Blame - docs/ProgrammersManual.html - platform/external/llvm

blob: 1931a680607f31db12a2df4462b661f41acb6b21 [file] [log] [blame]

Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1	<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
				2	"http://www.w3.org/TR/html4/strict.dtd">
				3	<html>
				4	<head>
Owen Anderson	60f8757	2009-06-16 17:40:28 +0000	[diff] [blame]	5	<meta http-equiv="Content-type" content="text/html;charset=UTF-8">
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	6	<title>LLVM Programmer's Manual</title>
				7	<link rel="stylesheet" href="llvm.css" type="text/css">
				8	</head>
				9	<body>
				10
				11	<div class="doc_title">
				12	LLVM Programmer's Manual
				13	</div>
				14
				15	<ol>
				16	<li><a href="#introduction">Introduction</a></li>
				17	<li><a href="#general">General Information</a>
				18	<ul>
				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>
				24
				25	-->
				26	</ul>
				27	</li>
				28	<li><a href="#apis">Important and useful LLVM APIs</a>
				29	<ul>
				30	<li><a href="#isa">The <tt>isa<></tt>, <tt>cast<></tt>
				31	and <tt>dyn_cast<></tt> templates</a> </li>
Daniel Dunbar	e3572ba	2009-07-25 04:41:11 +0000	[diff] [blame]	32	<li><a href="#string_apis">Passing strings (the <tt>StringRef</tt>
				33	and <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>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	38	<li><a href="#DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt>
				39	option</a>
				40	<ul>
				41	<li><a href="#DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt>
				42	and the <tt>-debug-only</tt> option</a> </li>
				43	</ul>
				44	</li>
				45	<li><a href="#Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
				46	option</a></li>
				47	<!--
				48	<li>The <tt>InstVisitor</tt> template
				49	<li>The general graph API
				50	-->
				51	<li><a href="#ViewGraph">Viewing graphs while debugging code</a></li>
				52	</ul>
				53	</li>
				54	<li><a href="#datastructure">Picking the Right Data Structure for a Task</a>
				55	<ul>
				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"><vector></a></li>
				62	<li><a href="#dss_deque"><deque></a></li>
				63	<li><a href="#dss_list"><list></a></li>
Gabor Greif	bb17f65	2009-02-27 11:37:41 +0000	[diff] [blame]	64	<li><a href="#dss_ilist">llvm/ADT/ilist.h</a></li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	65	<li><a href="#dss_other">Other Sequential Container Options</a></li>
				66	</ul></li>
				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 Lattner	77ab46d	2007-09-30 00:58:59 +0000	[diff] [blame]	72	<li><a href="#dss_denseset">"llvm/ADT/DenseSet.h"</a></li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	73	<li><a href="#dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a></li>
				74	<li><a href="#dss_set"><set></a></li>
				75	<li><a href="#dss_setvector">"llvm/ADT/SetVector.h"</a></li>
				76	<li><a href="#dss_uniquevector">"llvm/ADT/UniqueVector.h"</a></li>
				77	<li><a href="#dss_otherset">Other Set-Like ContainerOptions</a></li>
				78	</ul></li>
				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>
				82	<li><a href="#dss_stringmap">"llvm/ADT/StringMap.h"</a></li>
				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"><map></a></li>
				86	<li><a href="#dss_othermap">Other Map-Like Container Options</a></li>
				87	</ul></li>
Daniel Berlin	7ea44dc	2007-09-24 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>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	93	</ul>
				94	</li>
				95	<li><a href="#common">Helpful Hints for Common Operations</a>
				96	<ul>
				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
				100	in a <tt>Function</tt></a> </li>
				101	<li><a href="#iterate_basicblock">Iterating over the <tt>Instruction</tt>s
				102	in a <tt>BasicBlock</tt></a> </li>
				103	<li><a href="#iterate_institer">Iterating over the <tt>Instruction</tt>s
				104	in a <tt>Function</tt></a> </li>
				105	<li><a href="#iterate_convert">Turning an iterator into a
				106	class pointer</a> </li>
				107	<li><a href="#iterate_complex">Finding call sites: a more
				108	complex example</a> </li>
				109	<li><a href="#calls_and_invokes">Treating calls and invokes
				110	the same way</a> </li>
				111	<li><a href="#iterate_chains">Iterating over def-use &
				112	use-def chains</a> </li>
Chris Lattner	0665e1f	2008-01-03 16:56:04 +0000	[diff] [blame]	113	<li><a href="#iterate_preds">Iterating over predecessors &
				114	successors of blocks</a></li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +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>
				123	with another <tt>Value</tt></a> </li>
				124	<li><a href="#schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a> </li>
				125	</ul>
				126	</li>
Jeffrey Yasskin	7ebb6ac	2009-04-30 22:33:41 +0000	[diff] [blame]	127	<li><a href="#create_types">How to Create Types</a></li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +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>
				135	-->
				136	</ul>
				137	</li>
				138
Owen Anderson	e8370c0	2009-06-16 01:17:16 +0000	[diff] [blame]	139	<li><a href="#threading">Threads and LLVM</a>
				140	<ul>
Owen Anderson	b418690	2009-06-16 18:04:19 +0000	[diff] [blame]	141	<li><a href="#startmultithreaded">Entering and Exiting Multithreaded Mode
				142	</a></li>
Owen Anderson	e8370c0	2009-06-16 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
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	148	<li><a href="#advanced">Advanced Topics</a>
				149	<ul>
				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 Greif	92e8776	2008-06-16 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>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	160	</ul></li>
				161
				162	<li><a href="#coreclasses">The Core LLVM Class Hierarchy Reference</a>
				163	<ul>
				164	<li><a href="#Type">The <tt>Type</tt> class</a> </li>
				165	<li><a href="#Module">The <tt>Module</tt> class</a></li>
				166	<li><a href="#Value">The <tt>Value</tt> class</a>
				167	<ul>
				168	<li><a href="#User">The <tt>User</tt> class</a>
				169	<ul>
				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>
				174	<ul>
				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>
				181	</ul>
				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>
				186	</li>
				187	</ul>
				188	</li>
				189	</ol>
				190
				191	<div class="doc_author">
				192	<p>Written by <a href="mailto:sabre@nondot.org">Chris Lattner</a>,
				193	<a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a>,
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	194	<a href="mailto:ggreif@gmail.com">Gabor Greif</a>,
Owen Anderson	e8370c0	2009-06-16 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>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	198	</div>
				199
				200	<!-- *********************************************************************** -->
				201	<div class="doc_section">
				202	<a name="introduction">Introduction </a>
				203	</div>
				204	<!-- *********************************************************************** -->
				205
				206	<div class="doc_text">
				207
				208	<p>This document is meant to highlight some of the important classes and
				209	interfaces available in the LLVM source-base. This manual is not
				210	intended to explain what LLVM is, how it works, and what LLVM code looks
				211	like. It assumes that you know the basics of LLVM and are interested
				212	in writing transformations or otherwise analyzing or manipulating the
				213	code.</p>
				214
				215	<p>This document should get you oriented so that you can find your
				216	way in the continuously growing source code that makes up the LLVM
				217	infrastructure. Note that this manual is not intended to serve as a
				218	replacement for reading the source code, so if you think there should be
				219	a method in one of these classes to do something, but it's not listed,
				220	check the source. Links to the <a href="/doxygen/">doxygen</a> sources
				221	are provided to make this as easy as possible.</p>
				222
				223	<p>The first section of this document describes general information that is
				224	useful to know when working in the LLVM infrastructure, and the second describes
				225	the Core LLVM classes. In the future this manual will be extended with
				226	information describing how to use extension libraries, such as dominator
				227	information, CFG traversal routines, and useful utilities like the <tt><a
				228	href="/doxygen/InstVisitor_8h-source.html">InstVisitor</a></tt> template.</p>
				229
				230	</div>
				231
				232	<!-- *********************************************************************** -->
				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
				241	in 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),
				253	perhaps much more than you are used to, or have seen before. Because of
				254	this, you might want to do a little background reading in the
				255	techniques used and capabilities of the library. There are many good
				256	pages that discuss the STL, and several books on the subject that you
				257	can 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
				264	reference</a> - an excellent reference for the STL and other parts of the
				265	standard C++ library.</li>
				266
				267	<li><a href="http://www.tempest-sw.com/cpp/">C++ In a Nutshell</a> - This is an
Gabor Greif	e39f0e7	2009-03-12 09:47:03 +0000	[diff] [blame]	268	O'Reilly book in the making. It has a decent Standard Library
				269	Reference that rivals Dinkumware's, and is unfortunately no longer free since the
				270	book has been published.</li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	271
				272	<li><a href="http://www.parashift.com/c++-faq-lite/">C++ Frequently Asked
				273	Questions</a></li>
				274
				275	<li><a href="http://www.sgi.com/tech/stl/">SGI's STL Programmer's Guide</a> -
				276	Contains a useful <a
				277	href="http://www.sgi.com/tech/stl/stl_introduction.html">Introduction to the
				278	STL</a>.</li>
				279
				280	<li><a href="http://www.research.att.com/%7Ebs/C++.html">Bjarne Stroustrup's C++
				281	Page</a></li>
				282
				283	<li><a href="http://64.78.49.204/">
				284	Bruce Eckel's Thinking in C++, 2nd ed. Volume 2 Revision 4.0 (even better, get
				285	the book).</a></li>
				286
				287	</ol>
				288
				289	<p>You are also encouraged to take a look at the <a
				290	href="CodingStandards.html">LLVM Coding Standards</a> guide which focuses on how
				291	to 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
				302	<ol>
				303	<li><a href="http://www.psc.edu/%7Esemke/cvs_branches.html">CVS
				304	Branch and Tag Primer</a></li>
				305	<li><a href="http://www.fortran-2000.com/ArnaudRecipes/sharedlib.html">Using
				306	static and shared libraries across platforms</a></li>
				307	</ol>
				308
				309	</div>
				310
				311	<!-- *********************************************************************** -->
				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
				320	know about when writing transformations.</p>
				321
				322	</div>
				323
				324	<!-- ======================================================================= -->
				325	<div class="doc_subsection">
				326	<a name="isa">The <tt>isa<></tt>, <tt>cast<></tt> and
				327	<tt>dyn_cast<></tt> templates</a>
				328	</div>
				329
				330	<div class="doc_text">
				331
				332	<p>The LLVM source-base makes extensive use of a custom form of RTTI.
				333	These templates have many similarities to the C++ <tt>dynamic_cast<></tt>
				334	operator, but they don't have some drawbacks (primarily stemming from
				335	the fact that <tt>dynamic_cast<></tt> only works on classes that
				336	have a v-table). Because they are used so often, you must know what they
				337	do and how they work. All of these templates are defined in the <a
				338	href="/doxygen/Casting_8h-source.html"><tt>llvm/Support/Casting.h</tt></a>
				339	file (note that you very rarely have to include this file directly).</p>
				340
				341	<dl>
				342	<dt><tt>isa<></tt>: </dt>
				343
				344	<dd><p>The <tt>isa<></tt> operator works exactly like the Java
				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
				347	be very useful for constraint checking of various sorts (example below).</p>
				348	</dd>
				349
				350	<dt><tt>cast<></tt>: </dt>
				351
				352	<dd><p>The <tt>cast<></tt> operator is a "checked cast" operation. It
Chris Lattner	1d5610a	2008-06-20 05:03:17 +0000	[diff] [blame]	353	converts a pointer or reference from a base class to a derived class, causing
Dan Gohman	f17a25c	2007-07-18 16:29:46 +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<></tt>
				357	and <tt>cast<></tt> template is:</p>
				358
				359	<div class="doc_code">
				360	<pre>
				361	static bool isLoopInvariant(const <a href="#Value">Value</a> V, const Loop L) {
				362	if (isa<<a href="#Constant">Constant</a>>(V) \|\| isa<<a href="#Argument">Argument</a>>(V) \|\| isa<<a href="#GlobalValue">GlobalValue</a>>(V))
				363	return true;
				364
				365	// <i>Otherwise, it must be an instruction...</i>
				366	return !L->contains(cast<<a href="#Instruction">Instruction</a>>(V)->getParent());
				367	}
				368	</pre>
				369	</div>
				370
				371	<p>Note that you should <b>not</b> use an <tt>isa<></tt> test followed
				372	by a <tt>cast<></tt>, for that use the <tt>dyn_cast<></tt>
				373	operator.</p>
				374
				375	</dd>
				376
				377	<dt><tt>dyn_cast<></tt>:</dt>
				378
				379	<dd><p>The <tt>dyn_cast<></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
				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
				383	much like the <tt>dynamic_cast<></tt> operator in C++, and should be
				384	used in the same circumstances. Typically, the <tt>dyn_cast<></tt>
				385	operator is used in an <tt>if</tt> statement or some other flow control
				386	statement like this:</p>
				387
				388	<div class="doc_code">
				389	<pre>
				390	if (<a href="#AllocationInst">AllocationInst</a> *AI = dyn_cast<<a href="#AllocationInst">AllocationInst</a>>(Val)) {
				391	// <i>...</i>
				392	}
				393	</pre>
				394	</div>
				395
				396	<p>This form of the <tt>if</tt> statement effectively combines together a call
				397	to <tt>isa<></tt> and a call to <tt>cast<></tt> into one
				398	statement, which is very convenient.</p>
				399
				400	<p>Note that the <tt>dyn_cast<></tt> operator, like C++'s
				401	<tt>dynamic_cast<></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>
				406
				407	</dd>
				408
				409	<dt><tt>cast_or_null<></tt>: </dt>
				410
				411	<dd><p>The <tt>cast_or_null<></tt> operator works just like the
				412	<tt>cast<></tt> operator, except that it allows for a null pointer as an
				413	argument (which it then propagates). This can sometimes be useful, allowing
				414	you to combine several null checks into one.</p></dd>
				415
				416	<dt><tt>dyn_cast_or_null<></tt>: </dt>
				417
				418	<dd><p>The <tt>dyn_cast_or_null<></tt> operator works just like the
				419	<tt>dyn_cast<></tt> operator, except that it allows for a null pointer
				420	as an argument (which it then propagates). This can sometimes be useful,
				421	allowing you to combine several null checks into one.</p></dd>
				422
				423	</dl>
				424
				425	<p>These five templates can be used with any classes, whether they have a
				426	v-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
				428	to. Describing this is currently outside the scope of this document, but there
				429	are lots of examples in the LLVM source base.</p>
				430
				431	</div>
				432
Daniel Dunbar	e3572ba	2009-07-25 04:41:11 +0000	[diff] [blame]	433
				434	<!-- ======================================================================= -->
				435	<div class="doc_subsection">
				436	<a name="string_apis">Passing strings (the <tt>StringRef</tt>
				437	and <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
				443	several important APIs which take string. Several important examples are the
				444	Value class -- which has names for instructions, functions, etc. -- and the
				445	StringMap 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
				448	may have embedded null characters. Therefore, they cannot simply take
				449	a <tt>const char *</tt>, and taking a <tt>const std::string&</tt> requires
				450	clients to perform a heap allocation which is usually unnecessary. Instead,
				451	many LLVM APIs use a <tt>const StringRef&</tt> or a <tt>const Twine&</tt> for
				452	passing strings efficiently.</p>
				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
				465	on <tt>std:string</tt>, but does not require heap allocation.</p>
				466
				467	It can be implicitly constructed using either a C style null-terminated string
				468	or an <tt>std::string</tt>, or explicitly with a character pointer and length.
				469	For example, the <tt>StringRef</tt> find function is declared as:</p>
				470	<div class="doc_code">
				471	iterator find(const StringRef &Key);
				472	</div>
				473
				474	<p>and clients can call it using any one of:</p>
				475
				476	<div class="doc_code">
				477	<pre>
				478	Map.find("foo"); <i>// Lookup "foo"</i>
				479	Map.find(std::string("bar")); <i>// Lookup "bar"</i>
				480	Map.find(StringRef("\0baz", 4)); <i>// Lookup "\0baz"</i>
				481	</pre>
				482	</div>
				483
				484	<p>Similarly, APIs which need to return a string may return a <tt>StringRef</tt>
				485	instance, which can be used directly or converted to an <tt>std::string</tt>
				486	using the <tt>str</tt> member function. See
				487	"<tt><a href="/doxygen/classllvm_1_1StringRef_8h-source.html">llvm/ADT/StringRef.h</a></tt>"
				488	for more information.</p>
				489
				490	<p>You should rarely use the <tt>StringRef</tt> class directly, because it contains
				491	pointers to external memory it is not generally safe to store an instance of the
				492	class (since the external storage may be freed).</p>
				493
				494	</div>
				495
				496	<!-- _______________________________________________________________________ -->
				497	<div class="doc_subsubsection">
				498	<a name="Twine">The <tt>Twine</tt> class</a>
				499	</div>
				500
				501	<div class="doc_text">
				502
				503	<p>The <tt>Twine</tt> class is an efficient way for APIs to accept concatenated
				504	strings. For example, a common LLVM paradigm is to name one instruction based on
				505	the name of another instruction with a suffix, for example:</p>
				506
				507	<div class="doc_code">
				508	<pre>
				509	New = CmpInst::Create(<i>...</i>, SO->getName() + ".cmp");
				510	</pre>
				511	</div>
				512
				513	<p>The <tt>Twine</tt> class is effectively a
				514	lightweight <a href="http://en.wikipedia.org/wiki/Rope_(computer_science)">rope</a>
				515	which points to temporary (stack allocated) objects. Twines can be implicitly
				516	constructed as the result of the plus operator applied to strings (i.e., a C
				517	strings, an <tt>std::string</tt>, or a <tt>StringRef</tt>). The twine delays the
				518	actual concatentation of strings until it is actually required, at which point
				519	it can be efficiently rendered directly into a character array. This avoids
				520	unnecessary heap allocation involved in constructing the temporary results of
				521	string concatenation. See
				522	"<tt><a href="/doxygen/classllvm_1_1Twine_8h-source.html">llvm/ADT/Twine.h</a></tt>"
				523	for more information.</p></tt>
				524
				525	<p>As with a <tt>StringRef</tt>, <tt>Twine</tt> objects point to external memory
				526	and should almost never be stored or mentioned directly. They are intended
				527	solely for use when defining a function which should be able to efficiently
				528	accept concatenated strings.</p>
				529
				530	</div>
				531
				532
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	533	<!-- ======================================================================= -->
				534	<div class="doc_subsection">
				535	<a name="DEBUG">The <tt>DEBUG()</tt> macro and <tt>-debug</tt> option</a>
				536	</div>
				537
				538	<div class="doc_text">
				539
				540	<p>Often when working on your pass you will put a bunch of debugging printouts
				541	and other code into your pass. After you get it working, you want to remove
				542	it, but you may need it again in the future (to work out new bugs that you run
				543	across).</p>
				544
				545	<p> Naturally, because of this, you don't want to delete the debug printouts,
				546	but you don't want them to always be noisy. A standard compromise is to comment
				547	them out, allowing you to enable them if you need them in the future.</p>
				548
				549	<p>The "<tt><a href="/doxygen/Debug_8h-source.html">llvm/Support/Debug.h</a></tt>"
				550	file provides a macro named <tt>DEBUG()</tt> that is a much nicer solution to
				551	this problem. Basically, you can put arbitrary code into the argument of the
				552	<tt>DEBUG</tt> macro, and it is only executed if '<tt>opt</tt>' (or any other
				553	tool) is run with the '<tt>-debug</tt>' command line argument:</p>
				554
				555	<div class="doc_code">
				556	<pre>
Daniel Dunbar	e32f9b2	2009-07-25 01:55:32 +0000	[diff] [blame]	557	DEBUG(errs() << "I am here!\n");
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	558	</pre>
				559	</div>
				560
				561	<p>Then you can run your pass like this:</p>
				562
				563	<div class="doc_code">
				564	<pre>
				565	$ opt < a.bc > /dev/null -mypass
				566	<i><no output></i>
				567	$ opt < a.bc > /dev/null -mypass -debug
				568	I am here!
				569	</pre>
				570	</div>
				571
				572	<p>Using the <tt>DEBUG()</tt> macro instead of a home-brewed solution allows you
				573	to not have to create "yet another" command line option for the debug output for
				574	your pass. Note that <tt>DEBUG()</tt> macros are disabled for optimized builds,
				575	so they do not cause a performance impact at all (for the same reason, they
				576	should also not contain side-effects!).</p>
				577
				578	<p>One additional nice thing about the <tt>DEBUG()</tt> macro is that you can
				579	enable or disable it directly in gdb. Just use "<tt>set DebugFlag=0</tt>" or
				580	"<tt>set DebugFlag=1</tt>" from the gdb if the program is running. If the
				581	program hasn't been started yet, you can always just run it with
				582	<tt>-debug</tt>.</p>
				583
				584	</div>
				585
				586	<!-- _______________________________________________________________________ -->
				587	<div class="doc_subsubsection">
				588	<a name="DEBUG_TYPE">Fine grained debug info with <tt>DEBUG_TYPE</tt> and
				589	the <tt>-debug-only</tt> option</a>
				590	</div>
				591
				592	<div class="doc_text">
				593
				594	<p>Sometimes you may find yourself in a situation where enabling <tt>-debug</tt>
				595	just turns on <b>too much</b> information (such as when working on the code
				596	generator). If you want to enable debug information with more fine-grained
				597	control, you define the <tt>DEBUG_TYPE</tt> macro and the <tt>-debug</tt> only
				598	option as follows:</p>
				599
				600	<div class="doc_code">
				601	<pre>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	602	#undef DEBUG_TYPE
Daniel Dunbar	e32f9b2	2009-07-25 01:55:32 +0000	[diff] [blame]	603	DEBUG(errs() << "No debug type\n");
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	604	#define DEBUG_TYPE "foo"
Daniel Dunbar	e32f9b2	2009-07-25 01:55:32 +0000	[diff] [blame]	605	DEBUG(errs() << "'foo' debug type\n");
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	606	#undef DEBUG_TYPE
				607	#define DEBUG_TYPE "bar"
Daniel Dunbar	e32f9b2	2009-07-25 01:55:32 +0000	[diff] [blame]	608	DEBUG(errs() << "'bar' debug type\n"));
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	609	#undef DEBUG_TYPE
				610	#define DEBUG_TYPE ""
Daniel Dunbar	e32f9b2	2009-07-25 01:55:32 +0000	[diff] [blame]	611	DEBUG(errs() << "No debug type (2)\n");
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	612	</pre>
				613	</div>
				614
				615	<p>Then you can run your pass like this:</p>
				616
				617	<div class="doc_code">
				618	<pre>
				619	$ opt < a.bc > /dev/null -mypass
				620	<i><no output></i>
				621	$ opt < a.bc > /dev/null -mypass -debug
				622	No debug type
				623	'foo' debug type
				624	'bar' debug type
				625	No debug type (2)
				626	$ opt < a.bc > /dev/null -mypass -debug-only=foo
				627	'foo' debug type
				628	$ opt < a.bc > /dev/null -mypass -debug-only=bar
				629	'bar' debug type
				630	</pre>
				631	</div>
				632
				633	<p>Of course, in practice, you should only set <tt>DEBUG_TYPE</tt> at the top of
				634	a file, to specify the debug type for the entire module (if you do this before
				635	you <tt>#include "llvm/Support/Debug.h"</tt>, you don't have to insert the ugly
				636	<tt>#undef</tt>'s). Also, you should use names more meaningful than "foo" and
				637	"bar", because there is no system in place to ensure that names do not
				638	conflict. If two different modules use the same string, they will all be turned
				639	on when the name is specified. This allows, for example, all debug information
				640	for instruction scheduling to be enabled with <tt>-debug-type=InstrSched</tt>,
				641	even if the source lives in multiple files.</p>
				642
				643	</div>
				644
				645	<!-- ======================================================================= -->
				646	<div class="doc_subsection">
				647	<a name="Statistic">The <tt>Statistic</tt> class & <tt>-stats</tt>
				648	option</a>
				649	</div>
				650
				651	<div class="doc_text">
				652
				653	<p>The "<tt><a
				654	href="/doxygen/Statistic_8h-source.html">llvm/ADT/Statistic.h</a></tt>" file
				655	provides a class named <tt>Statistic</tt> that is used as a unified way to
				656	keep track of what the LLVM compiler is doing and how effective various
				657	optimizations are. It is useful to see what optimizations are contributing to
				658	making a particular program run faster.</p>
				659
				660	<p>Often you may run your pass on some big program, and you're interested to see
				661	how many times it makes a certain transformation. Although you can do this with
				662	hand inspection, or some ad-hoc method, this is a real pain and not very useful
				663	for big programs. Using the <tt>Statistic</tt> class makes it very easy to
				664	keep track of this information, and the calculated information is presented in a
				665	uniform manner with the rest of the passes being executed.</p>
				666
				667	<p>There are many examples of <tt>Statistic</tt> uses, but the basics of using
				668	it are as follows:</p>
				669
				670	<ol>
				671	<li><p>Define your statistic like this:</p>
				672
				673	<div class="doc_code">
				674	<pre>
				675	#define <a href="#DEBUG_TYPE">DEBUG_TYPE</a> "mypassname" <i>// This goes before any #includes.</i>
				676	STATISTIC(NumXForms, "The # of times I did stuff");
				677	</pre>
				678	</div>
				679
				680	<p>The <tt>STATISTIC</tt> macro defines a static variable, whose name is
				681	specified by the first argument. The pass name is taken from the DEBUG_TYPE
				682	macro, and the description is taken from the second argument. The variable
				683	defined ("NumXForms" in this case) acts like an unsigned integer.</p></li>
				684
				685	<li><p>Whenever you make a transformation, bump the counter:</p>
				686
				687	<div class="doc_code">
				688	<pre>
				689	++NumXForms; // <i>I did stuff!</i>
				690	</pre>
				691	</div>
				692
				693	</li>
				694	</ol>
				695
				696	<p>That's all you have to do. To get '<tt>opt</tt>' to print out the
				697	statistics gathered, use the '<tt>-stats</tt>' option:</p>
				698
				699	<div class="doc_code">
				700	<pre>
				701	$ opt -stats -mypassname < program.bc > /dev/null
				702	<i>... statistics output ...</i>
				703	</pre>
				704	</div>
				705
				706	<p> When running <tt>opt</tt> on a C file from the SPEC benchmark
				707	suite, it gives a report that looks like this:</p>
				708
				709	<div class="doc_code">
				710	<pre>
				711	7646 bitcodewriter - Number of normal instructions
				712	725 bitcodewriter - Number of oversized instructions
				713	129996 bitcodewriter - Number of bitcode bytes written
				714	2817 raise - Number of insts DCEd or constprop'd
				715	3213 raise - Number of cast-of-self removed
				716	5046 raise - Number of expression trees converted
				717	75 raise - Number of other getelementptr's formed
				718	138 raise - Number of load/store peepholes
				719	42 deadtypeelim - Number of unused typenames removed from symtab
				720	392 funcresolve - Number of varargs functions resolved
				721	27 globaldce - Number of global variables removed
				722	2 adce - Number of basic blocks removed
				723	134 cee - Number of branches revectored
				724	49 cee - Number of setcc instruction eliminated
				725	532 gcse - Number of loads removed
				726	2919 gcse - Number of instructions removed
				727	86 indvars - Number of canonical indvars added
				728	87 indvars - Number of aux indvars removed
				729	25 instcombine - Number of dead inst eliminate
				730	434 instcombine - Number of insts combined
				731	248 licm - Number of load insts hoisted
				732	1298 licm - Number of insts hoisted to a loop pre-header
				733	3 licm - Number of insts hoisted to multiple loop preds (bad, no loop pre-header)
				734	75 mem2reg - Number of alloca's promoted
				735	1444 cfgsimplify - Number of blocks simplified
				736	</pre>
				737	</div>
				738
				739	<p>Obviously, with so many optimizations, having a unified framework for this
				740	stuff is very nice. Making your pass fit well into the framework makes it more
				741	maintainable and useful.</p>
				742
				743	</div>
				744
				745	<!-- ======================================================================= -->
				746	<div class="doc_subsection">
				747	<a name="ViewGraph">Viewing graphs while debugging code</a>
				748	</div>
				749
				750	<div class="doc_text">
				751
				752	<p>Several of the important data structures in LLVM are graphs: for example
				753	CFGs made out of LLVM <a href="#BasicBlock">BasicBlock</a>s, CFGs made out of
				754	LLVM <a href="CodeGenerator.html#machinebasicblock">MachineBasicBlock</a>s, and
				755	<a href="CodeGenerator.html#selectiondag_intro">Instruction Selection
				756	DAGs</a>. In many cases, while debugging various parts of the compiler, it is
				757	nice to instantly visualize these graphs.</p>
				758
				759	<p>LLVM provides several callbacks that are available in a debug build to do
				760	exactly that. If you call the <tt>Function::viewCFG()</tt> method, for example,
				761	the current LLVM tool will pop up a window containing the CFG for the function
				762	where each basic block is a node in the graph, and each node contains the
				763	instructions in the block. Similarly, there also exists
				764	<tt>Function::viewCFGOnly()</tt> (does not include the instructions), the
				765	<tt>MachineFunction::viewCFG()</tt> and <tt>MachineFunction::viewCFGOnly()</tt>,
				766	and the <tt>SelectionDAG::viewGraph()</tt> methods. Within GDB, for example,
				767	you can usually use something like <tt>call DAG.viewGraph()</tt> to pop
				768	up a window. Alternatively, you can sprinkle calls to these functions in your
				769	code in places you want to debug.</p>
				770
				771	<p>Getting this to work requires a small amount of configuration. On Unix
				772	systems with X11, install the <a href="http://www.graphviz.org">graphviz</a>
				773	toolkit, and make sure 'dot' and 'gv' are in your path. If you are running on
				774	Mac OS/X, download and install the Mac OS/X <a
				775	href="http://www.pixelglow.com/graphviz/">Graphviz program</a>, and add
				776	<tt>/Applications/Graphviz.app/Contents/MacOS/</tt> (or wherever you install
				777	it) to your path. Once in your system and path are set up, rerun the LLVM
				778	configure script and rebuild LLVM to enable this functionality.</p>
				779
				780	<p><tt>SelectionDAG</tt> has been extended to make it easier to locate
				781	<i>interesting</i> nodes in large complex graphs. From gdb, if you
				782	<tt>call DAG.setGraphColor(<i>node</i>, "<i>color</i>")</tt>, then the
				783	next <tt>call DAG.viewGraph()</tt> would highlight the node in the
				784	specified color (choices of colors can be found at <a
				785	href="http://www.graphviz.org/doc/info/colors.html">colors</a>.) More
				786	complex node attributes can be provided with <tt>call
				787	DAG.setGraphAttrs(<i>node</i>, "<i>attributes</i>")</tt> (choices can be
				788	found at <a href="http://www.graphviz.org/doc/info/attrs.html">Graph
				789	Attributes</a>.) If you want to restart and clear all the current graph
				790	attributes, then you can <tt>call DAG.clearGraphAttrs()</tt>. </p>
				791
				792	</div>
				793
				794	<!-- *********************************************************************** -->
				795	<div class="doc_section">
				796	<a name="datastructure">Picking the Right Data Structure for a Task</a>
				797	</div>
				798	<!-- *********************************************************************** -->
				799
				800	<div class="doc_text">
				801
				802	<p>LLVM has a plethora of data structures in the <tt>llvm/ADT/</tt> directory,
				803	and we commonly use STL data structures. This section describes the trade-offs
				804	you should consider when you pick one.</p>
				805
				806	<p>
				807	The first step is a choose your own adventure: do you want a sequential
				808	container, a set-like container, or a map-like container? The most important
				809	thing when choosing a container is the algorithmic properties of how you plan to
				810	access the container. Based on that, you should use:</p>
				811
				812	<ul>
				813	<li>a <a href="#ds_map">map-like</a> container if you need efficient look-up
				814	of an value based on another value. Map-like containers also support
				815	efficient queries for containment (whether a key is in the map). Map-like
				816	containers generally do not support efficient reverse mapping (values to
				817	keys). If you need that, use two maps. Some map-like containers also
				818	support efficient iteration through the keys in sorted order. Map-like
				819	containers are the most expensive sort, only use them if you need one of
				820	these capabilities.</li>
				821
				822	<li>a <a href="#ds_set">set-like</a> container if you need to put a bunch of
				823	stuff into a container that automatically eliminates duplicates. Some
				824	set-like containers support efficient iteration through the elements in
				825	sorted order. Set-like containers are more expensive than sequential
				826	containers.
				827	</li>
				828
				829	<li>a <a href="#ds_sequential">sequential</a> container provides
				830	the most efficient way to add elements and keeps track of the order they are
				831	added to the collection. They permit duplicates and support efficient
				832	iteration, but do not support efficient look-up based on a key.
				833	</li>
				834
Daniel Berlin	7ea44dc	2007-09-24 17:52:25 +0000	[diff] [blame]	835	<li>a <a href="#ds_bit">bit</a> container provides an efficient way to store and
				836	perform set operations on sets of numeric id's, while automatically
				837	eliminating duplicates. Bit containers require a maximum of 1 bit for each
				838	identifier you want to store.
				839	</li>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	840	</ul>
				841
				842	<p>
				843	Once the proper category of container is determined, you can fine tune the
				844	memory use, constant factors, and cache behaviors of access by intelligently
				845	picking a member of the category. Note that constant factors and cache behavior
				846	can be a big deal. If you have a vector that usually only contains a few
				847	elements (but could contain many), for example, it's much better to use
				848	<a href="#dss_smallvector">SmallVector</a> than <a href="#dss_vector">vector</a>
				849	. Doing so avoids (relatively) expensive malloc/free calls, which dwarf the
				850	cost of adding the elements to the container. </p>
				851
				852	</div>
				853
				854	<!-- ======================================================================= -->
				855	<div class="doc_subsection">
				856	<a name="ds_sequential">Sequential Containers (std::vector, std::list, etc)</a>
				857	</div>
				858
				859	<div class="doc_text">
				860	There are a variety of sequential containers available for you, based on your
				861	needs. Pick the first in this section that will do what you want.
				862	</div>
				863
				864	<!-- _______________________________________________________________________ -->
				865	<div class="doc_subsubsection">
				866	<a name="dss_fixedarrays">Fixed Size Arrays</a>
				867	</div>
				868
				869	<div class="doc_text">
				870	<p>Fixed size arrays are very simple and very fast. They are good if you know
				871	exactly how many elements you have, or you have a (low) upper bound on how many
				872	you have.</p>
				873	</div>
				874
				875	<!-- _______________________________________________________________________ -->
				876	<div class="doc_subsubsection">
				877	<a name="dss_heaparrays">Heap Allocated Arrays</a>
				878	</div>
				879
				880	<div class="doc_text">
				881	<p>Heap allocated arrays (new[] + delete[]) are also simple. They are good if
				882	the number of elements is variable, if you know how many elements you will need
				883	before the array is allocated, and if the array is usually large (if not,
				884	consider a <a href="#dss_smallvector">SmallVector</a>). The cost of a heap
				885	allocated array is the cost of the new/delete (aka malloc/free). Also note that
				886	if you are allocating an array of a type with a constructor, the constructor and
				887	destructors will be run for every element in the array (re-sizable vectors only
				888	construct those elements actually used).</p>
				889	</div>
				890
				891	<!-- _______________________________________________________________________ -->
				892	<div class="doc_subsubsection">
				893	<a name="dss_smallvector">"llvm/ADT/SmallVector.h"</a>
				894	</div>
				895
				896	<div class="doc_text">
				897	<p><tt>SmallVector<Type, N></tt> is a simple class that looks and smells
				898	just like <tt>vector<Type></tt>:
				899	it supports efficient iteration, lays out elements in memory order (so you can
				900	do pointer arithmetic between elements), supports efficient push_back/pop_back
				901	operations, supports efficient random access to its elements, etc.</p>
				902
				903	<p>The advantage of SmallVector is that it allocates space for
				904	some number of elements (N) <b>in the object itself</b>. Because of this, if
				905	the SmallVector is dynamically smaller than N, no malloc is performed. This can
				906	be a big win in cases where the malloc/free call is far more expensive than the
				907	code that fiddles around with the elements.</p>
				908
				909	<p>This is good for vectors that are "usually small" (e.g. the number of
				910	predecessors/successors of a block is usually less than 8). On the other hand,
				911	this makes the size of the SmallVector itself large, so you don't want to
				912	allocate lots of them (doing so will waste a lot of space). As such,
				913	SmallVectors are most useful when on the stack.</p>
				914
				915	<p>SmallVector also provides a nice portable and efficient replacement for
				916	<tt>alloca</tt>.</p>
				917
				918	</div>
				919
				920	<!-- _______________________________________________________________________ -->
				921	<div class="doc_subsubsection">
				922	<a name="dss_vector"><vector></a>
				923	</div>
				924
				925	<div class="doc_text">
				926	<p>
				927	std::vector is well loved and respected. It is useful when SmallVector isn't:
				928	when the size of the vector is often large (thus the small optimization will
				929	rarely be a benefit) or if you will be allocating many instances of the vector
				930	itself (which would waste space for elements that aren't in the container).
				931	vector is also useful when interfacing with code that expects vectors :).
				932	</p>
				933
				934	<p>One worthwhile note about std::vector: avoid code like this:</p>
				935
				936	<div class="doc_code">
				937	<pre>
				938	for ( ... ) {
				939	std::vector<foo> V;
				940	use V;
				941	}
				942	</pre>
				943	</div>
				944
				945	<p>Instead, write this as:</p>
				946
				947	<div class="doc_code">
				948	<pre>
				949	std::vector<foo> V;
				950	for ( ... ) {
				951	use V;
				952	V.clear();
				953	}
				954	</pre>
				955	</div>
				956
				957	<p>Doing so will save (at least) one heap allocation and free per iteration of
				958	the loop.</p>
				959
				960	</div>
				961
				962	<!-- _______________________________________________________________________ -->
				963	<div class="doc_subsubsection">
				964	<a name="dss_deque"><deque></a>
				965	</div>
				966
				967	<div class="doc_text">
				968	<p>std::deque is, in some senses, a generalized version of std::vector. Like
				969	std::vector, it provides constant time random access and other similar
				970	properties, but it also provides efficient access to the front of the list. It
				971	does not guarantee continuity of elements within memory.</p>
				972
				973	<p>In exchange for this extra flexibility, std::deque has significantly higher
				974	constant factor costs than std::vector. If possible, use std::vector or
				975	something cheaper.</p>
				976	</div>
				977
				978	<!-- _______________________________________________________________________ -->
				979	<div class="doc_subsubsection">
				980	<a name="dss_list"><list></a>
				981	</div>
				982
				983	<div class="doc_text">
				984	<p>std::list is an extremely inefficient class that is rarely useful.
				985	It performs a heap allocation for every element inserted into it, thus having an
				986	extremely high constant factor, particularly for small data types. std::list
				987	also only supports bidirectional iteration, not random access iteration.</p>
				988
				989	<p>In exchange for this high cost, std::list supports efficient access to both
				990	ends of the list (like std::deque, but unlike std::vector or SmallVector). In
				991	addition, the iterator invalidation characteristics of std::list are stronger
				992	than that of a vector class: inserting or removing an element into the list does
				993	not invalidate iterator or pointers to other elements in the list.</p>
				994	</div>
				995
				996	<!-- _______________________________________________________________________ -->
				997	<div class="doc_subsubsection">
Gabor Greif	bb17f65	2009-02-27 11:37:41 +0000	[diff] [blame]	998	<a name="dss_ilist">llvm/ADT/ilist.h</a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	999	</div>
				1000
				1001	<div class="doc_text">
				1002	<p><tt>ilist<T></tt> implements an 'intrusive' doubly-linked list. It is
				1003	intrusive, because it requires the element to store and provide access to the
				1004	prev/next pointers for the list.</p>
				1005
Gabor Greif	b6b21ec	2009-02-27 12:02:19 +0000	[diff] [blame]	1006	<p><tt>ilist</tt> has the same drawbacks as <tt>std::list</tt>, and additionally
				1007	requires an <tt>ilist_traits</tt> implementation for the element type, but it
				1008	provides some novel characteristics. In particular, it can efficiently store
				1009	polymorphic objects, the traits class is informed when an element is inserted or
Gabor Greif	e39f0e7	2009-03-12 09:47:03 +0000	[diff] [blame]	1010	removed from the list, and <tt>ilist</tt>s are guaranteed to support a
				1011	constant-time splice operation.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1012
Gabor Greif	e39f0e7	2009-03-12 09:47:03 +0000	[diff] [blame]	1013	<p>These properties are exactly what we want for things like
				1014	<tt>Instruction</tt>s and basic blocks, which is why these are implemented with
				1015	<tt>ilist</tt>s.</p>
Gabor Greif	bb17f65	2009-02-27 11:37:41 +0000	[diff] [blame]	1016
				1017	Related classes of interest are explained in the following subsections:
				1018	<ul>
Gabor Greif	a17abe1	2009-02-27 13:28:07 +0000	[diff] [blame]	1019	<li><a href="#dss_ilist_traits">ilist_traits</a></li>
Gabor Greif	b6b21ec	2009-02-27 12:02:19 +0000	[diff] [blame]	1020	<li><a href="#dss_iplist">iplist</a></li>
Gabor Greif	bb17f65	2009-02-27 11:37:41 +0000	[diff] [blame]	1021	<li><a href="#dss_ilist_node">llvm/ADT/ilist_node.h</a></li>
Gabor Greif	d970d69	2009-03-12 10:30:31 +0000	[diff] [blame]	1022	<li><a href="#dss_ilist_sentinel">Sentinels</a></li>
Gabor Greif	bb17f65	2009-02-27 11:37:41 +0000	[diff] [blame]	1023	</ul>
				1024	</div>
				1025
				1026	<!-- _______________________________________________________________________ -->
				1027	<div class="doc_subsubsection">
Gabor Greif	a17abe1	2009-02-27 13:28:07 +0000	[diff] [blame]	1028	<a name="dss_ilist_traits">ilist_traits</a>
				1029	</div>
				1030
				1031	<div class="doc_text">
				1032	<p><tt>ilist_traits<T></tt> is <tt>ilist<T></tt>'s customization
				1033	mechanism. <tt>iplist<T></tt> (and consequently <tt>ilist<T></tt>)
				1034	publicly derive from this traits class.</p>
				1035	</div>
				1036
				1037	<!-- _______________________________________________________________________ -->
				1038	<div class="doc_subsubsection">
Gabor Greif	b6b21ec	2009-02-27 12:02:19 +0000	[diff] [blame]	1039	<a name="dss_iplist">iplist</a>
				1040	</div>
				1041
				1042	<div class="doc_text">
				1043	<p><tt>iplist<T></tt> is <tt>ilist<T></tt>'s base and as such
Gabor Greif	e39f0e7	2009-03-12 09:47:03 +0000	[diff] [blame]	1044	supports a slightly narrower interface. Notably, inserters from
				1045	<tt>T&</tt> are absent.</p>
Gabor Greif	a17abe1	2009-02-27 13:28:07 +0000	[diff] [blame]	1046
				1047	<p><tt>ilist_traits<T></tt> is a public base of this class and can be
				1048	used for a wide variety of customizations.</p>
Gabor Greif	b6b21ec	2009-02-27 12:02:19 +0000	[diff] [blame]	1049	</div>
				1050
				1051	<!-- _______________________________________________________________________ -->
				1052	<div class="doc_subsubsection">
Gabor Greif	bb17f65	2009-02-27 11:37:41 +0000	[diff] [blame]	1053	<a name="dss_ilist_node">llvm/ADT/ilist_node.h</a>
				1054	</div>
				1055
				1056	<div class="doc_text">
				1057	<p><tt>ilist_node<T></tt> implements a the forward and backward links
				1058	that are expected by the <tt>ilist<T></tt> (and analogous containers)
				1059	in the default manner.</p>
				1060
				1061	<p><tt>ilist_node<T></tt>s are meant to be embedded in the node type
Gabor Greif	e39f0e7	2009-03-12 09:47:03 +0000	[diff] [blame]	1062	<tt>T</tt>, usually <tt>T</tt> publicly derives from
				1063	<tt>ilist_node<T></tt>.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1064	</div>
				1065
				1066	<!-- _______________________________________________________________________ -->
				1067	<div class="doc_subsubsection">
Gabor Greif	d970d69	2009-03-12 10:30:31 +0000	[diff] [blame]	1068	<a name="dss_ilist_sentinel">Sentinels</a>
				1069	</div>
				1070
				1071	<div class="doc_text">
				1072	<p><tt>ilist</tt>s have another speciality that must be considered. To be a good
				1073	citizen in the C++ ecosystem, it needs to support the standard container
				1074	operations, such as <tt>begin</tt> and <tt>end</tt> iterators, etc. Also, the
				1075	<tt>operator--</tt> must work correctly on the <tt>end</tt> iterator in the
				1076	case of non-empty <tt>ilist</tt>s.</p>
				1077
				1078	<p>The only sensible solution to this problem is to allocate a so-called
				1079	<i>sentinel</i> along with the intrusive list, which serves as the <tt>end</tt>
				1080	iterator, providing the back-link to the last element. However conforming to the
				1081	C++ convention it is illegal to <tt>operator++</tt> beyond the sentinel and it
				1082	also must not be dereferenced.</p>
				1083
				1084	<p>These constraints allow for some implementation freedom to the <tt>ilist</tt>
				1085	how to allocate and store the sentinel. The corresponding policy is dictated
				1086	by <tt>ilist_traits<T></tt>. By default a <tt>T</tt> gets heap-allocated
				1087	whenever the need for a sentinel arises.</p>
				1088
				1089	<p>While the default policy is sufficient in most cases, it may break down when
				1090	<tt>T</tt> does not provide a default constructor. Also, in the case of many
				1091	instances of <tt>ilist</tt>s, the memory overhead of the associated sentinels
				1092	is wasted. To alleviate the situation with numerous and voluminous
				1093	<tt>T</tt>-sentinels, sometimes a trick is employed, leading to <i>ghostly
				1094	sentinels</i>.</p>
				1095
				1096	<p>Ghostly sentinels are obtained by specially-crafted <tt>ilist_traits<T></tt>
				1097	which superpose the sentinel with the <tt>ilist</tt> instance in memory. Pointer
				1098	arithmetic is used to obtain the sentinel, which is relative to the
				1099	<tt>ilist</tt>'s <tt>this</tt> pointer. The <tt>ilist</tt> is augmented by an
				1100	extra pointer, which serves as the back-link of the sentinel. This is the only
				1101	field in the ghostly sentinel which can be legally accessed.</p>
				1102	</div>
				1103
				1104	<!-- _______________________________________________________________________ -->
				1105	<div class="doc_subsubsection">
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1106	<a name="dss_other">Other Sequential Container options</a>
				1107	</div>
				1108
				1109	<div class="doc_text">
				1110	<p>Other STL containers are available, such as std::string.</p>
				1111
				1112	<p>There are also various STL adapter classes such as std::queue,
				1113	std::priority_queue, std::stack, etc. These provide simplified access to an
				1114	underlying container but don't affect the cost of the container itself.</p>
				1115
				1116	</div>
				1117
				1118
				1119	<!-- ======================================================================= -->
				1120	<div class="doc_subsection">
				1121	<a name="ds_set">Set-Like Containers (std::set, SmallSet, SetVector, etc)</a>
				1122	</div>
				1123
				1124	<div class="doc_text">
				1125
				1126	<p>Set-like containers are useful when you need to canonicalize multiple values
				1127	into a single representation. There are several different choices for how to do
				1128	this, providing various trade-offs.</p>
				1129
				1130	</div>
				1131
				1132
				1133	<!-- _______________________________________________________________________ -->
				1134	<div class="doc_subsubsection">
				1135	<a name="dss_sortedvectorset">A sorted 'vector'</a>
				1136	</div>
				1137
				1138	<div class="doc_text">
				1139
				1140	<p>If you intend to insert a lot of elements, then do a lot of queries, a
				1141	great approach is to use a vector (or other sequential container) with
				1142	std::sort+std::unique to remove duplicates. This approach works really well if
				1143	your usage pattern has these two distinct phases (insert then query), and can be
				1144	coupled with a good choice of <a href="#ds_sequential">sequential container</a>.
				1145	</p>
				1146
				1147	<p>
				1148	This combination provides the several nice properties: the result data is
				1149	contiguous in memory (good for cache locality), has few allocations, is easy to
				1150	address (iterators in the final vector are just indices or pointers), and can be
				1151	efficiently queried with a standard binary or radix search.</p>
				1152
				1153	</div>
				1154
				1155	<!-- _______________________________________________________________________ -->
				1156	<div class="doc_subsubsection">
				1157	<a name="dss_smallset">"llvm/ADT/SmallSet.h"</a>
				1158	</div>
				1159
				1160	<div class="doc_text">
				1161
				1162	<p>If you have a set-like data structure that is usually small and whose elements
				1163	are reasonably small, a <tt>SmallSet<Type, N></tt> is a good choice. This set
				1164	has space for N elements in place (thus, if the set is dynamically smaller than
				1165	N, no malloc traffic is required) and accesses them with a simple linear search.
				1166	When the set grows beyond 'N' elements, it allocates a more expensive representation that
				1167	guarantees efficient access (for most types, it falls back to std::set, but for
				1168	pointers it uses something far better, <a
				1169	href="#dss_smallptrset">SmallPtrSet</a>).</p>
				1170
				1171	<p>The magic of this class is that it handles small sets extremely efficiently,
				1172	but gracefully handles extremely large sets without loss of efficiency. The
				1173	drawback is that the interface is quite small: it supports insertion, queries
				1174	and erasing, but does not support iteration.</p>
				1175
				1176	</div>
				1177
				1178	<!-- _______________________________________________________________________ -->
				1179	<div class="doc_subsubsection">
				1180	<a name="dss_smallptrset">"llvm/ADT/SmallPtrSet.h"</a>
				1181	</div>
				1182
				1183	<div class="doc_text">
				1184
				1185	<p>SmallPtrSet has all the advantages of SmallSet (and a SmallSet of pointers is
				1186	transparently implemented with a SmallPtrSet), but also supports iterators. If
				1187	more than 'N' insertions are performed, a single quadratically
				1188	probed hash table is allocated and grows as needed, providing extremely
				1189	efficient access (constant time insertion/deleting/queries with low constant
				1190	factors) and is very stingy with malloc traffic.</p>
				1191
				1192	<p>Note that, unlike std::set, the iterators of SmallPtrSet are invalidated
				1193	whenever an insertion occurs. Also, the values visited by the iterators are not
				1194	visited in sorted order.</p>
				1195
				1196	</div>
				1197
				1198	<!-- _______________________________________________________________________ -->
				1199	<div class="doc_subsubsection">
Chris Lattner	77ab46d	2007-09-30 00:58:59 +0000	[diff] [blame]	1200	<a name="dss_denseset">"llvm/ADT/DenseSet.h"</a>
				1201	</div>
				1202
				1203	<div class="doc_text">
				1204
				1205	<p>
				1206	DenseSet is a simple quadratically probed hash table. It excels at supporting
				1207	small values: it uses a single allocation to hold all of the pairs that
				1208	are currently inserted in the set. DenseSet is a great way to unique small
				1209	values that are not simple pointers (use <a
				1210	href="#dss_smallptrset">SmallPtrSet</a> for pointers). Note that DenseSet has
				1211	the same requirements for the value type that <a
				1212	href="#dss_densemap">DenseMap</a> has.
				1213	</p>
				1214
				1215	</div>
				1216
				1217	<!-- _______________________________________________________________________ -->
				1218	<div class="doc_subsubsection">
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1219	<a name="dss_FoldingSet">"llvm/ADT/FoldingSet.h"</a>
				1220	</div>
				1221
				1222	<div class="doc_text">
				1223
				1224	<p>
				1225	FoldingSet is an aggregate class that is really good at uniquing
				1226	expensive-to-create or polymorphic objects. It is a combination of a chained
				1227	hash table with intrusive links (uniqued objects are required to inherit from
				1228	FoldingSetNode) that uses <a href="#dss_smallvector">SmallVector</a> as part of
				1229	its ID process.</p>
				1230
				1231	<p>Consider a case where you want to implement a "getOrCreateFoo" method for
				1232	a complex object (for example, a node in the code generator). The client has a
				1233	description of what it wants to generate (it knows the opcode and all the
				1234	operands), but we don't want to 'new' a node, then try inserting it into a set
				1235	only to find out it already exists, at which point we would have to delete it
				1236	and return the node that already exists.
				1237	</p>
				1238
				1239	<p>To support this style of client, FoldingSet perform a query with a
				1240	FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
				1241	element that we want to query for. The query either returns the element
				1242	matching the ID or it returns an opaque ID that indicates where insertion should
				1243	take place. Construction of the ID usually does not require heap traffic.</p>
				1244
				1245	<p>Because FoldingSet uses intrusive links, it can support polymorphic objects
				1246	in the set (for example, you can have SDNode instances mixed with LoadSDNodes).
				1247	Because the elements are individually allocated, pointers to the elements are
				1248	stable: inserting or removing elements does not invalidate any pointers to other
				1249	elements.
				1250	</p>
				1251
				1252	</div>
				1253
				1254	<!-- _______________________________________________________________________ -->
				1255	<div class="doc_subsubsection">
				1256	<a name="dss_set"><set></a>
				1257	</div>
				1258
				1259	<div class="doc_text">
				1260
				1261	<p><tt>std::set</tt> is a reasonable all-around set class, which is decent at
				1262	many things but great at nothing. std::set allocates memory for each element
				1263	inserted (thus it is very malloc intensive) and typically stores three pointers
				1264	per element in the set (thus adding a large amount of per-element space
				1265	overhead). It offers guaranteed log(n) performance, which is not particularly
				1266	fast from a complexity standpoint (particularly if the elements of the set are
				1267	expensive to compare, like strings), and has extremely high constant factors for
				1268	lookup, insertion and removal.</p>
				1269
				1270	<p>The advantages of std::set are that its iterators are stable (deleting or
				1271	inserting an element from the set does not affect iterators or pointers to other
				1272	elements) and that iteration over the set is guaranteed to be in sorted order.
				1273	If the elements in the set are large, then the relative overhead of the pointers
				1274	and malloc traffic is not a big deal, but if the elements of the set are small,
				1275	std::set is almost never a good choice.</p>
				1276
				1277	</div>
				1278
				1279	<!-- _______________________________________________________________________ -->
				1280	<div class="doc_subsubsection">
				1281	<a name="dss_setvector">"llvm/ADT/SetVector.h"</a>
				1282	</div>
				1283
				1284	<div class="doc_text">
				1285	<p>LLVM's SetVector<Type> is an adapter class that combines your choice of
				1286	a set-like container along with a <a href="#ds_sequential">Sequential
				1287	Container</a>. The important property
				1288	that this provides is efficient insertion with uniquing (duplicate elements are
				1289	ignored) with iteration support. It implements this by inserting elements into
				1290	both a set-like container and the sequential container, using the set-like
				1291	container for uniquing and the sequential container for iteration.
				1292	</p>
				1293
				1294	<p>The difference between SetVector and other sets is that the order of
				1295	iteration is guaranteed to match the order of insertion into the SetVector.
				1296	This property is really important for things like sets of pointers. Because
				1297	pointer values are non-deterministic (e.g. vary across runs of the program on
				1298	different machines), iterating over the pointers in the set will
				1299	not be in a well-defined order.</p>
				1300
				1301	<p>
				1302	The drawback of SetVector is that it requires twice as much space as a normal
				1303	set and has the sum of constant factors from the set-like container and the
				1304	sequential container that it uses. Use it only if you need to iterate over
				1305	the elements in a deterministic order. SetVector is also expensive to delete
				1306	elements out of (linear time), unless you use it's "pop_back" method, which is
				1307	faster.
				1308	</p>
				1309
				1310	<p>SetVector is an adapter class that defaults to using std::vector and std::set
				1311	for the underlying containers, so it is quite expensive. However,
				1312	<tt>"llvm/ADT/SetVector.h"</tt> also provides a SmallSetVector class, which
				1313	defaults to using a SmallVector and SmallSet of a specified size. If you use
				1314	this, and if your sets are dynamically smaller than N, you will save a lot of
				1315	heap traffic.</p>
				1316
				1317	</div>
				1318
				1319	<!-- _______________________________________________________________________ -->
				1320	<div class="doc_subsubsection">
				1321	<a name="dss_uniquevector">"llvm/ADT/UniqueVector.h"</a>
				1322	</div>
				1323
				1324	<div class="doc_text">
				1325
				1326	<p>
				1327	UniqueVector is similar to <a href="#dss_setvector">SetVector</a>, but it
				1328	retains a unique ID for each element inserted into the set. It internally
				1329	contains a map and a vector, and it assigns a unique ID for each value inserted
				1330	into the set.</p>
				1331
				1332	<p>UniqueVector is very expensive: its cost is the sum of the cost of
				1333	maintaining both the map and vector, it has high complexity, high constant
				1334	factors, and produces a lot of malloc traffic. It should be avoided.</p>
				1335
				1336	</div>
				1337
				1338
				1339	<!-- _______________________________________________________________________ -->
				1340	<div class="doc_subsubsection">
				1341	<a name="dss_otherset">Other Set-Like Container Options</a>
				1342	</div>
				1343
				1344	<div class="doc_text">
				1345
				1346	<p>
				1347	The STL provides several other options, such as std::multiset and the various
Chris Lattner	86a63d0	2009-03-09 05:20:45 +0000	[diff] [blame]	1348	"hash_set" like containers (whether from C++ TR1 or from the SGI library). We
				1349	never use hash_set and unordered_set because they are generally very expensive
				1350	(each insertion requires a malloc) and very non-portable.
				1351	</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1352
				1353	<p>std::multiset is useful if you're not interested in elimination of
				1354	duplicates, but has all the drawbacks of std::set. A sorted vector (where you
				1355	don't delete duplicate entries) or some other approach is almost always
				1356	better.</p>
				1357
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1358	</div>
				1359
				1360	<!-- ======================================================================= -->
				1361	<div class="doc_subsection">
				1362	<a name="ds_map">Map-Like Containers (std::map, DenseMap, etc)</a>
				1363	</div>
				1364
				1365	<div class="doc_text">
				1366	Map-like containers are useful when you want to associate data to a key. As
				1367	usual, there are a lot of different ways to do this. :)
				1368	</div>
				1369
				1370	<!-- _______________________________________________________________________ -->
				1371	<div class="doc_subsubsection">
				1372	<a name="dss_sortedvectormap">A sorted 'vector'</a>
				1373	</div>
				1374
				1375	<div class="doc_text">
				1376
				1377	<p>
				1378	If your usage pattern follows a strict insert-then-query approach, you can
				1379	trivially use the same approach as <a href="#dss_sortedvectorset">sorted vectors
				1380	for set-like containers</a>. The only difference is that your query function
				1381	(which uses std::lower_bound to get efficient log(n) lookup) should only compare
				1382	the key, not both the key and value. This yields the same advantages as sorted
				1383	vectors for sets.
				1384	</p>
				1385	</div>
				1386
				1387	<!-- _______________________________________________________________________ -->
				1388	<div class="doc_subsubsection">
				1389	<a name="dss_stringmap">"llvm/ADT/StringMap.h"</a>
				1390	</div>
				1391
				1392	<div class="doc_text">
				1393
				1394	<p>
				1395	Strings are commonly used as keys in maps, and they are difficult to support
				1396	efficiently: they are variable length, inefficient to hash and compare when
				1397	long, expensive to copy, etc. StringMap is a specialized container designed to
				1398	cope with these issues. It supports mapping an arbitrary range of bytes to an
				1399	arbitrary other object.</p>
				1400
				1401	<p>The StringMap implementation uses a quadratically-probed hash table, where
				1402	the buckets store a pointer to the heap allocated entries (and some other
				1403	stuff). The entries in the map must be heap allocated because the strings are
				1404	variable length. The string data (key) and the element object (value) are
				1405	stored in the same allocation with the string data immediately after the element
				1406	object. This container guarantees the "<tt>(char*)(&Value+1)</tt>" points
				1407	to the key string for a value.</p>
				1408
				1409	<p>The StringMap is very fast for several reasons: quadratic probing is very
				1410	cache efficient for lookups, the hash value of strings in buckets is not
				1411	recomputed when lookup up an element, StringMap rarely has to touch the
				1412	memory for unrelated objects when looking up a value (even when hash collisions
				1413	happen), hash table growth does not recompute the hash values for strings
				1414	already in the table, and each pair in the map is store in a single allocation
				1415	(the string data is stored in the same allocation as the Value of a pair).</p>
				1416
				1417	<p>StringMap also provides query methods that take byte ranges, so it only ever
				1418	copies a string if a value is inserted into the table.</p>
				1419	</div>
				1420
				1421	<!-- _______________________________________________________________________ -->
				1422	<div class="doc_subsubsection">
				1423	<a name="dss_indexedmap">"llvm/ADT/IndexedMap.h"</a>
				1424	</div>
				1425
				1426	<div class="doc_text">
				1427	<p>
				1428	IndexedMap is a specialized container for mapping small dense integers (or
				1429	values that can be mapped to small dense integers) to some other type. It is
				1430	internally implemented as a vector with a mapping function that maps the keys to
				1431	the dense integer range.
				1432	</p>
				1433
				1434	<p>
				1435	This is useful for cases like virtual registers in the LLVM code generator: they
				1436	have a dense mapping that is offset by a compile-time constant (the first
				1437	virtual register ID).</p>
				1438
				1439	</div>
				1440
				1441	<!-- _______________________________________________________________________ -->
				1442	<div class="doc_subsubsection">
				1443	<a name="dss_densemap">"llvm/ADT/DenseMap.h"</a>
				1444	</div>
				1445
				1446	<div class="doc_text">
				1447
				1448	<p>
				1449	DenseMap is a simple quadratically probed hash table. It excels at supporting
				1450	small keys and values: it uses a single allocation to hold all of the pairs that
				1451	are currently inserted in the map. DenseMap is a great way to map pointers to
				1452	pointers, or map other small types to each other.
				1453	</p>
				1454
				1455	<p>
				1456	There are several aspects of DenseMap that you should be aware of, however. The
				1457	iterators in a densemap are invalidated whenever an insertion occurs, unlike
				1458	map. Also, because DenseMap allocates space for a large number of key/value
				1459	pairs (it starts with 64 by default), it will waste a lot of space if your keys
				1460	or values are large. Finally, you must implement a partial specialization of
Chris Lattner	92eea07	2007-09-17 18:34:04 +0000	[diff] [blame]	1461	DenseMapInfo for the key that you want, if it isn't already supported. This
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1462	is required to tell DenseMap about two special marker values (which can never be
				1463	inserted into the map) that it needs internally.</p>
				1464
				1465	</div>
				1466
				1467	<!-- _______________________________________________________________________ -->
				1468	<div class="doc_subsubsection">
				1469	<a name="dss_map"><map></a>
				1470	</div>
				1471
				1472	<div class="doc_text">
				1473
				1474	<p>
				1475	std::map has similar characteristics to <a href="#dss_set">std::set</a>: it uses
				1476	a single allocation per pair inserted into the map, it offers log(n) lookup with
				1477	an extremely large constant factor, imposes a space penalty of 3 pointers per
				1478	pair in the map, etc.</p>
				1479
				1480	<p>std::map is most useful when your keys or values are very large, if you need
				1481	to iterate over the collection in sorted order, or if you need stable iterators
				1482	into the map (i.e. they don't get invalidated if an insertion or deletion of
				1483	another element takes place).</p>
				1484
				1485	</div>
				1486
				1487	<!-- _______________________________________________________________________ -->
				1488	<div class="doc_subsubsection">
				1489	<a name="dss_othermap">Other Map-Like Container Options</a>
				1490	</div>
				1491
				1492	<div class="doc_text">
				1493
				1494	<p>
				1495	The STL provides several other options, such as std::multimap and the various
Chris Lattner	86a63d0	2009-03-09 05:20:45 +0000	[diff] [blame]	1496	"hash_map" like containers (whether from C++ TR1 or from the SGI library). We
				1497	never use hash_set and unordered_set because they are generally very expensive
				1498	(each insertion requires a malloc) and very non-portable.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1499
				1500	<p>std::multimap is useful if you want to map a key to multiple values, but has
				1501	all the drawbacks of std::map. A sorted vector or some other approach is almost
				1502	always better.</p>
				1503
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1504	</div>
				1505
Daniel Berlin	7ea44dc	2007-09-24 17:52:25 +0000	[diff] [blame]	1506	<!-- ======================================================================= -->
				1507	<div class="doc_subsection">
				1508	<a name="ds_bit">Bit storage containers (BitVector, SparseBitVector)</a>
				1509	</div>
				1510
				1511	<div class="doc_text">
Chris Lattner	62a4eae	2007-09-25 22:37:50 +0000	[diff] [blame]	1512	<p>Unlike the other containers, there are only two bit storage containers, and
				1513	choosing when to use each is relatively straightforward.</p>
				1514
				1515	<p>One additional option is
				1516	<tt>std::vector<bool></tt>: we discourage its use for two reasons 1) the
				1517	implementation in many common compilers (e.g. commonly available versions of
				1518	GCC) is extremely inefficient and 2) the C++ standards committee is likely to
				1519	deprecate this container and/or change it significantly somehow. In any case,
				1520	please don't use it.</p>
Daniel Berlin	7ea44dc	2007-09-24 17:52:25 +0000	[diff] [blame]	1521	</div>
				1522
				1523	<!-- _______________________________________________________________________ -->
				1524	<div class="doc_subsubsection">
				1525	<a name="dss_bitvector">BitVector</a>
				1526	</div>
				1527
				1528	<div class="doc_text">
				1529	<p> The BitVector container provides a fixed size set of bits for manipulation.
				1530	It supports individual bit setting/testing, as well as set operations. The set
				1531	operations take time O(size of bitvector), but operations are performed one word
				1532	at a time, instead of one bit at a time. This makes the BitVector very fast for
				1533	set operations compared to other containers. Use the BitVector when you expect
				1534	the number of set bits to be high (IE a dense set).
				1535	</p>
				1536	</div>
				1537
				1538	<!-- _______________________________________________________________________ -->
				1539	<div class="doc_subsubsection">
				1540	<a name="dss_sparsebitvector">SparseBitVector</a>
				1541	</div>
				1542
				1543	<div class="doc_text">
				1544	<p> The SparseBitVector container is much like BitVector, with one major
				1545	difference: Only the bits that are set, are stored. This makes the
				1546	SparseBitVector much more space efficient than BitVector when the set is sparse,
				1547	as well as making set operations O(number of set bits) instead of O(size of
				1548	universe). 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
				1549	(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).
				1550	</p>
				1551	</div>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1552
				1553	<!-- *********************************************************************** -->
				1554	<div class="doc_section">
				1555	<a name="common">Helpful Hints for Common Operations</a>
				1556	</div>
				1557	<!-- *********************************************************************** -->
				1558
				1559	<div class="doc_text">
				1560
				1561	<p>This section describes how to perform some very simple transformations of
				1562	LLVM code. This is meant to give examples of common idioms used, showing the
				1563	practical side of LLVM transformations. <p> Because this is a "how-to" section,
				1564	you should also read about the main classes that you will be working with. The
				1565	<a href="#coreclasses">Core LLVM Class Hierarchy Reference</a> contains details
				1566	and descriptions of the main classes that you should know about.</p>
				1567
				1568	</div>
				1569
				1570	<!-- NOTE: this section should be heavy on example code -->
				1571	<!-- ======================================================================= -->
				1572	<div class="doc_subsection">
				1573	<a name="inspection">Basic Inspection and Traversal Routines</a>
				1574	</div>
				1575
				1576	<div class="doc_text">
				1577
				1578	<p>The LLVM compiler infrastructure have many different data structures that may
				1579	be traversed. Following the example of the C++ standard template library, the
				1580	techniques used to traverse these various data structures are all basically the
				1581	same. For a enumerable sequence of values, the <tt>XXXbegin()</tt> function (or
				1582	method) returns an iterator to the start of the sequence, the <tt>XXXend()</tt>
				1583	function returns an iterator pointing to one past the last valid element of the
				1584	sequence, and there is some <tt>XXXiterator</tt> data type that is common
				1585	between the two operations.</p>
				1586
				1587	<p>Because the pattern for iteration is common across many different aspects of
				1588	the program representation, the standard template library algorithms may be used
				1589	on them, and it is easier to remember how to iterate. First we show a few common
				1590	examples of the data structures that need to be traversed. Other data
				1591	structures are traversed in very similar ways.</p>
				1592
				1593	</div>
				1594
				1595	<!-- _______________________________________________________________________ -->
				1596	<div class="doc_subsubsection">
				1597	<a name="iterate_function">Iterating over the </a><a
				1598	href="#BasicBlock"><tt>BasicBlock</tt></a>s in a <a
				1599	href="#Function"><tt>Function</tt></a>
				1600	</div>
				1601
				1602	<div class="doc_text">
				1603
				1604	<p>It's quite common to have a <tt>Function</tt> instance that you'd like to
				1605	transform in some way; in particular, you'd like to manipulate its
				1606	<tt>BasicBlock</tt>s. To facilitate this, you'll need to iterate over all of
				1607	the <tt>BasicBlock</tt>s that constitute the <tt>Function</tt>. The following is
				1608	an example that prints the name of a <tt>BasicBlock</tt> and the number of
				1609	<tt>Instruction</tt>s it contains:</p>
				1610
				1611	<div class="doc_code">
				1612	<pre>
				1613	// <i>func is a pointer to a Function instance</i>
				1614	for (Function::iterator i = func->begin(), e = func->end(); i != e; ++i)
				1615	// <i>Print out the name of the basic block if it has one, and then the</i>
				1616	// <i>number of instructions that it contains</i>
				1617	llvm::cerr << "Basic block (name=" << i->getName() << ") has "
				1618	<< i->size() << " instructions.\n";
				1619	</pre>
				1620	</div>
				1621
				1622	<p>Note that i can be used as if it were a pointer for the purposes of
				1623	invoking member functions of the <tt>Instruction</tt> class. This is
				1624	because the indirection operator is overloaded for the iterator
				1625	classes. In the above code, the expression <tt>i->size()</tt> is
				1626	exactly equivalent to <tt>(*i).size()</tt> just like you'd expect.</p>
				1627
				1628	</div>
				1629
				1630	<!-- _______________________________________________________________________ -->
				1631	<div class="doc_subsubsection">
				1632	<a name="iterate_basicblock">Iterating over the </a><a
				1633	href="#Instruction"><tt>Instruction</tt></a>s in a <a
				1634	href="#BasicBlock"><tt>BasicBlock</tt></a>
				1635	</div>
				1636
				1637	<div class="doc_text">
				1638
				1639	<p>Just like when dealing with <tt>BasicBlock</tt>s in <tt>Function</tt>s, it's
				1640	easy to iterate over the individual instructions that make up
				1641	<tt>BasicBlock</tt>s. Here's a code snippet that prints out each instruction in
				1642	a <tt>BasicBlock</tt>:</p>
				1643
				1644	<div class="doc_code">
				1645	<pre>
				1646	// <i>blk is a pointer to a BasicBlock instance</i>
				1647	for (BasicBlock::iterator i = blk->begin(), e = blk->end(); i != e; ++i)
				1648	// <i>The next statement works since operator<<(ostream&,...)</i>
				1649	// <i>is overloaded for Instruction&</i>
				1650	llvm::cerr << *i << "\n";
				1651	</pre>
				1652	</div>
				1653
				1654	<p>However, this isn't really the best way to print out the contents of a
				1655	<tt>BasicBlock</tt>! Since the ostream operators are overloaded for virtually
				1656	anything you'll care about, you could have just invoked the print routine on the
				1657	basic block itself: <tt>llvm::cerr << *blk << "\n";</tt>.</p>
				1658
				1659	</div>
				1660
				1661	<!-- _______________________________________________________________________ -->
				1662	<div class="doc_subsubsection">
				1663	<a name="iterate_institer">Iterating over the </a><a
				1664	href="#Instruction"><tt>Instruction</tt></a>s in a <a
				1665	href="#Function"><tt>Function</tt></a>
				1666	</div>
				1667
				1668	<div class="doc_text">
				1669
				1670	<p>If you're finding that you commonly iterate over a <tt>Function</tt>'s
				1671	<tt>BasicBlock</tt>s and then that <tt>BasicBlock</tt>'s <tt>Instruction</tt>s,
				1672	<tt>InstIterator</tt> should be used instead. You'll need to include <a
				1673	href="/doxygen/InstIterator_8h-source.html"><tt>llvm/Support/InstIterator.h</tt></a>,
				1674	and then instantiate <tt>InstIterator</tt>s explicitly in your code. Here's a
				1675	small example that shows how to dump all instructions in a function to the standard error stream:<p>
				1676
				1677	<div class="doc_code">
				1678	<pre>
				1679	#include "<a href="/doxygen/InstIterator_8h-source.html">llvm/Support/InstIterator.h</a>"
				1680
				1681	// <i>F is a pointer to a Function instance</i>
Chris Lattner	fd62ee7	2008-06-04 18:20:42 +0000	[diff] [blame]	1682	for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
				1683	llvm::cerr << *I << "\n";
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1684	</pre>
				1685	</div>
				1686
				1687	<p>Easy, isn't it? You can also use <tt>InstIterator</tt>s to fill a
				1688	work list with its initial contents. For example, if you wanted to
				1689	initialize a work list to contain all instructions in a <tt>Function</tt>
				1690	F, all you would need to do is something like:</p>
				1691
				1692	<div class="doc_code">
				1693	<pre>
				1694	std::set<Instruction*> worklist;
Chris Lattner	fd62ee7	2008-06-04 18:20:42 +0000	[diff] [blame]	1695	// or better yet, SmallPtrSet<Instruction*, 64> worklist;
				1696
				1697	for (inst_iterator I = inst_begin(F), E = inst_end(F); I != E; ++I)
				1698	worklist.insert(&*I);
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1699	</pre>
				1700	</div>
				1701
				1702	<p>The STL set <tt>worklist</tt> would now contain all instructions in the
				1703	<tt>Function</tt> pointed to by F.</p>
				1704
				1705	</div>
				1706
				1707	<!-- _______________________________________________________________________ -->
				1708	<div class="doc_subsubsection">
				1709	<a name="iterate_convert">Turning an iterator into a class pointer (and
				1710	vice-versa)</a>
				1711	</div>
				1712
				1713	<div class="doc_text">
				1714
				1715	<p>Sometimes, it'll be useful to grab a reference (or pointer) to a class
				1716	instance when all you've got at hand is an iterator. Well, extracting
				1717	a reference or a pointer from an iterator is very straight-forward.
				1718	Assuming that <tt>i</tt> is a <tt>BasicBlock::iterator</tt> and <tt>j</tt>
				1719	is a <tt>BasicBlock::const_iterator</tt>:</p>
				1720
				1721	<div class="doc_code">
				1722	<pre>
				1723	Instruction& inst = *i; // <i>Grab reference to instruction reference</i>
				1724	Instruction* pinst = &*i; // <i>Grab pointer to instruction reference</i>
				1725	const Instruction& inst = *j;
				1726	</pre>
				1727	</div>
				1728
				1729	<p>However, the iterators you'll be working with in the LLVM framework are
				1730	special: they will automatically convert to a ptr-to-instance type whenever they
				1731	need to. Instead of dereferencing the iterator and then taking the address of
				1732	the result, you can simply assign the iterator to the proper pointer type and
				1733	you get the dereference and address-of operation as a result of the assignment
				1734	(behind the scenes, this is a result of overloading casting mechanisms). Thus
				1735	the last line of the last example,</p>
				1736
				1737	<div class="doc_code">
				1738	<pre>
Chris Lattner	0665e1f	2008-01-03 16:56:04 +0000	[diff] [blame]	1739	Instruction pinst = &i;
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1740	</pre>
				1741	</div>
				1742
				1743	<p>is semantically equivalent to</p>
				1744
				1745	<div class="doc_code">
				1746	<pre>
Chris Lattner	0665e1f	2008-01-03 16:56:04 +0000	[diff] [blame]	1747	Instruction *pinst = i;
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1748	</pre>
				1749	</div>
				1750
				1751	<p>It's also possible to turn a class pointer into the corresponding iterator,
				1752	and this is a constant time operation (very efficient). The following code
				1753	snippet illustrates use of the conversion constructors provided by LLVM
				1754	iterators. By using these, you can explicitly grab the iterator of something
				1755	without actually obtaining it via iteration over some structure:</p>
				1756
				1757	<div class="doc_code">
				1758	<pre>
				1759	void printNextInstruction(Instruction* inst) {
				1760	BasicBlock::iterator it(inst);
				1761	++it; // <i>After this line, it refers to the instruction after *inst</i>
				1762	if (it != inst->getParent()->end()) llvm::cerr << *it << "\n";
				1763	}
				1764	</pre>
				1765	</div>
				1766
				1767	</div>
				1768
				1769	<!--_______________________________________________________________________-->
				1770	<div class="doc_subsubsection">
				1771	<a name="iterate_complex">Finding call sites: a slightly more complex
				1772	example</a>
				1773	</div>
				1774
				1775	<div class="doc_text">
				1776
				1777	<p>Say that you're writing a FunctionPass and would like to count all the
				1778	locations in the entire module (that is, across every <tt>Function</tt>) where a
				1779	certain function (i.e., some <tt>Function</tt>*) is already in scope. As you'll
				1780	learn later, you may want to use an <tt>InstVisitor</tt> to accomplish this in a
				1781	much more straight-forward manner, but this example will allow us to explore how
				1782	you'd do it if you didn't have <tt>InstVisitor</tt> around. In pseudo-code, this
				1783	is what we want to do:</p>
				1784
				1785	<div class="doc_code">
				1786	<pre>
				1787	initialize callCounter to zero
				1788	for each Function f in the Module
				1789	for each BasicBlock b in f
				1790	for each Instruction i in b
				1791	if (i is a CallInst and calls the given function)
				1792	increment callCounter
				1793	</pre>
				1794	</div>
				1795
				1796	<p>And the actual code is (remember, because we're writing a
				1797	<tt>FunctionPass</tt>, our <tt>FunctionPass</tt>-derived class simply has to
				1798	override the <tt>runOnFunction</tt> method):</p>
				1799
				1800	<div class="doc_code">
				1801	<pre>
				1802	Function* targetFunc = ...;
				1803
				1804	class OurFunctionPass : public FunctionPass {
				1805	public:
				1806	OurFunctionPass(): callCounter(0) { }
				1807
				1808	virtual runOnFunction(Function& F) {
				1809	for (Function::iterator b = F.begin(), be = F.end(); b != be; ++b) {
Eric Christopher	5fbecf7	2008-11-08 08:20:49 +0000	[diff] [blame]	1810	for (BasicBlock::iterator i = b->begin(), ie = b->end(); i != ie; ++i) {
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1811	if (<a href="#CallInst">CallInst</a>* callInst = <a href="#isa">dyn_cast</a><<a
				1812	href="#CallInst">CallInst</a>>(&*i)) {
				1813	// <i>We know we've encountered a call instruction, so we</i>
				1814	// <i>need to determine if it's a call to the</i>
Chris Lattner	0665e1f	2008-01-03 16:56:04 +0000	[diff] [blame]	1815	// <i>function pointed to by m_func or not.</i>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1816	if (callInst->getCalledFunction() == targetFunc)
				1817	++callCounter;
				1818	}
				1819	}
				1820	}
				1821	}
				1822
				1823	private:
Chris Lattner	0665e1f	2008-01-03 16:56:04 +0000	[diff] [blame]	1824	unsigned callCounter;
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1825	};
				1826	</pre>
				1827	</div>
				1828
				1829	</div>
				1830
				1831	<!--_______________________________________________________________________-->
				1832	<div class="doc_subsubsection">
				1833	<a name="calls_and_invokes">Treating calls and invokes the same way</a>
				1834	</div>
				1835
				1836	<div class="doc_text">
				1837
				1838	<p>You may have noticed that the previous example was a bit oversimplified in
				1839	that it did not deal with call sites generated by 'invoke' instructions. In
				1840	this, and in other situations, you may find that you want to treat
				1841	<tt>CallInst</tt>s and <tt>InvokeInst</tt>s the same way, even though their
				1842	most-specific common base class is <tt>Instruction</tt>, which includes lots of
				1843	less closely-related things. For these cases, LLVM provides a handy wrapper
				1844	class called <a
				1845	href="http://llvm.org/doxygen/classllvm_1_1CallSite.html"><tt>CallSite</tt></a>.
				1846	It is essentially a wrapper around an <tt>Instruction</tt> pointer, with some
				1847	methods that provide functionality common to <tt>CallInst</tt>s and
				1848	<tt>InvokeInst</tt>s.</p>
				1849
				1850	<p>This class has "value semantics": it should be passed by value, not by
				1851	reference and it should not be dynamically allocated or deallocated using
				1852	<tt>operator new</tt> or <tt>operator delete</tt>. It is efficiently copyable,
				1853	assignable and constructable, with costs equivalents to that of a bare pointer.
				1854	If you look at its definition, it has only a single pointer member.</p>
				1855
				1856	</div>
				1857
				1858	<!--_______________________________________________________________________-->
				1859	<div class="doc_subsubsection">
				1860	<a name="iterate_chains">Iterating over def-use & use-def chains</a>
				1861	</div>
				1862
				1863	<div class="doc_text">
				1864
				1865	<p>Frequently, we might have an instance of the <a
				1866	href="/doxygen/classllvm_1_1Value.html">Value Class</a> and we want to
				1867	determine which <tt>User</tt>s use the <tt>Value</tt>. The list of all
				1868	<tt>User</tt>s of a particular <tt>Value</tt> is called a <i>def-use</i> chain.
				1869	For example, let's say we have a <tt>Function*</tt> named <tt>F</tt> to a
				1870	particular function <tt>foo</tt>. Finding all of the instructions that
				1871	<i>use</i> <tt>foo</tt> is as simple as iterating over the <i>def-use</i> chain
				1872	of <tt>F</tt>:</p>
				1873
				1874	<div class="doc_code">
				1875	<pre>
Chris Lattner	0665e1f	2008-01-03 16:56:04 +0000	[diff] [blame]	1876	Function *F = ...;
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1877
				1878	for (Value::use_iterator i = F->use_begin(), e = F->use_end(); i != e; ++i)
				1879	if (Instruction Inst = dyn_cast<Instruction>(i)) {
				1880	llvm::cerr << "F is used in instruction:\n";
				1881	llvm::cerr << *Inst << "\n";
				1882	}
				1883	</pre>
				1884	</div>
				1885
				1886	<p>Alternately, it's common to have an instance of the <a
				1887	href="/doxygen/classllvm_1_1User.html">User Class</a> and need to know what
				1888	<tt>Value</tt>s are used by it. The list of all <tt>Value</tt>s used by a
				1889	<tt>User</tt> is known as a <i>use-def</i> chain. Instances of class
				1890	<tt>Instruction</tt> are common <tt>User</tt>s, so we might want to iterate over
				1891	all of the values that a particular instruction uses (that is, the operands of
				1892	the particular <tt>Instruction</tt>):</p>
				1893
				1894	<div class="doc_code">
				1895	<pre>
Chris Lattner	0665e1f	2008-01-03 16:56:04 +0000	[diff] [blame]	1896	Instruction *pi = ...;
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1897
				1898	for (User::op_iterator i = pi->op_begin(), e = pi->op_end(); i != e; ++i) {
Chris Lattner	0665e1f	2008-01-03 16:56:04 +0000	[diff] [blame]	1899	Value v = i;
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1900	// <i>...</i>
				1901	}
				1902	</pre>
				1903	</div>
				1904
				1905	<!--
				1906	def-use chains ("finding all users of"): Value::use_begin/use_end
				1907	use-def chains ("finding all values used"): User::op_begin/op_end [op=operand]
				1908	-->
				1909
				1910	</div>
				1911
Chris Lattner	0665e1f	2008-01-03 16:56:04 +0000	[diff] [blame]	1912	<!--_______________________________________________________________________-->
				1913	<div class="doc_subsubsection">
				1914	<a name="iterate_preds">Iterating over predecessors &
				1915	successors of blocks</a>
				1916	</div>
				1917
				1918	<div class="doc_text">
				1919
				1920	<p>Iterating over the predecessors and successors of a block is quite easy
				1921	with the routines defined in <tt>"llvm/Support/CFG.h"</tt>. Just use code like
				1922	this to iterate over all predecessors of BB:</p>
				1923
				1924	<div class="doc_code">
				1925	<pre>
				1926	#include "llvm/Support/CFG.h"
				1927	BasicBlock *BB = ...;
				1928
				1929	for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
				1930	BasicBlock Pred = PI;
				1931	// <i>...</i>
				1932	}
				1933	</pre>
				1934	</div>
				1935
				1936	<p>Similarly, to iterate over successors use
				1937	succ_iterator/succ_begin/succ_end.</p>
				1938
				1939	</div>
				1940
				1941
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1942	<!-- ======================================================================= -->
				1943	<div class="doc_subsection">
				1944	<a name="simplechanges">Making simple changes</a>
				1945	</div>
				1946
				1947	<div class="doc_text">
				1948
				1949	<p>There are some primitive transformation operations present in the LLVM
				1950	infrastructure that are worth knowing about. When performing
				1951	transformations, it's fairly common to manipulate the contents of basic
				1952	blocks. This section describes some of the common methods for doing so
				1953	and gives example code.</p>
				1954
				1955	</div>
				1956
				1957	<!--_______________________________________________________________________-->
				1958	<div class="doc_subsubsection">
				1959	<a name="schanges_creating">Creating and inserting new
				1960	<tt>Instruction</tt>s</a>
				1961	</div>
				1962
				1963	<div class="doc_text">
				1964
				1965	<p><i>Instantiating Instructions</i></p>
				1966
				1967	<p>Creation of <tt>Instruction</tt>s is straight-forward: simply call the
				1968	constructor for the kind of instruction to instantiate and provide the necessary
				1969	parameters. For example, an <tt>AllocaInst</tt> only <i>requires</i> a
				1970	(const-ptr-to) <tt>Type</tt>. Thus:</p>
				1971
				1972	<div class="doc_code">
				1973	<pre>
Nick Lewycky	c004e5b	2007-12-03 01:52:52 +0000	[diff] [blame]	1974	AllocaInst* ai = new AllocaInst(Type::Int32Ty);
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	1975	</pre>
				1976	</div>
				1977
				1978	<p>will create an <tt>AllocaInst</tt> instance that represents the allocation of
				1979	one integer in the current stack frame, at run time. Each <tt>Instruction</tt>
				1980	subclass is likely to have varying default parameters which change the semantics
				1981	of the instruction, so refer to the <a
				1982	href="/doxygen/classllvm_1_1Instruction.html">doxygen documentation for the subclass of
				1983	Instruction</a> that you're interested in instantiating.</p>
				1984
				1985	<p><i>Naming values</i></p>
				1986
				1987	<p>It is very useful to name the values of instructions when you're able to, as
				1988	this facilitates the debugging of your transformations. If you end up looking
				1989	at generated LLVM machine code, you definitely want to have logical names
				1990	associated with the results of instructions! By supplying a value for the
				1991	<tt>Name</tt> (default) parameter of the <tt>Instruction</tt> constructor, you
				1992	associate a logical name with the result of the instruction's execution at
				1993	run time. For example, say that I'm writing a transformation that dynamically
				1994	allocates space for an integer on the stack, and that integer is going to be
				1995	used as some kind of index by some other code. To accomplish this, I place an
				1996	<tt>AllocaInst</tt> at the first point in the first <tt>BasicBlock</tt> of some
				1997	<tt>Function</tt>, and I'm intending to use it within the same
				1998	<tt>Function</tt>. I might do:</p>
				1999
				2000	<div class="doc_code">
				2001	<pre>
Nick Lewycky	c004e5b	2007-12-03 01:52:52 +0000	[diff] [blame]	2002	AllocaInst* pa = new AllocaInst(Type::Int32Ty, 0, "indexLoc");
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2003	</pre>
				2004	</div>
				2005
				2006	<p>where <tt>indexLoc</tt> is now the logical name of the instruction's
				2007	execution value, which is a pointer to an integer on the run time stack.</p>
				2008
				2009	<p><i>Inserting instructions</i></p>
				2010
				2011	<p>There are essentially two ways to insert an <tt>Instruction</tt>
				2012	into an existing sequence of instructions that form a <tt>BasicBlock</tt>:</p>
				2013
				2014	<ul>
				2015	<li>Insertion into an explicit instruction list
				2016
				2017	<p>Given a <tt>BasicBlock* pb</tt>, an <tt>Instruction* pi</tt> within that
				2018	<tt>BasicBlock</tt>, and a newly-created instruction we wish to insert
				2019	before <tt>*pi</tt>, we do the following: </p>
				2020
				2021	<div class="doc_code">
				2022	<pre>
				2023	BasicBlock *pb = ...;
				2024	Instruction *pi = ...;
				2025	Instruction *newInst = new Instruction(...);
				2026
				2027	pb->getInstList().insert(pi, newInst); // <i>Inserts newInst before pi in pb</i>
				2028	</pre>
				2029	</div>
				2030
				2031	<p>Appending to the end of a <tt>BasicBlock</tt> is so common that
				2032	the <tt>Instruction</tt> class and <tt>Instruction</tt>-derived
				2033	classes provide constructors which take a pointer to a
				2034	<tt>BasicBlock</tt> to be appended to. For example code that
				2035	looked like: </p>
				2036
				2037	<div class="doc_code">
				2038	<pre>
				2039	BasicBlock *pb = ...;
				2040	Instruction *newInst = new Instruction(...);
				2041
				2042	pb->getInstList().push_back(newInst); // <i>Appends newInst to pb</i>
				2043	</pre>
				2044	</div>
				2045
				2046	<p>becomes: </p>
				2047
				2048	<div class="doc_code">
				2049	<pre>
				2050	BasicBlock *pb = ...;
				2051	Instruction *newInst = new Instruction(..., pb);
				2052	</pre>
				2053	</div>
				2054
				2055	<p>which is much cleaner, especially if you are creating
				2056	long instruction streams.</p></li>
				2057
				2058	<li>Insertion into an implicit instruction list
				2059
				2060	<p><tt>Instruction</tt> instances that are already in <tt>BasicBlock</tt>s
				2061	are implicitly associated with an existing instruction list: the instruction
				2062	list of the enclosing basic block. Thus, we could have accomplished the same
				2063	thing as the above code without being given a <tt>BasicBlock</tt> by doing:
				2064	</p>
				2065
				2066	<div class="doc_code">
				2067	<pre>
				2068	Instruction *pi = ...;
				2069	Instruction *newInst = new Instruction(...);
				2070
				2071	pi->getParent()->getInstList().insert(pi, newInst);
				2072	</pre>
				2073	</div>
				2074
				2075	<p>In fact, this sequence of steps occurs so frequently that the
				2076	<tt>Instruction</tt> class and <tt>Instruction</tt>-derived classes provide
				2077	constructors which take (as a default parameter) a pointer to an
				2078	<tt>Instruction</tt> which the newly-created <tt>Instruction</tt> should
				2079	precede. That is, <tt>Instruction</tt> constructors are capable of
				2080	inserting the newly-created instance into the <tt>BasicBlock</tt> of a
				2081	provided instruction, immediately before that instruction. Using an
				2082	<tt>Instruction</tt> constructor with a <tt>insertBefore</tt> (default)
				2083	parameter, the above code becomes:</p>
				2084
				2085	<div class="doc_code">
				2086	<pre>
				2087	Instruction* pi = ...;
				2088	Instruction* newInst = new Instruction(..., pi);
				2089	</pre>
				2090	</div>
				2091
				2092	<p>which is much cleaner, especially if you're creating a lot of
				2093	instructions and adding them to <tt>BasicBlock</tt>s.</p></li>
				2094	</ul>
				2095
				2096	</div>
				2097
				2098	<!--_______________________________________________________________________-->
				2099	<div class="doc_subsubsection">
				2100	<a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a>
				2101	</div>
				2102
				2103	<div class="doc_text">
				2104
				2105	<p>Deleting an instruction from an existing sequence of instructions that form a
				2106	<a href="#BasicBlock"><tt>BasicBlock</tt></a> is very straight-forward. First,
				2107	you must have a pointer to the instruction that you wish to delete. Second, you
				2108	need to obtain the pointer to that instruction's basic block. You use the
				2109	pointer to the basic block to get its list of instructions and then use the
				2110	erase function to remove your instruction. For example:</p>
				2111
				2112	<div class="doc_code">
				2113	<pre>
				2114	<a href="#Instruction">Instruction</a> *I = .. ;
Chris Lattner	3db8f77	2008-02-15 22:57:17 +0000	[diff] [blame]	2115	I->eraseFromParent();
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2116	</pre>
				2117	</div>
				2118
				2119	</div>
				2120
				2121	<!--_______________________________________________________________________-->
				2122	<div class="doc_subsubsection">
				2123	<a name="schanges_replacing">Replacing an <tt>Instruction</tt> with another
				2124	<tt>Value</tt></a>
				2125	</div>
				2126
				2127	<div class="doc_text">
				2128
				2129	<p><i>Replacing individual instructions</i></p>
				2130
				2131	<p>Including "<a href="/doxygen/BasicBlockUtils_8h-source.html">llvm/Transforms/Utils/BasicBlockUtils.h</a>"
				2132	permits use of two very useful replace functions: <tt>ReplaceInstWithValue</tt>
				2133	and <tt>ReplaceInstWithInst</tt>.</p>
				2134
				2135	<h4><a name="schanges_deleting">Deleting <tt>Instruction</tt>s</a></h4>
				2136
				2137	<ul>
				2138	<li><tt>ReplaceInstWithValue</tt>
				2139
Nick Lewycky	48d4b03	2008-09-15 06:31:52 +0000	[diff] [blame]	2140	<p>This function replaces all uses of a given instruction with a value,
				2141	and then removes the original instruction. The following example
				2142	illustrates the replacement of the result of a particular
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2143	<tt>AllocaInst</tt> that allocates memory for a single integer with a null
				2144	pointer to an integer.</p>
				2145
				2146	<div class="doc_code">
				2147	<pre>
				2148	AllocaInst* instToReplace = ...;
				2149	BasicBlock::iterator ii(instToReplace);
				2150
				2151	ReplaceInstWithValue(instToReplace->getParent()->getInstList(), ii,
Daniel Dunbar	8ce7962	2008-10-03 22:17:25 +0000	[diff] [blame]	2152	Constant::getNullValue(PointerType::getUnqual(Type::Int32Ty)));
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2153	</pre></div></li>
				2154
				2155	<li><tt>ReplaceInstWithInst</tt>
				2156
				2157	<p>This function replaces a particular instruction with another
Nick Lewycky	48d4b03	2008-09-15 06:31:52 +0000	[diff] [blame]	2158	instruction, inserting the new instruction into the basic block at the
				2159	location where the old instruction was, and replacing any uses of the old
				2160	instruction with the new instruction. The following example illustrates
				2161	the replacement of one <tt>AllocaInst</tt> with another.</p>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2162
				2163	<div class="doc_code">
				2164	<pre>
				2165	AllocaInst* instToReplace = ...;
				2166	BasicBlock::iterator ii(instToReplace);
				2167
				2168	ReplaceInstWithInst(instToReplace->getParent()->getInstList(), ii,
Nick Lewycky	c004e5b	2007-12-03 01:52:52 +0000	[diff] [blame]	2169	new AllocaInst(Type::Int32Ty, 0, "ptrToReplacedInt"));
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2170	</pre></div></li>
				2171	</ul>
				2172
				2173	<p><i>Replacing multiple uses of <tt>User</tt>s and <tt>Value</tt>s</i></p>
				2174
				2175	<p>You can use <tt>Value::replaceAllUsesWith</tt> and
				2176	<tt>User::replaceUsesOfWith</tt> to change more than one use at a time. See the
				2177	doxygen documentation for the <a href="/doxygen/classllvm_1_1Value.html">Value Class</a>
				2178	and <a href="/doxygen/classllvm_1_1User.html">User Class</a>, respectively, for more
				2179	information.</p>
				2180
				2181	<!-- Value::replaceAllUsesWith User::replaceUsesOfWith Point out:
				2182	include/llvm/Transforms/Utils/ especially BasicBlockUtils.h with:
				2183	ReplaceInstWithValue, ReplaceInstWithInst -->
				2184
				2185	</div>
				2186
				2187	<!--_______________________________________________________________________-->
				2188	<div class="doc_subsubsection">
				2189	<a name="schanges_deletingGV">Deleting <tt>GlobalVariable</tt>s</a>
				2190	</div>
				2191
				2192	<div class="doc_text">
				2193
				2194	<p>Deleting a global variable from a module is just as easy as deleting an
				2195	Instruction. First, you must have a pointer to the global variable that you wish
				2196	to delete. You use this pointer to erase it from its parent, the module.
				2197	For example:</p>
				2198
				2199	<div class="doc_code">
				2200	<pre>
				2201	<a href="#GlobalVariable">GlobalVariable</a> *GV = .. ;
				2202
				2203	GV->eraseFromParent();
				2204	</pre>
				2205	</div>
				2206
				2207	</div>
				2208
Jeffrey Yasskin	7ebb6ac	2009-04-30 22:33:41 +0000	[diff] [blame]	2209	<!-- ======================================================================= -->
				2210	<div class="doc_subsection">
				2211	<a name="create_types">How to Create Types</a>
				2212	</div>
				2213
				2214	<div class="doc_text">
				2215
				2216	<p>In generating IR, you may need some complex types. If you know these types
Misha Brukman	0672513	2009-05-01 20:40:51 +0000	[diff] [blame]	2217	statically, you can use <tt>TypeBuilder<...>::get()</tt>, defined
Jeffrey Yasskin	7ebb6ac	2009-04-30 22:33:41 +0000	[diff] [blame]	2218	in <tt>llvm/Support/TypeBuilder.h</tt>, to retrieve them. <tt>TypeBuilder</tt>
				2219	has two forms depending on whether you're building types for cross-compilation
Misha Brukman	0672513	2009-05-01 20:40:51 +0000	[diff] [blame]	2220	or native library use. <tt>TypeBuilder<T, true></tt> requires
Jeffrey Yasskin	7ebb6ac	2009-04-30 22:33:41 +0000	[diff] [blame]	2221	that <tt>T</tt> be independent of the host environment, meaning that it's built
				2222	out of types from
				2223	the <a href="/doxygen/namespacellvm_1_1types.html"><tt>llvm::types</tt></a>
				2224	namespace and pointers, functions, arrays, etc. built of
Misha Brukman	0672513	2009-05-01 20:40:51 +0000	[diff] [blame]	2225	those. <tt>TypeBuilder<T, false></tt> additionally allows native C types
Jeffrey Yasskin	7ebb6ac	2009-04-30 22:33:41 +0000	[diff] [blame]	2226	whose size may depend on the host compiler. For example,</p>
				2227
				2228	<div class="doc_code">
				2229	<pre>
Misha Brukman	0672513	2009-05-01 20:40:51 +0000	[diff] [blame]	2230	FunctionType ft = TypeBuilder<types::i<8>(types::i<32>), true>::get();
Jeffrey Yasskin	7ebb6ac	2009-04-30 22:33:41 +0000	[diff] [blame]	2231	</pre>
				2232	</div>
				2233
				2234	<p>is easier to read and write than the equivalent</p>
				2235
				2236	<div class="doc_code">
				2237	<pre>
Owen Anderson	60f8757	2009-06-16 17:40:28 +0000	[diff] [blame]	2238	std::vector<const Type*> params;
Jeffrey Yasskin	7ebb6ac	2009-04-30 22:33:41 +0000	[diff] [blame]	2239	params.push_back(PointerType::getUnqual(Type::Int32Ty));
				2240	FunctionType *ft = FunctionType::get(Type::Int8Ty, params, false);
				2241	</pre>
				2242	</div>
				2243
				2244	<p>See the <a href="/doxygen/TypeBuilder_8h-source.html#l00001">class
				2245	comment</a> for more details.</p>
				2246
				2247	</div>
				2248
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2249	<!-- *********************************************************************** -->
				2250	<div class="doc_section">
Owen Anderson	e8370c0	2009-06-16 01:17:16 +0000	[diff] [blame]	2251	<a name="threading">Threads and LLVM</a>
				2252	</div>
				2253	<!-- *********************************************************************** -->
				2254
				2255	<div class="doc_text">
				2256	<p>
				2257	This section describes the interaction of the LLVM APIs with multithreading,
				2258	both on the part of client applications, and in the JIT, in the hosted
				2259	application.
				2260	</p>
				2261
				2262	<p>
				2263	Note that LLVM's support for multithreading is still relatively young. Up
				2264	through version 2.5, the execution of threaded hosted applications was
				2265	supported, but not threaded client access to the APIs. While this use case is
				2266	now supported, clients <em>must</em> adhere to the guidelines specified below to
				2267	ensure proper operation in multithreaded mode.
				2268	</p>
				2269
				2270	<p>
				2271	Note that, on Unix-like platforms, LLVM requires the presence of GCC's atomic
				2272	intrinsics in order to support threaded operation. If you need a
				2273	multhreading-capable LLVM on a platform without a suitably modern system
				2274	compiler, consider compiling LLVM and LLVM-GCC in single-threaded mode, and
				2275	using the resultant compiler to build a copy of LLVM with multithreading
				2276	support.
				2277	</p>
				2278	</div>
				2279
				2280	<!-- ======================================================================= -->
				2281	<div class="doc_subsection">
Owen Anderson	b418690	2009-06-16 18:04:19 +0000	[diff] [blame]	2282	<a name="startmultithreaded">Entering and Exiting Multithreaded Mode</a>
Owen Anderson	e8370c0	2009-06-16 01:17:16 +0000	[diff] [blame]	2283	</div>
				2284
				2285	<div class="doc_text">
				2286
				2287	<p>
				2288	In order to properly protect its internal data structures while avoiding
Owen Anderson	b418690	2009-06-16 18:04:19 +0000	[diff] [blame]	2289	excessive locking overhead in the single-threaded case, the LLVM must intialize
				2290	certain data structures necessary to provide guards around its internals. To do
				2291	so, the client program must invoke <tt>llvm_start_multithreaded()</tt> before
				2292	making any concurrent LLVM API calls. To subsequently tear down these
				2293	structures, use the <tt>llvm_stop_multithreaded()</tt> call. You can also use
				2294	the <tt>llvm_is_multithreaded()</tt> call to check the status of multithreaded
				2295	mode.
Owen Anderson	e8370c0	2009-06-16 01:17:16 +0000	[diff] [blame]	2296	</p>
				2297
				2298	<p>
Owen Anderson	b418690	2009-06-16 18:04:19 +0000	[diff] [blame]	2299	Note that both of these calls must be made <em>in isolation</em>. That is to
				2300	say that no other LLVM API calls may be executing at any time during the
				2301	execution of <tt>llvm_start_multithreaded()</tt> or <tt>llvm_stop_multithreaded
				2302	</tt>. It's is the client's responsibility to enforce this isolation.
				2303	</p>
				2304
				2305	<p>
				2306	The return value of <tt>llvm_start_multithreaded()</tt> indicates the success or
				2307	failure of the initialization. Failure typically indicates that your copy of
				2308	LLVM was built without multithreading support, typically because GCC atomic
				2309	intrinsics were not found in your system compiler. In this case, the LLVM API
				2310	will not be safe for concurrent calls. However, it <em>will</em> be safe for
				2311	hosting threaded applications in the JIT, though care must be taken to ensure
				2312	that side exits and the like do not accidentally result in concurrent LLVM API
				2313	calls.
Owen Anderson	e8370c0	2009-06-16 01:17:16 +0000	[diff] [blame]	2314	</p>
				2315	</div>
				2316
				2317	<!-- ======================================================================= -->
				2318	<div class="doc_subsection">
				2319	<a name="shutdown">Ending Execution with <tt>llvm_shutdown()</tt></a>
				2320	</div>
				2321
				2322	<div class="doc_text">
				2323	<p>
				2324	When you are done using the LLVM APIs, you should call <tt>llvm_shutdown()</tt>
Owen Anderson	b418690	2009-06-16 18:04:19 +0000	[diff] [blame]	2325	to deallocate memory used for internal structures. This will also invoke
				2326	<tt>llvm_stop_multithreaded()</tt> if LLVM is operating in multithreaded mode.
				2327	As such, <tt>llvm_shutdown()</tt> requires the same isolation guarantees as
				2328	<tt>llvm_stop_multithreaded()</tt>.
Owen Anderson	e8370c0	2009-06-16 01:17:16 +0000	[diff] [blame]	2329	</p>
				2330
				2331	<p>
				2332	Note that, if you use scope-based shutdown, you can use the
				2333	<tt>llvm_shutdown_obj</tt> class, which calls <tt>llvm_shutdown()</tt> in its
				2334	destructor.
				2335	</div>
				2336
				2337	<!-- ======================================================================= -->
				2338	<div class="doc_subsection">
				2339	<a name="managedstatic">Lazy Initialization with <tt>ManagedStatic</tt></a>
				2340	</div>
				2341
				2342	<div class="doc_text">
				2343	<p>
				2344	<tt>ManagedStatic</tt> is a utility class in LLVM used to implement static
				2345	initialization of static resources, such as the global type tables. Before the
				2346	invocation of <tt>llvm_shutdown()</tt>, it implements a simple lazy
				2347	initialization scheme. Once <tt>llvm_start_multithreaded()</tt> returns,
				2348	however, it uses double-checked locking to implement thread-safe lazy
				2349	initialization.
				2350	</p>
				2351
				2352	<p>
				2353	Note that, because no other threads are allowed to issue LLVM API calls before
				2354	<tt>llvm_start_multithreaded()</tt> returns, it is possible to have
				2355	<tt>ManagedStatic</tt>s of <tt>llvm::sys::Mutex</tt>s.
				2356	</p>
Owen Anderson	b418690	2009-06-16 18:04:19 +0000	[diff] [blame]	2357
				2358	<p>
				2359	The <tt>llvm_acquire_global_lock()</tt> and <tt>llvm_release_global_lock</tt>
				2360	APIs provide access to the global lock used to implement the double-checked
				2361	locking for lazy initialization. These should only be used internally to LLVM,
				2362	and only if you know what you're doing!
				2363	</p>
Owen Anderson	e8370c0	2009-06-16 01:17:16 +0000	[diff] [blame]	2364	</div>
				2365
				2366	<!-- *********************************************************************** -->
				2367	<div class="doc_section">
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2368	<a name="advanced">Advanced Topics</a>
				2369	</div>
				2370	<!-- *********************************************************************** -->
				2371
				2372	<div class="doc_text">
				2373	<p>
				2374	This section describes some of the advanced or obscure API's that most clients
				2375	do not need to be aware of. These API's tend manage the inner workings of the
				2376	LLVM system, and only need to be accessed in unusual circumstances.
				2377	</p>
				2378	</div>
				2379
				2380	<!-- ======================================================================= -->
				2381	<div class="doc_subsection">
				2382	<a name="TypeResolve">LLVM Type Resolution</a>
				2383	</div>
				2384
				2385	<div class="doc_text">
				2386
				2387	<p>
				2388	The LLVM type system has a very simple goal: allow clients to compare types for
				2389	structural equality with a simple pointer comparison (aka a shallow compare).
				2390	This goal makes clients much simpler and faster, and is used throughout the LLVM
				2391	system.
				2392	</p>
				2393
				2394	<p>
				2395	Unfortunately achieving this goal is not a simple matter. In particular,
				2396	recursive types and late resolution of opaque types makes the situation very
				2397	difficult to handle. Fortunately, for the most part, our implementation makes
				2398	most clients able to be completely unaware of the nasty internal details. The
				2399	primary case where clients are exposed to the inner workings of it are when
				2400	building a recursive type. In addition to this case, the LLVM bitcode reader,
				2401	assembly parser, and linker also have to be aware of the inner workings of this
				2402	system.
				2403	</p>
				2404
				2405	<p>
				2406	For our purposes below, we need three concepts. First, an "Opaque Type" is
				2407	exactly as defined in the <a href="LangRef.html#t_opaque">language
				2408	reference</a>. Second an "Abstract Type" is any type which includes an
				2409	opaque type as part of its type graph (for example "<tt>{ opaque, i32 }</tt>").
				2410	Third, a concrete type is a type that is not an abstract type (e.g. "<tt>{ i32,
				2411	float }</tt>").
				2412	</p>
				2413
				2414	</div>
				2415
				2416	<!-- ______________________________________________________________________ -->
				2417	<div class="doc_subsubsection">
				2418	<a name="BuildRecType">Basic Recursive Type Construction</a>
				2419	</div>
				2420
				2421	<div class="doc_text">
				2422
				2423	<p>
				2424	Because the most common question is "how do I build a recursive type with LLVM",
				2425	we answer it now and explain it as we go. Here we include enough to cause this
				2426	to be emitted to an output .ll file:
				2427	</p>
				2428
				2429	<div class="doc_code">
				2430	<pre>
				2431	%mylist = type { %mylist*, i32 }
				2432	</pre>
				2433	</div>
				2434
				2435	<p>
				2436	To build this, use the following LLVM APIs:
				2437	</p>
				2438
				2439	<div class="doc_code">
				2440	<pre>
				2441	// <i>Create the initial outer struct</i>
				2442	<a href="#PATypeHolder">PATypeHolder</a> StructTy = OpaqueType::get();
				2443	std::vector<const Type*> Elts;
Daniel Dunbar	8ce7962	2008-10-03 22:17:25 +0000	[diff] [blame]	2444	Elts.push_back(PointerType::getUnqual(StructTy));
Nick Lewycky	c004e5b	2007-12-03 01:52:52 +0000	[diff] [blame]	2445	Elts.push_back(Type::Int32Ty);
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2446	StructType *NewSTy = StructType::get(Elts);
				2447
				2448	// <i>At this point, NewSTy = "{ opaque*, i32 }". Tell VMCore that</i>
				2449	// <i>the struct and the opaque type are actually the same.</i>
				2450	cast<OpaqueType>(StructTy.get())-><a href="#refineAbstractTypeTo">refineAbstractTypeTo</a>(NewSTy);
				2451
				2452	// <i>NewSTy is potentially invalidated, but StructTy (a <a href="#PATypeHolder">PATypeHolder</a>) is</i>
				2453	// <i>kept up-to-date</i>
				2454	NewSTy = cast<StructType>(StructTy.get());
				2455
				2456	// <i>Add a name for the type to the module symbol table (optional)</i>
				2457	MyModule->addTypeName("mylist", NewSTy);
				2458	</pre>
				2459	</div>
				2460
				2461	<p>
				2462	This code shows the basic approach used to build recursive types: build a
				2463	non-recursive type using 'opaque', then use type unification to close the cycle.
				2464	The type unification step is performed by the <tt><a
				2465	href="#refineAbstractTypeTo">refineAbstractTypeTo</a></tt> method, which is
				2466	described next. After that, we describe the <a
				2467	href="#PATypeHolder">PATypeHolder class</a>.
				2468	</p>
				2469
				2470	</div>
				2471
				2472	<!-- ______________________________________________________________________ -->
				2473	<div class="doc_subsubsection">
				2474	<a name="refineAbstractTypeTo">The <tt>refineAbstractTypeTo</tt> method</a>
				2475	</div>
				2476
				2477	<div class="doc_text">
				2478	<p>
				2479	The <tt>refineAbstractTypeTo</tt> method starts the type unification process.
				2480	While this method is actually a member of the DerivedType class, it is most
				2481	often used on OpaqueType instances. Type unification is actually a recursive
				2482	process. After unification, types can become structurally isomorphic to
				2483	existing types, and all duplicates are deleted (to preserve pointer equality).
				2484	</p>
				2485
				2486	<p>
				2487	In the example above, the OpaqueType object is definitely deleted.
				2488	Additionally, if there is an "{ \2*, i32}" type already created in the system,
				2489	the pointer and struct type created are <b>also</b> deleted. Obviously whenever
				2490	a type is deleted, any "Type*" pointers in the program are invalidated. As
				2491	such, it is safest to avoid having <i>any</i> "Type*" pointers to abstract types
				2492	live across a call to <tt>refineAbstractTypeTo</tt> (note that non-abstract
				2493	types can never move or be deleted). To deal with this, the <a
				2494	href="#PATypeHolder">PATypeHolder</a> class is used to maintain a stable
				2495	reference to a possibly refined type, and the <a
				2496	href="#AbstractTypeUser">AbstractTypeUser</a> class is used to update more
				2497	complex datastructures.
				2498	</p>
				2499
				2500	</div>
				2501
				2502	<!-- ______________________________________________________________________ -->
				2503	<div class="doc_subsubsection">
				2504	<a name="PATypeHolder">The PATypeHolder Class</a>
				2505	</div>
				2506
				2507	<div class="doc_text">
				2508	<p>
				2509	PATypeHolder is a form of a "smart pointer" for Type objects. When VMCore
				2510	happily goes about nuking types that become isomorphic to existing types, it
				2511	automatically updates all PATypeHolder objects to point to the new type. In the
				2512	example above, this allows the code to maintain a pointer to the resultant
				2513	resolved recursive type, even though the Type*'s are potentially invalidated.
				2514	</p>
				2515
				2516	<p>
				2517	PATypeHolder is an extremely light-weight object that uses a lazy union-find
				2518	implementation to update pointers. For example the pointer from a Value to its
				2519	Type is maintained by PATypeHolder objects.
				2520	</p>
				2521
				2522	</div>
				2523
				2524	<!-- ______________________________________________________________________ -->
				2525	<div class="doc_subsubsection">
				2526	<a name="AbstractTypeUser">The AbstractTypeUser Class</a>
				2527	</div>
				2528
				2529	<div class="doc_text">
				2530
				2531	<p>
				2532	Some data structures need more to perform more complex updates when types get
				2533	resolved. To support this, a class can derive from the AbstractTypeUser class.
				2534	This class
				2535	allows it to get callbacks when certain types are resolved. To register to get
				2536	callbacks for a particular type, the DerivedType::{add/remove}AbstractTypeUser
				2537	methods can be called on a type. Note that these methods only work for <i>
				2538	abstract</i> types. Concrete types (those that do not include any opaque
				2539	objects) can never be refined.
				2540	</p>
				2541	</div>
				2542
				2543
				2544	<!-- ======================================================================= -->
				2545	<div class="doc_subsection">
				2546	<a name="SymbolTable">The <tt>ValueSymbolTable</tt> and
				2547	<tt>TypeSymbolTable</tt> classes</a>
				2548	</div>
				2549
				2550	<div class="doc_text">
				2551	<p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1ValueSymbolTable.html">
				2552	ValueSymbolTable</a></tt> class provides a symbol table that the <a
				2553	href="#Function"><tt>Function</tt></a> and <a href="#Module">
				2554	<tt>Module</tt></a> classes use for naming value definitions. The symbol table
				2555	can provide a name for any <a href="#Value"><tt>Value</tt></a>.
				2556	The <tt><a href="http://llvm.org/doxygen/classllvm_1_1TypeSymbolTable.html">
				2557	TypeSymbolTable</a></tt> class is used by the <tt>Module</tt> class to store
				2558	names for types.</p>
				2559
				2560	<p>Note that the <tt>SymbolTable</tt> class should not be directly accessed
				2561	by most clients. It should only be used when iteration over the symbol table
				2562	names themselves are required, which is very special purpose. Note that not
				2563	all LLVM
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2564	<tt><a href="#Value">Value</a></tt>s have names, and those without names (i.e. they have
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2565	an empty name) do not exist in the symbol table.
				2566	</p>
				2567
				2568	<p>These symbol tables support iteration over the values/types in the symbol
				2569	table with <tt>begin/end/iterator</tt> and supports querying to see if a
				2570	specific name is in the symbol table (with <tt>lookup</tt>). The
				2571	<tt>ValueSymbolTable</tt> class exposes no public mutator methods, instead,
				2572	simply call <tt>setName</tt> on a value, which will autoinsert it into the
				2573	appropriate symbol table. For types, use the Module::addTypeName method to
				2574	insert entries into the symbol table.</p>
				2575
				2576	</div>
				2577
				2578
				2579
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2580	<!-- ======================================================================= -->
				2581	<div class="doc_subsection">
				2582	<a name="UserLayout">The <tt>User</tt> and owned <tt>Use</tt> classes' memory layout</a>
				2583	</div>
				2584
				2585	<div class="doc_text">
				2586	<p>The <tt><a href="http://llvm.org/doxygen/classllvm_1_1User.html">
Gabor Greif	50626fc	2009-01-05 16:05:32 +0000	[diff] [blame]	2587	User</a></tt> class provides a basis for expressing the ownership of <tt>User</tt>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2588	towards other <tt><a href="http://llvm.org/doxygen/classllvm_1_1Value.html">
				2589	Value</a></tt>s. The <tt><a href="http://llvm.org/doxygen/classllvm_1_1Use.html">
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2590	Use</a></tt> helper class is employed to do the bookkeeping and to facilitate <i>O(1)</i>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2591	addition and removal.</p>
				2592
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2593	<!-- ______________________________________________________________________ -->
				2594	<div class="doc_subsubsection">
Gabor Greif	50626fc	2009-01-05 16:05:32 +0000	[diff] [blame]	2595	<a name="Use2User">Interaction and relationship between <tt>User</tt> and <tt>Use</tt> objects</a>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2596	</div>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2597
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2598	<div class="doc_text">
				2599	<p>
				2600	A subclass of <tt>User</tt> can choose between incorporating its <tt>Use</tt> objects
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2601	or refer to them out-of-line by means of a pointer. A mixed variant
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2602	(some <tt>Use</tt>s inline others hung off) is impractical and breaks the invariant
				2603	that the <tt>Use</tt> objects belonging to the same <tt>User</tt> form a contiguous array.
				2604	</p>
				2605	</div>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2606
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2607	<p>
				2608	We have 2 different layouts in the <tt>User</tt> (sub)classes:
				2609	<ul>
				2610	<li><p>Layout a)
				2611	The <tt>Use</tt> object(s) are inside (resp. at fixed offset) of the <tt>User</tt>
				2612	object and there are a fixed number of them.</p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2613
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2614	<li><p>Layout b)
				2615	The <tt>Use</tt> object(s) are referenced by a pointer to an
				2616	array from the <tt>User</tt> object and there may be a variable
				2617	number of them.</p>
				2618	</ul>
				2619	<p>
Gabor Greif	e247e36	2008-06-25 00:10:22 +0000	[diff] [blame]	2620	As of v2.4 each layout still possesses a direct pointer to the
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2621	start of the array of <tt>Use</tt>s. Though not mandatory for layout a),
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2622	we stick to this redundancy for the sake of simplicity.
Gabor Greif	e247e36	2008-06-25 00:10:22 +0000	[diff] [blame]	2623	The <tt>User</tt> object also stores the number of <tt>Use</tt> objects it
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2624	has. (Theoretically this information can also be calculated
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2625	given the scheme presented below.)</p>
				2626	<p>
				2627	Special forms of allocation operators (<tt>operator new</tt>)
Gabor Greif	e247e36	2008-06-25 00:10:22 +0000	[diff] [blame]	2628	enforce the following memory layouts:</p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2629
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2630	<ul>
Gabor Greif	e247e36	2008-06-25 00:10:22 +0000	[diff] [blame]	2631	<li><p>Layout a) is modelled by prepending the <tt>User</tt> object by the <tt>Use[]</tt> array.</p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2632
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2633	<pre>
				2634	...---.---.---.---.-------...
				2635	\| P \| P \| P \| P \| User
				2636	'''---'---'---'---'-------'''
				2637	</pre>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2638
Gabor Greif	e247e36	2008-06-25 00:10:22 +0000	[diff] [blame]	2639	<li><p>Layout b) is modelled by pointing at the <tt>Use[]</tt> array.</p>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2640	<pre>
				2641	.-------...
				2642	\| User
				2643	'-------'''
				2644	\|
				2645	v
				2646	.---.---.---.---...
				2647	\| P \| P \| P \| P \|
				2648	'---'---'---'---'''
				2649	</pre>
				2650	</ul>
				2651	<i>(In the above figures '<tt>P</tt>' stands for the <tt>Use**</tt> that
				2652	is stored in each <tt>Use</tt> object in the member <tt>Use::Prev</tt>)</i>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2653
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2654	<!-- ______________________________________________________________________ -->
				2655	<div class="doc_subsubsection">
Gabor Greif	50626fc	2009-01-05 16:05:32 +0000	[diff] [blame]	2656	<a name="Waymarking">The waymarking algorithm</a>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2657	</div>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2658
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2659	<div class="doc_text">
				2660	<p>
Gabor Greif	e247e36	2008-06-25 00:10:22 +0000	[diff] [blame]	2661	Since the <tt>Use</tt> objects are deprived of the direct (back)pointer to
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2662	their <tt>User</tt> objects, there must be a fast and exact method to
				2663	recover it. This is accomplished by the following scheme:</p>
				2664	</div>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2665
Gabor Greif	e247e36	2008-06-25 00:10:22 +0000	[diff] [blame]	2666	A bit-encoding in the 2 LSBits (least significant bits) of the <tt>Use::Prev</tt> allows to find the
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2667	start of the <tt>User</tt> object:
				2668	<ul>
				2669	<li><tt>00</tt> —> binary digit 0</li>
				2670	<li><tt>01</tt> —> binary digit 1</li>
				2671	<li><tt>10</tt> —> stop and calculate (<tt>s</tt>)</li>
				2672	<li><tt>11</tt> —> full stop (<tt>S</tt>)</li>
				2673	</ul>
				2674	<p>
				2675	Given a <tt>Use*</tt>, all we have to do is to walk till we get
				2676	a stop and we either have a <tt>User</tt> immediately behind or
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2677	we have to walk to the next stop picking up digits
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2678	and calculating the offset:</p>
				2679	<pre>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2680	.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.---.----------------
				2681	\| 1 \| s \| 1 \| 0 \| 1 \| 0 \| s \| 1 \| 1 \| 0 \| s \| 1 \| 1 \| s \| 1 \| S \| User (or User*)
				2682	'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'---'----------------
				2683	\|+15 \|+10 \|+6 \|+3 \|+1
				2684	\| \| \| \| \|__>
				2685	\| \| \| \|__________>
				2686	\| \| \|______________________>
				2687	\| \|______________________________________>
				2688	\|__________________________________________________________>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2689	</pre>
				2690	<p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2691	Only the significant number of bits need to be stored between the
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2692	stops, so that the <i>worst case is 20 memory accesses</i> when there are
				2693	1000 <tt>Use</tt> objects associated with a <tt>User</tt>.</p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2694
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2695	<!-- ______________________________________________________________________ -->
				2696	<div class="doc_subsubsection">
Gabor Greif	50626fc	2009-01-05 16:05:32 +0000	[diff] [blame]	2697	<a name="ReferenceImpl">Reference implementation</a>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2698	</div>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2699
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2700	<div class="doc_text">
				2701	<p>
				2702	The following literate Haskell fragment demonstrates the concept:</p>
				2703	</div>
				2704
				2705	<div class="doc_code">
				2706	<pre>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2707	> import Test.QuickCheck
				2708	>
				2709	> digits :: Int -> [Char] -> [Char]
				2710	> digits 0 acc = '0' : acc
				2711	> digits 1 acc = '1' : acc
				2712	> digits n acc = digits (n `div` 2) $ digits (n `mod` 2) acc
				2713	>
				2714	> dist :: Int -> [Char] -> [Char]
				2715	> dist 0 [] = ['S']
				2716	> dist 0 acc = acc
				2717	> dist 1 acc = let r = dist 0 acc in 's' : digits (length r) r
				2718	> dist n acc = dist (n - 1) $ dist 1 acc
				2719	>
				2720	> takeLast n ss = reverse $ take n $ reverse ss
				2721	>
				2722	> test = takeLast 40 $ dist 20 []
				2723	>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2724	</pre>
				2725	</div>
				2726	<p>
				2727	Printing <test> gives: <tt>"1s100000s11010s10100s1111s1010s110s11s1S"</tt></p>
				2728	<p>
				2729	The reverse algorithm computes the length of the string just by examining
				2730	a certain prefix:</p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2731
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2732	<div class="doc_code">
				2733	<pre>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2734	> pref :: [Char] -> Int
				2735	> pref "S" = 1
				2736	> pref ('s':'1':rest) = decode 2 1 rest
				2737	> pref (_:rest) = 1 + pref rest
				2738	>
				2739	> decode walk acc ('0':rest) = decode (walk + 1) (acc * 2) rest
				2740	> decode walk acc ('1':rest) = decode (walk + 1) (acc * 2 + 1) rest
				2741	> decode walk acc _ = walk + acc
				2742	>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2743	</pre>
				2744	</div>
				2745	<p>
				2746	Now, as expected, printing <pref test> gives <tt>40</tt>.</p>
				2747	<p>
				2748	We can <i>quickCheck</i> this with following property:</p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2749
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2750	<div class="doc_code">
				2751	<pre>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2752	> testcase = dist 2000 []
				2753	> testcaseLength = length testcase
				2754	>
				2755	> identityProp n = n > 0 && n <= testcaseLength ==> length arr == pref arr
				2756	> where arr = takeLast n testcase
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2757	>
				2758	</pre>
				2759	</div>
				2760	<p>
				2761	As expected <quickCheck identityProp> gives:</p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2762
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2763	<pre>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2764	*Main> quickCheck identityProp
				2765	OK, passed 100 tests.
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2766	</pre>
				2767	<p>
				2768	Let's be a bit more exhaustive:</p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2769
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2770	<div class="doc_code">
				2771	<pre>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2772	>
				2773	> deepCheck p = check (defaultConfig { configMaxTest = 500 }) p
				2774	>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2775	</pre>
				2776	</div>
				2777	<p>
				2778	And here is the result of <deepCheck identityProp>:</p>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2779
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2780	<pre>
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2781	*Main> deepCheck identityProp
				2782	OK, passed 500 tests.
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2783	</pre>
				2784
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2785	<!-- ______________________________________________________________________ -->
				2786	<div class="doc_subsubsection">
Gabor Greif	50626fc	2009-01-05 16:05:32 +0000	[diff] [blame]	2787	<a name="Tagging">Tagging considerations</a>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2788	</div>
				2789
				2790	<p>
				2791	To maintain the invariant that the 2 LSBits of each <tt>Use**</tt> in <tt>Use</tt>
				2792	never change after being set up, setters of <tt>Use::Prev</tt> must re-tag the
				2793	new <tt>Use**</tt> on every modification. Accordingly getters must strip the
				2794	tag bits.</p>
				2795	<p>
Gabor Greif	e247e36	2008-06-25 00:10:22 +0000	[diff] [blame]	2796	For layout b) instead of the <tt>User</tt> we find a pointer (<tt>User*</tt> with LSBit set).
				2797	Following this pointer brings us to the <tt>User</tt>. A portable trick ensures
				2798	that the first bytes of <tt>User</tt> (if interpreted as a pointer) never has
Gabor Greif	50626fc	2009-01-05 16:05:32 +0000	[diff] [blame]	2799	the LSBit set. (Portability is relying on the fact that all known compilers place the
				2800	<tt>vptr</tt> in the first word of the instances.)</p>
Gabor Greif	93b462b	2008-06-18 13:44:57 +0000	[diff] [blame]	2801
Gabor Greif	92e8776	2008-06-16 21:06:12 +0000	[diff] [blame]	2802	</div>
				2803
				2804	<!-- *********************************************************************** -->
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2805	<div class="doc_section">
				2806	<a name="coreclasses">The Core LLVM Class Hierarchy Reference </a>
				2807	</div>
				2808	<!-- *********************************************************************** -->
				2809
				2810	<div class="doc_text">
				2811	<p><tt>#include "<a href="/doxygen/Type_8h-source.html">llvm/Type.h</a>"</tt>
				2812	<br>doxygen info: <a href="/doxygen/classllvm_1_1Type.html">Type Class</a></p>
				2813
				2814	<p>The Core LLVM classes are the primary means of representing the program
				2815	being inspected or transformed. The core LLVM classes are defined in
				2816	header files in the <tt>include/llvm/</tt> directory, and implemented in
				2817	the <tt>lib/VMCore</tt> directory.</p>
				2818
				2819	</div>
				2820
				2821	<!-- ======================================================================= -->
				2822	<div class="doc_subsection">
				2823	<a name="Type">The <tt>Type</tt> class and Derived Types</a>
				2824	</div>
				2825
				2826	<div class="doc_text">
				2827
				2828	<p><tt>Type</tt> is a superclass of all type classes. Every <tt>Value</tt> has
				2829	a <tt>Type</tt>. <tt>Type</tt> cannot be instantiated directly but only
				2830	through its subclasses. Certain primitive types (<tt>VoidType</tt>,
				2831	<tt>LabelType</tt>, <tt>FloatType</tt> and <tt>DoubleType</tt>) have hidden
				2832	subclasses. They are hidden because they offer no useful functionality beyond
				2833	what the <tt>Type</tt> class offers except to distinguish themselves from
				2834	other subclasses of <tt>Type</tt>.</p>
				2835	<p>All other types are subclasses of <tt>DerivedType</tt>. Types can be
				2836	named, but this is not a requirement. There exists exactly
				2837	one instance of a given shape at any one time. This allows type equality to
				2838	be performed with address equality of the Type Instance. That is, given two
				2839	<tt>Type*</tt> values, the types are identical if the pointers are identical.
				2840	</p>
				2841	</div>
				2842
				2843	<!-- _______________________________________________________________________ -->
				2844	<div class="doc_subsubsection">
Gabor Greif	50626fc	2009-01-05 16:05:32 +0000	[diff] [blame]	2845	<a name="m_Type">Important Public Methods</a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2846	</div>
				2847
				2848	<div class="doc_text">
				2849
				2850	<ul>
				2851	<li><tt>bool isInteger() const</tt>: Returns true for any integer type.</li>
				2852
				2853	<li><tt>bool isFloatingPoint()</tt>: Return true if this is one of the two
				2854	floating point types.</li>
				2855
				2856	<li><tt>bool isAbstract()</tt>: Return true if the type is abstract (contains
				2857	an OpaqueType anywhere in its definition).</li>
				2858
				2859	<li><tt>bool isSized()</tt>: Return true if the type has known size. Things
				2860	that don't have a size are abstract types, labels and void.</li>
				2861
				2862	</ul>
				2863	</div>
				2864
				2865	<!-- _______________________________________________________________________ -->
				2866	<div class="doc_subsubsection">
Gabor Greif	50626fc	2009-01-05 16:05:32 +0000	[diff] [blame]	2867	<a name="derivedtypes">Important Derived Types</a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	2868	</div>
				2869	<div class="doc_text">
				2870	<dl>
				2871	<dt><tt>IntegerType</tt></dt>
				2872	<dd>Subclass of DerivedType that represents integer types of any bit width.
				2873	Any bit width between <tt>IntegerType::MIN_INT_BITS</tt> (1) and
				2874	<tt>IntegerType::MAX_INT_BITS</tt> (~8 million) can be represented.
				2875	<ul>
				2876	<li><tt>static const IntegerType* get(unsigned NumBits)</tt>: get an integer
				2877	type of a specific bit width.</li>
				2878	<li><tt>unsigned getBitWidth() const</tt>: Get the bit width of an integer
				2879	type.</li>
				2880	</ul>
				2881	</dd>
				2882	<dt><tt>SequentialType</tt></dt>
				2883	<dd>This is subclassed by ArrayType and PointerType
				2884	<ul>
				2885	<li><tt>const Type * getElementType() const</tt>: Returns the type of each
				2886	of the elements in the sequential type. </li>
				2887	</ul>
				2888	</dd>
				2889	<dt><tt>ArrayType</tt></dt>
				2890	<dd>This is a subclass of SequentialType and defines the interface for array
				2891	types.
				2892	<ul>
				2893	<li><tt>unsigned getNumElements() const</tt>: Returns the number of
				2894	elements in the array. </li>
				2895	</ul>
				2896	</dd>
				2897	<dt><tt>PointerType</tt></dt>
				2898	<dd>Subclass of SequentialType for pointer types.</dd>
				2899	<dt><tt>VectorType</tt></dt>
				2900	<dd>Subclass of SequentialType for vector types. A
				2901	vector type is similar to an ArrayType but is distinguished because it is
				2902	a first class type wherease ArrayType is not. Vector types are used for
				2903	vector operations and are usually small vectors of of an integer or floating
				2904	point type.</dd>
				2905	<dt><tt>StructType</tt></dt>
				2906	<dd>Subclass of DerivedTypes for struct types.</dd>
				2907	<dt><tt><a name="FunctionType">FunctionType</a></tt></dt>
				2908	<dd>Subclass of DerivedTypes for function types.
				2909	<ul>
				2910	<li><tt>bool isVarArg() const</tt>: Returns true if its a vararg
				2911	function</li>
				2912	<li><tt> const Type * getReturnType() const</tt>: Returns the
				2913	return type of the function.</li>
				2914	<li><tt>const Type * getParamType (unsigned i)</tt>: Returns
				2915	the type of the ith parameter.</li>
				2916	<li><tt> const unsigned getNumParams() const</tt>: Returns the
				2917	number of formal parameters.</li>
				2918	</ul>
				2919	</dd>
				2920	<dt><tt>OpaqueType</tt></dt>
				2921	<dd>Sublcass of DerivedType for abstract types. This class
				2922	defines no content and is used as a placeholder for some other type. Note
				2923	that OpaqueType is used (temporarily) during type resolution for forward
				2924	references of types. Once the referenced type is resolved, the OpaqueType
				2925	is replaced with the actual type. OpaqueType can also be used for data
				2926	abstraction. At link time opaque types can be resolved to actual types
				2927	of the same name.</dd>
				2928	</dl>
				2929	</div>
				2930
				2931
				2932
				2933	<!-- ======================================================================= -->
				2934	<div class="doc_subsection">
				2935	<a name="Module">The <tt>Module</tt> class</a>
				2936	</div>
				2937
				2938	<div class="doc_text">
				2939
				2940	<p><tt>#include "<a
				2941	href="/doxygen/Module_8h-source.html">llvm/Module.h</a>"</tt><br> doxygen info:
				2942	<a href="/doxygen/classllvm_1_1Module.html">Module Class</a></p>
				2943
				2944	<p>The <tt>Module</tt> class represents the top level structure present in LLVM
				2945	programs. An LLVM module is effectively either a translation unit of the
				2946	original program or a combination of several translation units merged by the
				2947	linker. The <tt>Module</tt> class keeps track of a list of <a
				2948	href="#Function"><tt>Function</tt></a>s, a list of <a
				2949	href="#GlobalVariable"><tt>GlobalVariable</tt></a>s, and a <a
				2950	href="#SymbolTable"><tt>SymbolTable</tt></a>. Additionally, it contains a few
				2951	helpful member functions that try to make common operations easy.</p>
				2952
				2953	</div>
				2954
				2955	<!-- _______________________________________________________________________ -->
				2956	<div class="doc_subsubsection">
				2957	<a name="m_Module">Important Public Members of the <tt>Module</tt> class</a>
				2958	</div>
				2959
				2960	<div class="doc_text">
				2961
				2962	<ul>
				2963	<li><tt>Module::Module(std::string name = "")</tt></li>
				2964	</ul>
				2965
				2966	<p>Constructing a <a href="#Module">Module</a> is easy. You can optionally
				2967	provide a name for it (probably based on the name of the translation unit).</p>
				2968
				2969	<ul>
				2970	<li><tt>Module::iterator</tt> - Typedef for function list iterator<br>
				2971	<tt>Module::const_iterator</tt> - Typedef for const_iterator.<br>
				2972
				2973	<tt>begin()</tt>, <tt>end()</tt>
				2974	<tt>size()</tt>, <tt>empty()</tt>
				2975
				2976	<p>These are forwarding methods that make it easy to access the contents of
				2977	a <tt>Module</tt> object's <a href="#Function"><tt>Function</tt></a>
				2978	list.</p></li>
				2979
				2980	<li><tt>Module::FunctionListType &getFunctionList()</tt>
				2981
				2982	<p> Returns the list of <a href="#Function"><tt>Function</tt></a>s. This is
				2983	necessary to use when you need to update the list or perform a complex
				2984	action that doesn't have a forwarding method.</p>
				2985
				2986	<p><!-- Global Variable --></p></li>
				2987	</ul>
				2988
				2989	<hr>
				2990
				2991	<ul>
				2992	<li><tt>Module::global_iterator</tt> - Typedef for global variable list iterator<br>
				2993
				2994	<tt>Module::const_global_iterator</tt> - Typedef for const_iterator.<br>
				2995
				2996	<tt>global_begin()</tt>, <tt>global_end()</tt>
				2997	<tt>global_size()</tt>, <tt>global_empty()</tt>
				2998
				2999	<p> These are forwarding methods that make it easy to access the contents of
				3000	a <tt>Module</tt> object's <a
				3001	href="#GlobalVariable"><tt>GlobalVariable</tt></a> list.</p></li>
				3002
				3003	<li><tt>Module::GlobalListType &getGlobalList()</tt>
				3004
				3005	<p>Returns the list of <a
				3006	href="#GlobalVariable"><tt>GlobalVariable</tt></a>s. This is necessary to
				3007	use when you need to update the list or perform a complex action that
				3008	doesn't have a forwarding method.</p>
				3009
				3010	<p><!-- Symbol table stuff --> </p></li>
				3011	</ul>
				3012
				3013	<hr>
				3014
				3015	<ul>
				3016	<li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
				3017
				3018	<p>Return a reference to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
				3019	for this <tt>Module</tt>.</p>
				3020
				3021	<p><!-- Convenience methods --></p></li>
				3022	</ul>
				3023
				3024	<hr>
				3025
				3026	<ul>
				3027	<li><tt><a href="#Function">Function</a> *getFunction(const std::string
				3028	&Name, const <a href="#FunctionType">FunctionType</a> *Ty)</tt>
				3029
				3030	<p>Look up the specified function in the <tt>Module</tt> <a
				3031	href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, return
				3032	<tt>null</tt>.</p></li>
				3033
				3034	<li><tt><a href="#Function">Function</a> *getOrInsertFunction(const
				3035	std::string &Name, const <a href="#FunctionType">FunctionType</a> *T)</tt>
				3036
				3037	<p>Look up the specified function in the <tt>Module</tt> <a
				3038	href="#SymbolTable"><tt>SymbolTable</tt></a>. If it does not exist, add an
				3039	external declaration for the function and return it.</p></li>
				3040
				3041	<li><tt>std::string getTypeName(const <a href="#Type">Type</a> *Ty)</tt>
				3042
				3043	<p>If there is at least one entry in the <a
				3044	href="#SymbolTable"><tt>SymbolTable</tt></a> for the specified <a
				3045	href="#Type"><tt>Type</tt></a>, return it. Otherwise return the empty
				3046	string.</p></li>
				3047
				3048	<li><tt>bool addTypeName(const std::string &Name, const <a
				3049	href="#Type">Type</a> *Ty)</tt>
				3050
				3051	<p>Insert an entry in the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
				3052	mapping <tt>Name</tt> to <tt>Ty</tt>. If there is already an entry for this
				3053	name, true is returned and the <a
				3054	href="#SymbolTable"><tt>SymbolTable</tt></a> is not modified.</p></li>
				3055	</ul>
				3056
				3057	</div>
				3058
				3059
				3060	<!-- ======================================================================= -->
				3061	<div class="doc_subsection">
				3062	<a name="Value">The <tt>Value</tt> class</a>
				3063	</div>
				3064
				3065	<div class="doc_text">
				3066
				3067	<p><tt>#include "<a href="/doxygen/Value_8h-source.html">llvm/Value.h</a>"</tt>
				3068	<br>
				3069	doxygen info: <a href="/doxygen/classllvm_1_1Value.html">Value Class</a></p>
				3070
				3071	<p>The <tt>Value</tt> class is the most important class in the LLVM Source
				3072	base. It represents a typed value that may be used (among other things) as an
				3073	operand to an instruction. There are many different types of <tt>Value</tt>s,
				3074	such as <a href="#Constant"><tt>Constant</tt></a>s,<a
				3075	href="#Argument"><tt>Argument</tt></a>s. Even <a
				3076	href="#Instruction"><tt>Instruction</tt></a>s and <a
				3077	href="#Function"><tt>Function</tt></a>s are <tt>Value</tt>s.</p>
				3078
				3079	<p>A particular <tt>Value</tt> may be used many times in the LLVM representation
				3080	for a program. For example, an incoming argument to a function (represented
				3081	with an instance of the <a href="#Argument">Argument</a> class) is "used" by
				3082	every instruction in the function that references the argument. To keep track
				3083	of this relationship, the <tt>Value</tt> class keeps a list of all of the <a
				3084	href="#User"><tt>User</tt></a>s that is using it (the <a
				3085	href="#User"><tt>User</tt></a> class is a base class for all nodes in the LLVM
				3086	graph that can refer to <tt>Value</tt>s). This use list is how LLVM represents
				3087	def-use information in the program, and is accessible through the <tt>use_</tt>*
				3088	methods, shown below.</p>
				3089
				3090	<p>Because LLVM is a typed representation, every LLVM <tt>Value</tt> is typed,
				3091	and this <a href="#Type">Type</a> is available through the <tt>getType()</tt>
				3092	method. In addition, all LLVM values can be named. The "name" of the
				3093	<tt>Value</tt> is a symbolic string printed in the LLVM code:</p>
				3094
				3095	<div class="doc_code">
				3096	<pre>
				3097	%<b>foo</b> = add i32 1, 2
				3098	</pre>
				3099	</div>
				3100
				3101	<p><a name="nameWarning">The name of this instruction is "foo".</a> <b>NOTE</b>
				3102	that the name of any value may be missing (an empty string), so names should
				3103	<b>ONLY</b> be used for debugging (making the source code easier to read,
				3104	debugging printouts), they should not be used to keep track of values or map
				3105	between them. For this purpose, use a <tt>std::map</tt> of pointers to the
				3106	<tt>Value</tt> itself instead.</p>
				3107
				3108	<p>One important aspect of LLVM is that there is no distinction between an SSA
				3109	variable and the operation that produces it. Because of this, any reference to
				3110	the value produced by an instruction (or the value available as an incoming
				3111	argument, for example) is represented as a direct pointer to the instance of
				3112	the class that
				3113	represents this value. Although this may take some getting used to, it
				3114	simplifies the representation and makes it easier to manipulate.</p>
				3115
				3116	</div>
				3117
				3118	<!-- _______________________________________________________________________ -->
				3119	<div class="doc_subsubsection">
				3120	<a name="m_Value">Important Public Members of the <tt>Value</tt> class</a>
				3121	</div>
				3122
				3123	<div class="doc_text">
				3124
				3125	<ul>
				3126	<li><tt>Value::use_iterator</tt> - Typedef for iterator over the
				3127	use-list<br>
				3128	<tt>Value::use_const_iterator</tt> - Typedef for const_iterator over
				3129	the use-list<br>
				3130	<tt>unsigned use_size()</tt> - Returns the number of users of the
				3131	value.<br>
				3132	<tt>bool use_empty()</tt> - Returns true if there are no users.<br>
				3133	<tt>use_iterator use_begin()</tt> - Get an iterator to the start of
				3134	the use-list.<br>
				3135	<tt>use_iterator use_end()</tt> - Get an iterator to the end of the
				3136	use-list.<br>
				3137	<tt><a href="#User">User</a> *use_back()</tt> - Returns the last
				3138	element in the list.
				3139	<p> These methods are the interface to access the def-use
				3140	information in LLVM. As with all other iterators in LLVM, the naming
				3141	conventions follow the conventions defined by the <a href="#stl">STL</a>.</p>
				3142	</li>
				3143	<li><tt><a href="#Type">Type</a> *getType() const</tt>
				3144	<p>This method returns the Type of the Value.</p>
				3145	</li>
				3146	<li><tt>bool hasName() const</tt><br>
				3147	<tt>std::string getName() const</tt><br>
				3148	<tt>void setName(const std::string &Name)</tt>
				3149	<p> This family of methods is used to access and assign a name to a <tt>Value</tt>,
				3150	be aware of the <a href="#nameWarning">precaution above</a>.</p>
				3151	</li>
				3152	<li><tt>void replaceAllUsesWith(Value *V)</tt>
				3153
				3154	<p>This method traverses the use list of a <tt>Value</tt> changing all <a
				3155	href="#User"><tt>User</tt>s</a> of the current value to refer to
				3156	"<tt>V</tt>" instead. For example, if you detect that an instruction always
				3157	produces a constant value (for example through constant folding), you can
				3158	replace all uses of the instruction with the constant like this:</p>
				3159
				3160	<div class="doc_code">
				3161	<pre>
				3162	Inst->replaceAllUsesWith(ConstVal);
				3163	</pre>
				3164	</div>
				3165
				3166	</ul>
				3167
				3168	</div>
				3169
				3170	<!-- ======================================================================= -->
				3171	<div class="doc_subsection">
				3172	<a name="User">The <tt>User</tt> class</a>
				3173	</div>
				3174
				3175	<div class="doc_text">
				3176
				3177	<p>
				3178	<tt>#include "<a href="/doxygen/User_8h-source.html">llvm/User.h</a>"</tt><br>
				3179	doxygen info: <a href="/doxygen/classllvm_1_1User.html">User Class</a><br>
				3180	Superclass: <a href="#Value"><tt>Value</tt></a></p>
				3181
				3182	<p>The <tt>User</tt> class is the common base class of all LLVM nodes that may
				3183	refer to <a href="#Value"><tt>Value</tt></a>s. It exposes a list of "Operands"
				3184	that are all of the <a href="#Value"><tt>Value</tt></a>s that the User is
				3185	referring to. The <tt>User</tt> class itself is a subclass of
				3186	<tt>Value</tt>.</p>
				3187
				3188	<p>The operands of a <tt>User</tt> point directly to the LLVM <a
				3189	href="#Value"><tt>Value</tt></a> that it refers to. Because LLVM uses Static
				3190	Single Assignment (SSA) form, there can only be one definition referred to,
				3191	allowing this direct connection. This connection provides the use-def
				3192	information in LLVM.</p>
				3193
				3194	</div>
				3195
				3196	<!-- _______________________________________________________________________ -->
				3197	<div class="doc_subsubsection">
				3198	<a name="m_User">Important Public Members of the <tt>User</tt> class</a>
				3199	</div>
				3200
				3201	<div class="doc_text">
				3202
				3203	<p>The <tt>User</tt> class exposes the operand list in two ways: through
				3204	an index access interface and through an iterator based interface.</p>
				3205
				3206	<ul>
				3207	<li><tt>Value *getOperand(unsigned i)</tt><br>
				3208	<tt>unsigned getNumOperands()</tt>
				3209	<p> These two methods expose the operands of the <tt>User</tt> in a
				3210	convenient form for direct access.</p></li>
				3211
				3212	<li><tt>User::op_iterator</tt> - Typedef for iterator over the operand
				3213	list<br>
				3214	<tt>op_iterator op_begin()</tt> - Get an iterator to the start of
				3215	the operand list.<br>
				3216	<tt>op_iterator op_end()</tt> - Get an iterator to the end of the
				3217	operand list.
				3218	<p> Together, these methods make up the iterator based interface to
				3219	the operands of a <tt>User</tt>.</p></li>
				3220	</ul>
				3221
				3222	</div>
				3223
				3224	<!-- ======================================================================= -->
				3225	<div class="doc_subsection">
				3226	<a name="Instruction">The <tt>Instruction</tt> class</a>
				3227	</div>
				3228
				3229	<div class="doc_text">
				3230
				3231	<p><tt>#include "</tt><tt><a
				3232	href="/doxygen/Instruction_8h-source.html">llvm/Instruction.h</a>"</tt><br>
				3233	doxygen info: <a href="/doxygen/classllvm_1_1Instruction.html">Instruction Class</a><br>
				3234	Superclasses: <a href="#User"><tt>User</tt></a>, <a
				3235	href="#Value"><tt>Value</tt></a></p>
				3236
				3237	<p>The <tt>Instruction</tt> class is the common base class for all LLVM
				3238	instructions. It provides only a few methods, but is a very commonly used
				3239	class. The primary data tracked by the <tt>Instruction</tt> class itself is the
				3240	opcode (instruction type) and the parent <a
				3241	href="#BasicBlock"><tt>BasicBlock</tt></a> the <tt>Instruction</tt> is embedded
				3242	into. To represent a specific type of instruction, one of many subclasses of
				3243	<tt>Instruction</tt> are used.</p>
				3244
				3245	<p> Because the <tt>Instruction</tt> class subclasses the <a
				3246	href="#User"><tt>User</tt></a> class, its operands can be accessed in the same
				3247	way as for other <a href="#User"><tt>User</tt></a>s (with the
				3248	<tt>getOperand()</tt>/<tt>getNumOperands()</tt> and
				3249	<tt>op_begin()</tt>/<tt>op_end()</tt> methods).</p> <p> An important file for
				3250	the <tt>Instruction</tt> class is the <tt>llvm/Instruction.def</tt> file. This
				3251	file contains some meta-data about the various different types of instructions
				3252	in LLVM. It describes the enum values that are used as opcodes (for example
				3253	<tt>Instruction::Add</tt> and <tt>Instruction::ICmp</tt>), as well as the
				3254	concrete sub-classes of <tt>Instruction</tt> that implement the instruction (for
				3255	example <tt><a href="#BinaryOperator">BinaryOperator</a></tt> and <tt><a
				3256	href="#CmpInst">CmpInst</a></tt>). Unfortunately, the use of macros in
				3257	this file confuses doxygen, so these enum values don't show up correctly in the
				3258	<a href="/doxygen/classllvm_1_1Instruction.html">doxygen output</a>.</p>
				3259
				3260	</div>
				3261
				3262	<!-- _______________________________________________________________________ -->
				3263	<div class="doc_subsubsection">
				3264	<a name="s_Instruction">Important Subclasses of the <tt>Instruction</tt>
				3265	class</a>
				3266	</div>
				3267	<div class="doc_text">
				3268	<ul>
				3269	<li><tt><a name="BinaryOperator">BinaryOperator</a></tt>
				3270	<p>This subclasses represents all two operand instructions whose operands
				3271	must be the same type, except for the comparison instructions.</p></li>
				3272	<li><tt><a name="CastInst">CastInst</a></tt>
				3273	<p>This subclass is the parent of the 12 casting instructions. It provides
				3274	common operations on cast instructions.</p>
				3275	<li><tt><a name="CmpInst">CmpInst</a></tt>
				3276	<p>This subclass respresents the two comparison instructions,
				3277	<a href="LangRef.html#i_icmp">ICmpInst</a> (integer opreands), and
				3278	<a href="LangRef.html#i_fcmp">FCmpInst</a> (floating point operands).</p>
				3279	<li><tt><a name="TerminatorInst">TerminatorInst</a></tt>
				3280	<p>This subclass is the parent of all terminator instructions (those which
				3281	can terminate a block).</p>
				3282	</ul>
				3283	</div>
				3284
				3285	<!-- _______________________________________________________________________ -->
				3286	<div class="doc_subsubsection">
				3287	<a name="m_Instruction">Important Public Members of the <tt>Instruction</tt>
				3288	class</a>
				3289	</div>
				3290
				3291	<div class="doc_text">
				3292
				3293	<ul>
				3294	<li><tt><a href="#BasicBlock">BasicBlock</a> *getParent()</tt>
				3295	<p>Returns the <a href="#BasicBlock"><tt>BasicBlock</tt></a> that
				3296	this <tt>Instruction</tt> is embedded into.</p></li>
				3297	<li><tt>bool mayWriteToMemory()</tt>
				3298	<p>Returns true if the instruction writes to memory, i.e. it is a
				3299	<tt>call</tt>,<tt>free</tt>,<tt>invoke</tt>, or <tt>store</tt>.</p></li>
				3300	<li><tt>unsigned getOpcode()</tt>
				3301	<p>Returns the opcode for the <tt>Instruction</tt>.</p></li>
				3302	<li><tt><a href="#Instruction">Instruction</a> *clone() const</tt>
				3303	<p>Returns another instance of the specified instruction, identical
				3304	in all ways to the original except that the instruction has no parent
				3305	(ie it's not embedded into a <a href="#BasicBlock"><tt>BasicBlock</tt></a>),
				3306	and it has no name</p></li>
				3307	</ul>
				3308
				3309	</div>
				3310
				3311	<!-- ======================================================================= -->
				3312	<div class="doc_subsection">
				3313	<a name="Constant">The <tt>Constant</tt> class and subclasses</a>
				3314	</div>
				3315
				3316	<div class="doc_text">
				3317
				3318	<p>Constant represents a base class for different types of constants. It
				3319	is subclassed by ConstantInt, ConstantArray, etc. for representing
				3320	the various types of Constants. <a href="#GlobalValue">GlobalValue</a> is also
				3321	a subclass, which represents the address of a global variable or function.
				3322	</p>
				3323
				3324	</div>
				3325
				3326	<!-- _______________________________________________________________________ -->
				3327	<div class="doc_subsubsection">Important Subclasses of Constant </div>
				3328	<div class="doc_text">
				3329	<ul>
				3330	<li>ConstantInt : This subclass of Constant represents an integer constant of
				3331	any width.
				3332	<ul>
				3333	<li><tt>const APInt& getValue() const</tt>: Returns the underlying
				3334	value of this constant, an APInt value.</li>
				3335	<li><tt>int64_t getSExtValue() const</tt>: Converts the underlying APInt
				3336	value to an int64_t via sign extension. If the value (not the bit width)
				3337	of the APInt is too large to fit in an int64_t, an assertion will result.
				3338	For this reason, use of this method is discouraged.</li>
				3339	<li><tt>uint64_t getZExtValue() const</tt>: Converts the underlying APInt
				3340	value to a uint64_t via zero extension. IF the value (not the bit width)
				3341	of the APInt is too large to fit in a uint64_t, an assertion will result.
				3342	For this reason, use of this method is discouraged.</li>
				3343	<li><tt>static ConstantInt* get(const APInt& Val)</tt>: Returns the
				3344	ConstantInt object that represents the value provided by <tt>Val</tt>.
				3345	The type is implied as the IntegerType that corresponds to the bit width
				3346	of <tt>Val</tt>.</li>
				3347	<li><tt>static ConstantInt* get(const Type *Ty, uint64_t Val)</tt>:
				3348	Returns the ConstantInt object that represents the value provided by
				3349	<tt>Val</tt> for integer type <tt>Ty</tt>.</li>
				3350	</ul>
				3351	</li>
				3352	<li>ConstantFP : This class represents a floating point constant.
				3353	<ul>
				3354	<li><tt>double getValue() const</tt>: Returns the underlying value of
				3355	this constant. </li>
				3356	</ul>
				3357	</li>
				3358	<li>ConstantArray : This represents a constant array.
				3359	<ul>
				3360	<li><tt>const std::vector<Use> &getValues() const</tt>: Returns
				3361	a vector of component constants that makeup this array. </li>
				3362	</ul>
				3363	</li>
				3364	<li>ConstantStruct : This represents a constant struct.
				3365	<ul>
				3366	<li><tt>const std::vector<Use> &getValues() const</tt>: Returns
				3367	a vector of component constants that makeup this array. </li>
				3368	</ul>
				3369	</li>
				3370	<li>GlobalValue : This represents either a global variable or a function. In
				3371	either case, the value is a constant fixed address (after linking).
				3372	</li>
				3373	</ul>
				3374	</div>
				3375
				3376
				3377	<!-- ======================================================================= -->
				3378	<div class="doc_subsection">
				3379	<a name="GlobalValue">The <tt>GlobalValue</tt> class</a>
				3380	</div>
				3381
				3382	<div class="doc_text">
				3383
				3384	<p><tt>#include "<a
				3385	href="/doxygen/GlobalValue_8h-source.html">llvm/GlobalValue.h</a>"</tt><br>
				3386	doxygen info: <a href="/doxygen/classllvm_1_1GlobalValue.html">GlobalValue
				3387	Class</a><br>
				3388	Superclasses: <a href="#Constant"><tt>Constant</tt></a>,
				3389	<a href="#User"><tt>User</tt></a>, <a href="#Value"><tt>Value</tt></a></p>
				3390
				3391	<p>Global values (<a href="#GlobalVariable"><tt>GlobalVariable</tt></a>s or <a
				3392	href="#Function"><tt>Function</tt></a>s) are the only LLVM values that are
				3393	visible in the bodies of all <a href="#Function"><tt>Function</tt></a>s.
				3394	Because they are visible at global scope, they are also subject to linking with
				3395	other globals defined in different translation units. To control the linking
				3396	process, <tt>GlobalValue</tt>s know their linkage rules. Specifically,
				3397	<tt>GlobalValue</tt>s know whether they have internal or external linkage, as
				3398	defined by the <tt>LinkageTypes</tt> enumeration.</p>
				3399
				3400	<p>If a <tt>GlobalValue</tt> has internal linkage (equivalent to being
				3401	<tt>static</tt> in C), it is not visible to code outside the current translation
				3402	unit, and does not participate in linking. If it has external linkage, it is
				3403	visible to external code, and does participate in linking. In addition to
				3404	linkage information, <tt>GlobalValue</tt>s keep track of which <a
				3405	href="#Module"><tt>Module</tt></a> they are currently part of.</p>
				3406
				3407	<p>Because <tt>GlobalValue</tt>s are memory objects, they are always referred to
				3408	by their <b>address</b>. As such, the <a href="#Type"><tt>Type</tt></a> of a
				3409	global is always a pointer to its contents. It is important to remember this
				3410	when using the <tt>GetElementPtrInst</tt> instruction because this pointer must
				3411	be dereferenced first. For example, if you have a <tt>GlobalVariable</tt> (a
				3412	subclass of <tt>GlobalValue)</tt> that is an array of 24 ints, type <tt>[24 x
				3413	i32]</tt>, then the <tt>GlobalVariable</tt> is a pointer to that array. Although
				3414	the address of the first element of this array and the value of the
				3415	<tt>GlobalVariable</tt> are the same, they have different types. The
				3416	<tt>GlobalVariable</tt>'s type is <tt>[24 x i32]</tt>. The first element's type
				3417	is <tt>i32.</tt> Because of this, accessing a global value requires you to
				3418	dereference the pointer with <tt>GetElementPtrInst</tt> first, then its elements
				3419	can be accessed. This is explained in the <a href="LangRef.html#globalvars">LLVM
				3420	Language Reference Manual</a>.</p>
				3421
				3422	</div>
				3423
				3424	<!-- _______________________________________________________________________ -->
				3425	<div class="doc_subsubsection">
				3426	<a name="m_GlobalValue">Important Public Members of the <tt>GlobalValue</tt>
				3427	class</a>
				3428	</div>
				3429
				3430	<div class="doc_text">
				3431
				3432	<ul>
				3433	<li><tt>bool hasInternalLinkage() const</tt><br>
				3434	<tt>bool hasExternalLinkage() const</tt><br>
				3435	<tt>void setInternalLinkage(bool HasInternalLinkage)</tt>
				3436	<p> These methods manipulate the linkage characteristics of the <tt>GlobalValue</tt>.</p>
				3437	<p> </p>
				3438	</li>
				3439	<li><tt><a href="#Module">Module</a> *getParent()</tt>
				3440	<p> This returns the <a href="#Module"><tt>Module</tt></a> that the
				3441	GlobalValue is currently embedded into.</p></li>
				3442	</ul>
				3443
				3444	</div>
				3445
				3446	<!-- ======================================================================= -->
				3447	<div class="doc_subsection">
				3448	<a name="Function">The <tt>Function</tt> class</a>
				3449	</div>
				3450
				3451	<div class="doc_text">
				3452
				3453	<p><tt>#include "<a
				3454	href="/doxygen/Function_8h-source.html">llvm/Function.h</a>"</tt><br> doxygen
				3455	info: <a href="/doxygen/classllvm_1_1Function.html">Function Class</a><br>
				3456	Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
				3457	<a href="#Constant"><tt>Constant</tt></a>,
				3458	<a href="#User"><tt>User</tt></a>,
				3459	<a href="#Value"><tt>Value</tt></a></p>
				3460
				3461	<p>The <tt>Function</tt> class represents a single procedure in LLVM. It is
				3462	actually one of the more complex classes in the LLVM heirarchy because it must
				3463	keep track of a large amount of data. The <tt>Function</tt> class keeps track
				3464	of a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, a list of formal
				3465	<a href="#Argument"><tt>Argument</tt></a>s, and a
				3466	<a href="#SymbolTable"><tt>SymbolTable</tt></a>.</p>
				3467
				3468	<p>The list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s is the most
				3469	commonly used part of <tt>Function</tt> objects. The list imposes an implicit
				3470	ordering of the blocks in the function, which indicate how the code will be
				3471	layed out by the backend. Additionally, the first <a
				3472	href="#BasicBlock"><tt>BasicBlock</tt></a> is the implicit entry node for the
				3473	<tt>Function</tt>. It is not legal in LLVM to explicitly branch to this initial
				3474	block. There are no implicit exit nodes, and in fact there may be multiple exit
				3475	nodes from a single <tt>Function</tt>. If the <a
				3476	href="#BasicBlock"><tt>BasicBlock</tt></a> list is empty, this indicates that
				3477	the <tt>Function</tt> is actually a function declaration: the actual body of the
				3478	function hasn't been linked in yet.</p>
				3479
				3480	<p>In addition to a list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s, the
				3481	<tt>Function</tt> class also keeps track of the list of formal <a
				3482	href="#Argument"><tt>Argument</tt></a>s that the function receives. This
				3483	container manages the lifetime of the <a href="#Argument"><tt>Argument</tt></a>
				3484	nodes, just like the <a href="#BasicBlock"><tt>BasicBlock</tt></a> list does for
				3485	the <a href="#BasicBlock"><tt>BasicBlock</tt></a>s.</p>
				3486
				3487	<p>The <a href="#SymbolTable"><tt>SymbolTable</tt></a> is a very rarely used
				3488	LLVM feature that is only used when you have to look up a value by name. Aside
				3489	from that, the <a href="#SymbolTable"><tt>SymbolTable</tt></a> is used
				3490	internally to make sure that there are not conflicts between the names of <a
				3491	href="#Instruction"><tt>Instruction</tt></a>s, <a
				3492	href="#BasicBlock"><tt>BasicBlock</tt></a>s, or <a
				3493	href="#Argument"><tt>Argument</tt></a>s in the function body.</p>
				3494
				3495	<p>Note that <tt>Function</tt> is a <a href="#GlobalValue">GlobalValue</a>
				3496	and therefore also a <a href="#Constant">Constant</a>. The value of the function
				3497	is its address (after linking) which is guaranteed to be constant.</p>
				3498	</div>
				3499
				3500	<!-- _______________________________________________________________________ -->
				3501	<div class="doc_subsubsection">
				3502	<a name="m_Function">Important Public Members of the <tt>Function</tt>
				3503	class</a>
				3504	</div>
				3505
				3506	<div class="doc_text">
				3507
				3508	<ul>
				3509	<li><tt>Function(const </tt><tt><a href="#FunctionType">FunctionType</a>
				3510	Ty, LinkageTypes Linkage, const std::string &N = "", Module Parent = 0)</tt>
				3511
				3512	<p>Constructor used when you need to create new <tt>Function</tt>s to add
				3513	the the program. The constructor must specify the type of the function to
				3514	create and what type of linkage the function should have. The <a
				3515	href="#FunctionType"><tt>FunctionType</tt></a> argument
				3516	specifies the formal arguments and return value for the function. The same
				3517	<a href="#FunctionType"><tt>FunctionType</tt></a> value can be used to
				3518	create multiple functions. The <tt>Parent</tt> argument specifies the Module
				3519	in which the function is defined. If this argument is provided, the function
				3520	will automatically be inserted into that module's list of
				3521	functions.</p></li>
				3522
Chris Lattner	5e70957	2008-11-25 18:34:50 +0000	[diff] [blame]	3523	<li><tt>bool isDeclaration()</tt>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3524
				3525	<p>Return whether or not the <tt>Function</tt> has a body defined. If the
				3526	function is "external", it does not have a body, and thus must be resolved
				3527	by linking with a function defined in a different translation unit.</p></li>
				3528
				3529	<li><tt>Function::iterator</tt> - Typedef for basic block list iterator<br>
				3530	<tt>Function::const_iterator</tt> - Typedef for const_iterator.<br>
				3531
				3532	<tt>begin()</tt>, <tt>end()</tt>
				3533	<tt>size()</tt>, <tt>empty()</tt>
				3534
				3535	<p>These are forwarding methods that make it easy to access the contents of
				3536	a <tt>Function</tt> object's <a href="#BasicBlock"><tt>BasicBlock</tt></a>
				3537	list.</p></li>
				3538
				3539	<li><tt>Function::BasicBlockListType &getBasicBlockList()</tt>
				3540
				3541	<p>Returns the list of <a href="#BasicBlock"><tt>BasicBlock</tt></a>s. This
				3542	is necessary to use when you need to update the list or perform a complex
				3543	action that doesn't have a forwarding method.</p></li>
				3544
				3545	<li><tt>Function::arg_iterator</tt> - Typedef for the argument list
				3546	iterator<br>
				3547	<tt>Function::const_arg_iterator</tt> - Typedef for const_iterator.<br>
				3548
				3549	<tt>arg_begin()</tt>, <tt>arg_end()</tt>
				3550	<tt>arg_size()</tt>, <tt>arg_empty()</tt>
				3551
				3552	<p>These are forwarding methods that make it easy to access the contents of
				3553	a <tt>Function</tt> object's <a href="#Argument"><tt>Argument</tt></a>
				3554	list.</p></li>
				3555
				3556	<li><tt>Function::ArgumentListType &getArgumentList()</tt>
				3557
				3558	<p>Returns the list of <a href="#Argument"><tt>Argument</tt></a>s. This is
				3559	necessary to use when you need to update the list or perform a complex
				3560	action that doesn't have a forwarding method.</p></li>
				3561
				3562	<li><tt><a href="#BasicBlock">BasicBlock</a> &getEntryBlock()</tt>
				3563
				3564	<p>Returns the entry <a href="#BasicBlock"><tt>BasicBlock</tt></a> for the
				3565	function. Because the entry block for the function is always the first
				3566	block, this returns the first block of the <tt>Function</tt>.</p></li>
				3567
				3568	<li><tt><a href="#Type">Type</a> *getReturnType()</tt><br>
				3569	<tt><a href="#FunctionType">FunctionType</a> *getFunctionType()</tt>
				3570
				3571	<p>This traverses the <a href="#Type"><tt>Type</tt></a> of the
				3572	<tt>Function</tt> and returns the return type of the function, or the <a
				3573	href="#FunctionType"><tt>FunctionType</tt></a> of the actual
				3574	function.</p></li>
				3575
				3576	<li><tt><a href="#SymbolTable">SymbolTable</a> *getSymbolTable()</tt>
				3577
				3578	<p> Return a pointer to the <a href="#SymbolTable"><tt>SymbolTable</tt></a>
				3579	for this <tt>Function</tt>.</p></li>
				3580	</ul>
				3581
				3582	</div>
				3583
				3584	<!-- ======================================================================= -->
				3585	<div class="doc_subsection">
				3586	<a name="GlobalVariable">The <tt>GlobalVariable</tt> class</a>
				3587	</div>
				3588
				3589	<div class="doc_text">
				3590
				3591	<p><tt>#include "<a
				3592	href="/doxygen/GlobalVariable_8h-source.html">llvm/GlobalVariable.h</a>"</tt>
				3593	<br>
				3594	doxygen info: <a href="/doxygen/classllvm_1_1GlobalVariable.html">GlobalVariable
				3595	Class</a><br>
				3596	Superclasses: <a href="#GlobalValue"><tt>GlobalValue</tt></a>,
				3597	<a href="#Constant"><tt>Constant</tt></a>,
				3598	<a href="#User"><tt>User</tt></a>,
				3599	<a href="#Value"><tt>Value</tt></a></p>
				3600
				3601	<p>Global variables are represented with the (suprise suprise)
				3602	<tt>GlobalVariable</tt> class. Like functions, <tt>GlobalVariable</tt>s are also
				3603	subclasses of <a href="#GlobalValue"><tt>GlobalValue</tt></a>, and as such are
				3604	always referenced by their address (global values must live in memory, so their
				3605	"name" refers to their constant address). See
				3606	<a href="#GlobalValue"><tt>GlobalValue</tt></a> for more on this. Global
				3607	variables may have an initial value (which must be a
				3608	<a href="#Constant"><tt>Constant</tt></a>), and if they have an initializer,
				3609	they may be marked as "constant" themselves (indicating that their contents
				3610	never change at runtime).</p>
				3611	</div>
				3612
				3613	<!-- _______________________________________________________________________ -->
				3614	<div class="doc_subsubsection">
				3615	<a name="m_GlobalVariable">Important Public Members of the
				3616	<tt>GlobalVariable</tt> class</a>
				3617	</div>
				3618
				3619	<div class="doc_text">
				3620
				3621	<ul>
				3622	<li><tt>GlobalVariable(const </tt><tt><a href="#Type">Type</a> *Ty, bool
				3623	isConstant, LinkageTypes& Linkage, <a href="#Constant">Constant</a>
				3624	Initializer = 0, const std::string &Name = "", Module Parent = 0)</tt>
				3625
				3626	<p>Create a new global variable of the specified type. If
				3627	<tt>isConstant</tt> is true then the global variable will be marked as
				3628	unchanging for the program. The Linkage parameter specifies the type of
Duncan Sands	19d161f	2009-03-07 15:45:40 +0000	[diff] [blame]	3629	linkage (internal, external, weak, linkonce, appending) for the variable.
				3630	If the linkage is InternalLinkage, WeakAnyLinkage, WeakODRLinkage,
				3631	LinkOnceAnyLinkage or LinkOnceODRLinkage,  then the resultant
				3632	global variable will have internal linkage. AppendingLinkage concatenates
				3633	together all instances (in different translation units) of the variable
				3634	into a single variable but is only applicable to arrays.  See
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3635	the <a href="LangRef.html#modulestructure">LLVM Language Reference</a> for
				3636	further details on linkage types. Optionally an initializer, a name, and the
				3637	module to put the variable into may be specified for the global variable as
				3638	well.</p></li>
				3639
				3640	<li><tt>bool isConstant() const</tt>
				3641
				3642	<p>Returns true if this is a global variable that is known not to
				3643	be modified at runtime.</p></li>
				3644
				3645	<li><tt>bool hasInitializer()</tt>
				3646
				3647	<p>Returns true if this <tt>GlobalVariable</tt> has an intializer.</p></li>
				3648
				3649	<li><tt><a href="#Constant">Constant</a> *getInitializer()</tt>
				3650
				3651	<p>Returns the intial value for a <tt>GlobalVariable</tt>. It is not legal
				3652	to call this method if there is no initializer.</p></li>
				3653	</ul>
				3654
				3655	</div>
				3656
				3657
				3658	<!-- ======================================================================= -->
				3659	<div class="doc_subsection">
				3660	<a name="BasicBlock">The <tt>BasicBlock</tt> class</a>
				3661	</div>
				3662
				3663	<div class="doc_text">
				3664
				3665	<p><tt>#include "<a
				3666	href="/doxygen/BasicBlock_8h-source.html">llvm/BasicBlock.h</a>"</tt><br>
Stefanus Du Toit	f017c85	2009-06-17 21:12:26 +0000	[diff] [blame]	3667	doxygen info: <a href="/doxygen/classllvm_1_1BasicBlock.html">BasicBlock
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3668	Class</a><br>
				3669	Superclass: <a href="#Value"><tt>Value</tt></a></p>
				3670
				3671	<p>This class represents a single entry multiple exit section of the code,
				3672	commonly known as a basic block by the compiler community. The
				3673	<tt>BasicBlock</tt> class maintains a list of <a
				3674	href="#Instruction"><tt>Instruction</tt></a>s, which form the body of the block.
				3675	Matching the language definition, the last element of this list of instructions
				3676	is always a terminator instruction (a subclass of the <a
				3677	href="#TerminatorInst"><tt>TerminatorInst</tt></a> class).</p>
				3678
				3679	<p>In addition to tracking the list of instructions that make up the block, the
				3680	<tt>BasicBlock</tt> class also keeps track of the <a
				3681	href="#Function"><tt>Function</tt></a> that it is embedded into.</p>
				3682
				3683	<p>Note that <tt>BasicBlock</tt>s themselves are <a
				3684	href="#Value"><tt>Value</tt></a>s, because they are referenced by instructions
				3685	like branches and can go in the switch tables. <tt>BasicBlock</tt>s have type
				3686	<tt>label</tt>.</p>
				3687
				3688	</div>
				3689
				3690	<!-- _______________________________________________________________________ -->
				3691	<div class="doc_subsubsection">
				3692	<a name="m_BasicBlock">Important Public Members of the <tt>BasicBlock</tt>
				3693	class</a>
				3694	</div>
				3695
				3696	<div class="doc_text">
				3697	<ul>
				3698
				3699	<li><tt>BasicBlock(const std::string &Name = "", </tt><tt><a
				3700	href="#Function">Function</a> *Parent = 0)</tt>
				3701
				3702	<p>The <tt>BasicBlock</tt> constructor is used to create new basic blocks for
				3703	insertion into a function. The constructor optionally takes a name for the new
				3704	block, and a <a href="#Function"><tt>Function</tt></a> to insert it into. If
				3705	the <tt>Parent</tt> parameter is specified, the new <tt>BasicBlock</tt> is
				3706	automatically inserted at the end of the specified <a
				3707	href="#Function"><tt>Function</tt></a>, if not specified, the BasicBlock must be
				3708	manually inserted into the <a href="#Function"><tt>Function</tt></a>.</p></li>
				3709
				3710	<li><tt>BasicBlock::iterator</tt> - Typedef for instruction list iterator<br>
				3711	<tt>BasicBlock::const_iterator</tt> - Typedef for const_iterator.<br>
				3712	<tt>begin()</tt>, <tt>end()</tt>, <tt>front()</tt>, <tt>back()</tt>,
				3713	<tt>size()</tt>, <tt>empty()</tt>
				3714	STL-style functions for accessing the instruction list.
				3715
				3716	<p>These methods and typedefs are forwarding functions that have the same
				3717	semantics as the standard library methods of the same names. These methods
				3718	expose the underlying instruction list of a basic block in a way that is easy to
				3719	manipulate. To get the full complement of container operations (including
				3720	operations to update the list), you must use the <tt>getInstList()</tt>
				3721	method.</p></li>
				3722
				3723	<li><tt>BasicBlock::InstListType &getInstList()</tt>
				3724
				3725	<p>This method is used to get access to the underlying container that actually
				3726	holds the Instructions. This method must be used when there isn't a forwarding
				3727	function in the <tt>BasicBlock</tt> class for the operation that you would like
				3728	to perform. Because there are no forwarding functions for "updating"
				3729	operations, you need to use this if you want to update the contents of a
				3730	<tt>BasicBlock</tt>.</p></li>
				3731
				3732	<li><tt><a href="#Function">Function</a> *getParent()</tt>
				3733
				3734	<p> Returns a pointer to <a href="#Function"><tt>Function</tt></a> the block is
				3735	embedded into, or a null pointer if it is homeless.</p></li>
				3736
				3737	<li><tt><a href="#TerminatorInst">TerminatorInst</a> *getTerminator()</tt>
				3738
				3739	<p> Returns a pointer to the terminator instruction that appears at the end of
				3740	the <tt>BasicBlock</tt>. If there is no terminator instruction, or if the last
				3741	instruction in the block is not a terminator, then a null pointer is
				3742	returned.</p></li>
				3743
				3744	</ul>
				3745
				3746	</div>
				3747
				3748
				3749	<!-- ======================================================================= -->
				3750	<div class="doc_subsection">
				3751	<a name="Argument">The <tt>Argument</tt> class</a>
				3752	</div>
				3753
				3754	<div class="doc_text">
				3755
				3756	<p>This subclass of Value defines the interface for incoming formal
				3757	arguments to a function. A Function maintains a list of its formal
				3758	arguments. An argument has a pointer to the parent Function.</p>
				3759
				3760	</div>
				3761
				3762	<!-- *********************************************************************** -->
				3763	<hr>
				3764	<address>
				3765	<a href="http://jigsaw.w3.org/css-validator/check/referer"><img
Misha Brukman	947321d	2008-12-11 17:34:48 +0000	[diff] [blame]	3766	src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3767	<a href="http://validator.w3.org/check/referer"><img
Gabor Greif	7dabe38	2008-06-18 14:05:31 +0000	[diff] [blame]	3768	src="http://www.w3.org/Icons/valid-html401" alt="Valid HTML 4.01 Strict"></a>
Dan Gohman	f17a25c	2007-07-18 16:29:46 +0000	[diff] [blame]	3769
				3770	<a href="mailto:dhurjati@cs.uiuc.edu">Dinakar Dhurjati</a> and
				3771	<a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
				3772	<a href="http://llvm.org">The LLVM Compiler Infrastructure</a><br>
				3773	Last modified: $Date$
				3774	</address>
				3775
				3776	</body>
				3777	</html>