in a different translation unit.</p></li>
3529
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3530 <li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]3531 <tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3532
Chris Lattner77d69242005-03-15 05:19:20 +0000[diff] [blame]3533 <tt>begin()</tt>, <tt>end()</tt>
3534 <tt>size()</tt>, <tt>empty()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3535
3536 <p>These are forwarding methods that make it easy to access the contents of
3537 a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
3538 list.</p></li>
3539
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3540 <li><tt>Function::BasicBlockListType &amp;getBasicBlockList()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3541
3542 <p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
3543 is necessary to use when you need to update the list or perform a complex
3544 action that doesn't have a forwarding method.</p></li>
3545
Chris Lattner89cc2652005-03-15 04:48:32 +0000[diff] [blame]3546 <li><tt>Function::arg_iterator</tt> - Typedef for the argument list
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3547iterator<br>
Chris Lattner89cc2652005-03-15 04:48:32 +0000[diff] [blame]3548 <tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3549
Chris Lattner77d69242005-03-15 05:19:20 +0000[diff] [blame]3550 <tt>arg_begin()</tt>, <tt>arg_end()</tt>
Chris Lattner89cc2652005-03-15 04:48:32 +0000[diff] [blame]3551 <tt>arg_size()</tt>, <tt>arg_empty()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3552
3553 <p>These are forwarding methods that make it easy to access the contents of
3554 a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
3555 list.</p></li>
3556
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3557 <li><tt>Function::ArgumentListType &amp;getArgumentList()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3558
3559 <p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
3560 necessary to use when you need to update the list or perform a complex
3561 action that doesn't have a forwarding method.</p></li>
3562
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3563 <li><tt><a href="#BasicBlock">BasicBlock</a> &amp;getEntryBlock()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3564
3565 <p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
3566 function. Because the entry block for the function is always the first
3567 block, this returns the first block of the <tt>Function</tt>.</p></li>
3568
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3569 <li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
3570 <tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3571
3572 <p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
3573 <tt>Function</tt> and returns the return type of the function, or the <a
3574 href="#FunctionType"><tt>FunctionType</tt></a> of the actual
3575 function.</p></li>
3576
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3577 <li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3578
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3579 <p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3580 for this <tt>Function</tt>.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3581</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3582
3583</div>
3584
3585<!-- ======================================================================= -->
3586<div class="doc_subsection">
3587 <a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
3588</div>
3589
3590<div class="doc_text">
3591
3592<p><tt>#include "<a
3593href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
3594<br>
Tanya Lattnera3da7772004-06-22 08:02:25 +0000[diff] [blame]3595doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
Reid Spencerbe5e85e2006-04-14 14:11:48 +0000[diff] [blame]3596 Class</a><br>
3597Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
3598<a href="#Constant"><tt>Constant</tt></a>,
3599<a href="#User"><tt>User</tt></a>,
3600<a href="#Value"><tt>Value</tt></a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3601
3602<p>Global variables are represented with the (suprise suprise)
3603<tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
3604subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
3605always referenced by their address (global values must live in memory, so their
Reid Spencerbe5e85e2006-04-14 14:11:48 +0000[diff] [blame]3606"name" refers to their constant address). See
3607<a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
3608variables may have an initial value (which must be a
3609<a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
3610they may be marked as "constant" themselves (indicating that their contents
3611never change at runtime).</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3612</div>
3613
3614<!-- _______________________________________________________________________ -->
3615<div class="doc_subsubsection">
3616 <a name="m_GlobalVariable">Important Public Members of the
3617 <tt>GlobalVariable</tt> class</a>
3618</div>
3619
3620<div class="doc_text">
3621
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3622<ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3623 <li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
3624 isConstant, LinkageTypes&amp; Linkage, <a href="#Constant">Constant</a>
3625 *Initializer = 0, const std::string &amp;Name = "", Module* Parent = 0)</tt>
3626
3627 <p>Create a new global variable of the specified type. If
3628 <tt>isConstant</tt> is true then the global variable will be marked as
3629 unchanging for the program. The Linkage parameter specifies the type of
Duncan Sands667d4b82009-03-07 15:45:40 +0000[diff] [blame]3630 linkage (internal, external, weak, linkonce, appending) for the variable.
3631 If the linkage is InternalLinkage, WeakAnyLinkage, WeakODRLinkage,
3632 LinkOnceAnyLinkage or LinkOnceODRLinkage,&nbsp; then the resultant
3633 global variable will have internal linkage. AppendingLinkage concatenates
3634 together all instances (in different translation units) of the variable
3635 into a single variable but is only applicable to arrays. &nbsp;See
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3636 the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
3637 further details on linkage types. Optionally an initializer, a name, and the
3638 module to put the variable into may be specified for the global variable as
3639 well.</p></li>
3640
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3641 <li><tt>bool isConstant() const</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3642
3643 <p>Returns true if this is a global variable that is known not to
3644 be modified at runtime.</p></li>
3645
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3646 <li><tt>bool hasInitializer()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3647
3648 <p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
3649
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3650 <li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3651
3652 <p>Returns the intial value for a <tt>GlobalVariable</tt>. It is not legal
3653 to call this method if there is no initializer.</p></li>
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3654</ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3655
3656</div>
3657
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3658
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3659<!-- ======================================================================= -->
3660<div class="doc_subsection">
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3661 <a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3662</div>
3663
3664<div class="doc_text">
3665
3666<p><tt>#include "<a
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3667href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
Stefanus Du Toit24e04112009-06-17 21:12:26 +0000[diff] [blame]3668doxygen info: <a href="/doxygen/classllvm_1_1BasicBlock.html">BasicBlock
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3669Class</a><br>
3670Superclass: <a href="#Value"><tt>Value</tt></a></p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3671
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3672<p>This class represents a single entry multiple exit section of the code,
3673commonly known as a basic block by the compiler community. The
3674<tt>BasicBlock</tt> class maintains a list of <a
3675href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
3676Matching the language definition, the last element of this list of instructions
3677is always a terminator instruction (a subclass of the <a
3678href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
3679
3680<p>In addition to tracking the list of instructions that make up the block, the
3681<tt>BasicBlock</tt> class also keeps track of the <a
3682href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
3683
3684<p>Note that <tt>BasicBlock</tt>s themselves are <a
3685href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
3686like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
3687<tt>label</tt>.</p>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3688
3689</div>
3690
3691<!-- _______________________________________________________________________ -->
3692<div class="doc_subsubsection">
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3693 <a name="m_BasicBlock">Important Public Members of the <tt>BasicBlock</tt>
3694 class</a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3695</div>
3696
3697<div class="doc_text">
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3698<ul>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3699
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3700<li><tt>BasicBlock(const std::string &amp;Name = "", </tt><tt><a
3701 href="#Function">Function</a> *Parent = 0)</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3702
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3703<p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
3704insertion into a function. The constructor optionally takes a name for the new
3705block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
3706the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
3707automatically inserted at the end of the specified <a
3708href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
3709manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3710
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3711<li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
3712<tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
3713<tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
3714<tt>size()</tt>, <tt>empty()</tt>
3715STL-style functions for accessing the instruction list.
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3716
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3717<p>These methods and typedefs are forwarding functions that have the same
3718semantics as the standard library methods of the same names. These methods
3719expose the underlying instruction list of a basic block in a way that is easy to
3720manipulate. To get the full complement of container operations (including
3721operations to update the list), you must use the <tt>getInstList()</tt>
3722method.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3723
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3724<li><tt>BasicBlock::InstListType &amp;getInstList()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3725
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3726<p>This method is used to get access to the underlying container that actually
3727holds the Instructions. This method must be used when there isn't a forwarding
3728function in the <tt>BasicBlock</tt> class for the operation that you would like
3729to perform. Because there are no forwarding functions for "updating"
3730operations, you need to use this if you want to update the contents of a
3731<tt>BasicBlock</tt>.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3732
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3733<li><tt><a href="#Function">Function</a> *getParent()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3734
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3735<p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
3736embedded into, or a null pointer if it is homeless.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3737
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3738<li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3739
Chris Lattner2b78d962007-02-03 20:02:25 +0000[diff] [blame]3740<p> Returns a pointer to the terminator instruction that appears at the end of
3741the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
3742instruction in the block is not a terminator, then a null pointer is
3743returned.</p></li>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3744
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3745</ul>
3746
3747</div>
3748
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3749
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3750<!-- ======================================================================= -->
3751<div class="doc_subsection">
3752 <a name="Argument">The <tt>Argument</tt> class</a>
3753</div>
3754
3755<div class="doc_text">
3756
3757<p>This subclass of Value defines the interface for incoming formal
Chris Lattner58360822005-01-17 00:12:04 +0000[diff] [blame]3758arguments to a function. A Function maintains a list of its formal
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3759arguments. An argument has a pointer to the parent Function.</p>
3760
3761</div>
3762
Chris Lattner9355b472002-09-06 02:50:58 +0000[diff] [blame]3763<!-- *********************************************************************** -->
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3764<hr>
3765<address>
3766 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman44408702008-12-11 17:34:48 +0000[diff] [blame]3767 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3768 <a href="http://validator.w3.org/check/referer"><img
Gabor Greifa9c0f2b2008-06-18 14:05:31 +0000[diff] [blame]3769 src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01 Strict"></a>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3770
3771 <a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
3772 <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
Reid Spencer05fe4b02006-03-14 05:39:39 +0000[diff] [blame]3773 <a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
Misha Brukman13fd15c2004-01-15 00:14:41 +0000[diff] [blame]3774 Last modified: $Date$
3775</address>
3776
Chris Lattner261efe92003-11-25 01:02:51 +0000[diff] [blame]3777</body>
3778</html>